YT YouTube2Text

YouTube2Text → Transcripts → Digital Economic Forum: AI and the Ghana Economy.

Digital Economic Forum: AI and the Ghana Economy. — Transcript

by JoyNews · 29,388 words · 4,792 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00shaping Ghana's digital economy.
0:02Tonight's edition is titled AI and the
0:05Ghana economy, opportunity or
0:08disruption. And I'm happy to have with
0:11us exploring this topic a lively studio
0:14audience from across the board.
0:17>> [music]
0:17>> We have AI skeptics, farmers deploying
0:20AI, and students fearing for their
0:22future because of the potential for AI
0:26to wipe out the jobs they are training
0:28for. So, ladies and gentlemen, thank you
0:31for taking time to join us. And I want
0:33to invite [music] our studio audience
0:35with a round of applause. I'm grateful
0:37>> [applause]
0:38>> that you could make time to join us.
0:40Now, we begin with a documentary that
0:43examines this question through the
0:45experiences, perspectives, and concerns
0:49of people already engaging with
0:52artificial intelligence.
0:54After the documentary, we'll hear
0:56initial reactions, have a short
0:58analytical presentation from Bright
1:00Simons,
1:01and bring together our panel for a
1:03deeper discussion. We will also invite
1:06questions from our lively studio
1:08audience. The aim is not to arrive at
1:11easy answers. It is to understand the
1:14choices before us, the risk we must
1:16manage, and the opportunities we cannot
1:19afford to miss. So, let's begin with the
1:21documentary. It explores one central
1:23question.
1:25Will artificial intelligence create a
1:27new economy, new opportunities for our
1:30economy, or will it disrupt the
1:32livelihoods, businesses, and
1:33institutions we depend on?
1:36Watch.
1:38>> It's another dreaded budget day, Kofi.
1:41The media budget review and the rest of
1:42the team preparing for PM Express would
1:45have to endure just to turn around a
1:48great research to piece
1:50by 9:00 p.m.
1:51>> So, you've got about over 500
1:53paragraphs, and you have to do some
1:55comparative analysis
1:57>> [snorts]
1:58>> to make sure that [music] everything
1:59that has been said, the fact-check
2:00element, you know, putting everything
2:03into proper perspective, the graphic
2:05visualization. Also, do your additional
2:09research on top of that, so you can ask
2:12relevant questions by 9:00 p.m. So,
2:14you're talking about sometimes 500 pages
2:17with appendices that five people poring
2:21over this document, reading word for
2:23word, you know, the way I I like to do
2:24my thing. I had to circle it and
2:26underline it, double-check, call
2:28somebody. And once you're doing that,
2:31you then have to cross-reference with
2:32previous budgets.
2:33>> analysis part is is the most difficult
2:36one.
2:37And sometimes you you get a a certain
2:40year's figure you want to compare to the
2:41previous year, you need to do the
2:43calculation manually.
2:44>> So, you're possibly doing 1,500 pages
2:47of analysis. You know, this is the
2:49research
2:49>> You also refine it, you know,
2:51presentable, visualize it for people to
2:53understand it. So, it's it's not just
2:56the text in the budget, but it's how you
2:58also present the information you get
3:00to your audience. This is a difficult
3:02work.
3:03>> And we can sit for hours just trying
3:05>> It's a difficult
3:05>> trying to do that. The keyword search,
3:07you know, sometimes you search it in
3:09your PDF and you see about, you know,
3:11400 results and you're trying to go
3:13through each and every one to see where
3:15exactly what you're looking for.
3:17>> We had to expand this unit because this
3:19unit the work was enormous. It was just
3:22note-taking,
3:24reading yourself 500 pages. And this is
3:28where the AI conversation becomes
3:30interesting.
3:32So, the job is done. Yeah, Kofi.
3:34And you can see here it's fast. It looks
3:38very efficient.
3:40And it's also producing the slides for
3:42us. And he's already telling you the
3:43interview brief, Minister of Finance,
3:45Dr. I was going to give me a brief on
3:47what to ask. And actually get all your
3:49stories also from
3:49>> Everything from here.
3:50>> Indeed, this
3:52is excellent.
3:53Um no human being could have done this
3:55in 10 minutes actually
3:57or less.
3:59Beautiful.
4:03To understand what this technology
4:05actually is,
4:06we went to a newsroom in Accra.
4:09>> Good afternoon. [music]
4:10I'm Margaret Tobin and you're watching
4:13News in Brief. From shifting economic
4:16policies to the zone 2025 Bulleting
4:19introduce Margaret Tobin, Ghana's first
4:21ever AI improved news custom. [music]
4:24We saw it as a need because
4:27the newsroom is under so much pressure.
4:30Imagine a newsroom
4:32in the afternoon everybody is out on
4:34assignment
4:35>> [music]
4:35>> and you have nobody to read the news.
4:36That's where Margaret steps in. So,
4:38Margaret is fed with the stories. It's a
4:41blend of AI and human intervention. So,
4:43Margaret is prepped for every bulletin
4:45[music]
4:46and everything she says goes through the
4:48right channels. Just like any other
4:50reader does.
4:51>> A machine that reads the news prepped
4:54like a colleague giving a name corrected
4:57[music] when she gets it wrong.
5:00This is what people mean when they say
5:02artificial intelligence.
5:06But if you strip away the noise, what is
5:08this thing actually?
5:10>> Mostly when [music] we say AI in
5:12academics, AI is just a system. It's a
5:15mathematical model.
5:16>> [music]
5:16>> Everything that AI is doing
5:18is statistics and probability.
5:20So, it's just learning data,
5:23learning patterns in the data,
5:25and trying to
5:27use what it has learned to make a
5:29predictions. [music]
5:30Look at an apprentice that goes to
5:33Makola. From the beginning, the
5:34apprentice knows [music] nothing.
5:36The apprentice cannot tell
5:38which of these tomatoes will go [music]
5:40rotten.
5:41Which of them will sell.
5:42But with time, when you follow the
5:45market [music] women, you deal with a
5:46lot of tomato baskets.
5:48They begin to build that instance. So,
5:51that is what
5:52AI is.
5:53Nothing
5:55fancy, nothing hidden.
5:57It's just learning from millions of
5:59data,
6:00understanding the pattern within the
6:02data, and then just make a prediction.
6:04Learning from what happened before to
6:07predict what happens next, [music] which
6:10means AI is only ever as good as what we
6:14feed it.
6:15>> AI is both the current [music] and the
6:16future. There's
6:18a huge economic, you know, linkage to
6:21what AI could provide [music] to Africa
6:23and Ghana. Africa is also a lot behind
6:25when it comes to economic development.
6:27This also offers us an opportunity to
6:29really catch up.
6:30>> And it is already [music] here. Not in
6:33the lab,
6:34in a field.
6:35>> FarmSense is our flagship product, and
6:37what we are trying to do with this
6:38product [music] is to give more
6:40intelligence to the farmers, more
6:42insights into [music] their soil. For
6:44most of the farmers on the continent,
6:45they do not have enough insights into
6:47their soil, what fertilizers to use,
6:50[music] what their soil is deficient in.
6:51>> Reading the soil, reading the sky,
6:55>> [music]
6:55>> telling a Ghanaian farmer when to plant
6:58and what to plant.
7:01Hold on to this farm.
7:03>> [music]
7:03>> We will come back to it because how this
7:05story ends may tell us everything.
7:07[music]
7:09Ghana has a plan.
7:11In 2025,
7:13the country launched its national
7:15artificial intelligence strategy, a
7:1810-year roadmap built [music] on eight
7:20pillars and aimed squarely at four
7:23things: agriculture, healthcare,
7:27education, and government [music]
7:28services. It is stewarded by the
7:31Minister for Communication, Digital
7:33Technology [music] and Innovation, Som
7:35George. There is money behind it, there
7:39is a timeline, and there is real
7:41ambition.
7:42>> Our AI strategy gives us a 10-year plan
7:45[music]
7:45of what we intend to achieve, measurable
7:48plans of what do we need to achieve, and
7:50we've started work on that. I want
7:52[music] to leave this ministry with a a
7:54legacy that we've left a digitally
7:56[music] inclusive and data-driven
7:59economy. That for me is the [music]
8:01focus.
8:02>> But between a strategy on paper and a
8:04country transformed, there are four
8:07[music] questions, and how Ghana answers
8:09them decides everything. [music]
8:16Question one, the machine itself.
8:18>> A lot of countries are interested in
8:20just building data centers. I went to
8:22cabinet and said, "We don't need a new
8:24just a new data center. We need an AI
8:26compute center that becomes the nerve
8:28center and the brain box in in
8:30artificial intelligence age, not just
8:32for Ghana, for but for the continent."
8:33So, we're building with capacity beyond
8:35[music] what we need as a country. We
8:37are on the cusp of
8:39starting our projects with [music]
8:40dedicated funding. Our turnaround time
8:43is about 24 months. On the continent, if
8:45you want to talk about the top five
8:47[music]
8:47on the continent, we are definitely in
8:50there, and our plan is to be number one.
8:51>> Ghana's [music] biggest homegrown
8:53digital payments company
8:56backs this idea.
8:57>> Ghana historically has been a first
8:59mover on a lot of major technology that
9:02has shaped our future. [music] And we
9:03had mobile telephony as far as 1992,
9:07right? And
9:08when a lot of African countries had not
9:10even
9:11>> [music]
9:11>> thought about it, the government has
9:13taken the right step forward in building
9:15a policy framework to say, "This is how
9:17we're going to do it."
9:18>> But endorsement is not the same as
9:21agreement, because the moment you leave
9:23the ministry and the boardroom, the
9:25questions get harder, [music]
9:27especially when Ghana plans to spend 250
9:31million dollars on its AI compute
9:34[music] center.
9:35>> It's a very good initiative, because we
9:38need compute to be able to do [music]
9:41most of these things. Okay.
9:43But, as I said, I don't know what went
9:46into
9:47their budget. Did they consider only the
9:51>> [music]
9:51>> purchase of the
9:53computing devices?
9:55Cuz the actual cost is not the purchase
9:57[music] of it. The actual cost is the
9:59running of it.
10:00Power, cooling, professionals to
10:03maintain and use it.
10:08Readiness is is uneven, as I said.
10:12At national level, at some level, when
10:14you look at our universities, our
10:15universities are doing their best.
10:17>> [music]
10:17>> They have these research labs. They're
10:18doing their best. Okay. Some local
10:21researchers, some local organizations
10:22are producing solutions, they're
10:23deploying solutions. Okay.
10:26But, on the
10:27national level,
10:29the readiness is uneven.
10:31Because we have challenges to
10:32>> Explain that when it's uneven, what do
10:33>> Uneven. We still have expensive data.
10:36>> [music]
10:36>> Our electricity is not stable. When it
10:38comes to digital literacy, we are
10:41limited. So, these [music] things,
10:43even if we don't if you don't fix some
10:45of these things,
10:47whatever solution you try to propose
10:48[music]
10:49will not work.
10:50When you have a broken system, and you
10:51think AI is going to solve the broken
10:53system, it won't solve the broken
10:54system.
10:55>> [music]
10:56>> So, if one, we talked about data. If our
10:58data is not well structured,
11:00>> [music]
11:00>> if our data is not reliable, if our
11:03infrastructure is not clear, if we don't
11:05have response when it comes to
11:07governance,
11:08policies, if you don't have the
11:10professionals [music]
11:11to be able to use and supervise these
11:13ones,
11:15then we can't actually build some
11:16[music] of these things.
11:18Okay. Cuz if you have a broken system,
11:20and you put an AI on it, all that you're
11:21trying to do is you're just
11:23fast-forwarding the way we get to
11:24failure.
11:27>> [music]
11:30>> And aside that,
11:31do we have the market after building it?
11:34Do we have the market? And then there's
11:36a central question [music] government
11:38must contend with. Not can we build it,
11:41but [music] who will run it?
11:48>> We wanted to recruit,
11:50I believe it was about 30 machine
11:52learning engineer. We just couldn't find
11:5430 in Ghana.
11:56We increased the budget almost about
11:59twice
12:00and we still couldn't find
12:04machine learning data scientist that
12:05were from
12:07from Ghana. We offered very generous
12:09terms. We still couldn't find cuz 30 is
12:11still a very small number.
12:14The money to set up
12:15>> [music]
12:15>> the the starting data centers and the
12:18architecture and the infrastructure
12:19around it is very important.
12:21But whilst we do that, I think we also
12:23have to look at the [music] the skills.
12:25Those digital skills have to really come
12:27up.
12:28>> Hapto [music] knows this because Hapto
12:31tried and learned the lessons.
12:34>> We set up a lab to see let's start
12:36building our own from scratch using
12:40the deep neural network [music] theory.
12:42We opened the lab in 2023. The biggest
12:45one that we announced was a ECG.
12:47That did give us [music] an opportunity
12:49to help them use AI to find
12:53people that could be stealing
12:54electricity.
12:55What we found
12:56>> [music]
12:56>> towards the end of 2024 into 2025 was
13:00that
13:01others were way ahead to the extent
13:03[music] they have commercialized their
13:04models and those models that are
13:06commercial and are selling for a
13:08fraction
13:10of the price like [music] 20 25 dollars
13:13per month
13:14are way more advanced. [music]
13:17So we had to pivot to also start buying
13:19cookies
13:21just like everybody else.
13:23>> [music]
13:23>> A Ghanaian company built his own
13:25artificial intelligence then abandoned
13:27it because the world had moved further,
13:31faster, and cheaper than it could.
13:35Government is now about to attempt the
13:37same thing
13:39with public money.
13:41Question two,
13:43the bill.
13:46Building national artificial
13:48intelligence capability is among the
13:50most expensive undertakings a country
13:53can commit to.
13:55In the United States, a single private
13:57data center program has been valued at
14:00around half a trillion dollars.
14:03Across Africa, the numbers are smaller,
14:05but they are still enormous.
14:09>> In Africa, I have seen, you know,
14:11countries like South Africa that has
14:13planned to invest about 3.7 billion
14:16dollars in their national AI strategy.
14:19I've seen Angola announcing about 300
14:22million dollars. I've seen Kenya wanting
14:25to invest around a billion dollars, and
14:27Nigeria wants to do around 350 billion
14:30dollars.
14:30>> The government has committed 270 million
14:33dollars.
14:34T50 million dollars towards the phase
14:36one of the AI compute center, and 20
14:38million dollars towards the setup of the
14:39AI fund.
14:43The AI fund is something that will need
14:45to look at multiple sources of funding.
14:47Can't be funded only by government. In
14:49many countries, you have private sector
14:51players
14:52who are investing in AI funds, [music]
14:54and the government is giving them
14:55incentives. We will need to have that
14:56conversation in the country. There's no
14:58way if government [music] has some very
15:02beautiful
15:03uh services, data, and AI, and all of
15:07that that we would we would bypass it
15:09and not use it. We would need to use it
15:11because we will appreciate it.
15:12>> So, definitely [music] the after is one
15:14of the I'm sure government will have to
15:15look at that. Yeah. Yeah, definitely
15:17some some incentive. These are moments
15:20that the two must pull together. Both
15:22the private sector and government must
15:23work hand in hand to
15:25to to be intentional [music] about what
15:27kind of future we going to create.
15:29>> When you talk to players and governments
15:32in this space, they accept
15:34the fact that AI technology is a huge
15:37opportunity for us,
15:39but you never quite see that reflected
15:41in say budgets.
15:42So, when the Minister of Finance comes
15:44to speak,
15:45you know, presents [music]
15:46the budget for the year, you don't quite
15:48see that reflected that hey, if you
15:50think that AI can move us forward, why
15:52is it not
15:53you know, showing it? You may be able to
15:55fund it for a month,
15:57paying $20 to $50,000 just to keep
16:00your computing infrastructure in place,
16:03but you can't do that for a long term.
16:05So, you realize that a lot of startups
16:06start, but then
16:08halfway into
16:10uh the year, they fizzle out somehow.
16:13>> Question three.
16:14This one may be the most important of
16:17all.
16:19Because a computer center is a very
16:21expensive brain, and a [music] brain
16:24with nothing to think about is just a
16:28very expensive building.
16:30>> AI really runs on a lot of data. Your AI
16:33models are only as good as the data sets
16:35that are being used to train the
16:36machines. Our data exchange hub, which
16:38is another thing that we're working on
16:40in the ministry here, that's putting
16:42together 48 data registries in the
16:44country into a seamless interconnection.
16:46>> Most of our data is unstructured and
16:49fragmented.
16:51Okay, if you pick a typical
16:53um local assembly, any government
16:55institution, and they collaborate,
16:58but their data is fragmented.
17:00So, if you really want to deploy an AI
17:02on top of some of these things, it won't
17:04solve the problem.
17:06We first
17:07need to solve the data problem.
17:10Models can just be built, but models
17:12without good data is useless.
17:14>> We struggled a a with data. At a point,
17:16we tried connect with other large
17:18companies to see if they can give us
17:21they can share data with us, we can
17:23share some of our data with them. We
17:24realized
17:25there was this there's still no
17:27framework in Ghana for sharing data. Do
17:30we sign NDAs? Do we copy our data in
17:33Excel and give it to you? There is
17:34there's no no framework. [music]
17:36What we know has been successful in the
17:39US and China
17:41and partly in Europe about these AI
17:43models is that there is a data sharing
17:45framework within [music] these
17:46countries. We just found that in Ghana
17:48it wasn't so
17:50there wasn't that many company willing
17:52to to share data [music] with us. It
17:54became another obstacle that didn't
17:56really help our success. At the least we
17:59should have
18:01a [music] data framework that will
18:05sort of call [music] on companies to
18:08actually not delete data but store them
18:11and have [music] a framework for
18:13uh
18:15incentivizing companies to store data
18:17[music] because at some point when we
18:19finish the AI data centers
18:22we will need a lot of data. But it's a
18:25step in [music] the in the in the right
18:26direction. We will need that. If we had
18:28that, I'm sure our AI lab would have
18:31been a a success.
18:33Uh yeah, but we didn't have that so
18:36yeah.
18:37>> And yet, [music] despite the
18:39fragmentation despite the missing
18:42framework Hapto
18:43>> [music]
18:44>> found a way
18:45by turning inward to the only data it
18:48fully controlled.
18:50Its own.
18:51>> How can I help you today?
18:54>> Oh.
18:55>> Uh was this order delivered?
18:58My last order for Panadol, was it
19:00delivered?
19:03>> Let me check that that for you right
19:05away. This usually takes about a minute.
19:09You might hear some silence while I'm
19:11pulling, but I'm still here.
19:15>> The network quality is not
19:17>> So, that's that's AI?
19:18>> Yeah, it is AI.
19:20>> It's all It's all
19:22That is not bad at all.
19:23>> Yeah.
19:25>> Yeah, the network quality
19:26>> Thanks. I
19:28So, let me get back to helping you with
19:29your Panadol order. Let me check the
19:32delivery status for you.
19:34>> So, it's not to be checking.
19:35>> Mhm.
19:37>> So, at this point, to go into
19:39the data
19:40>> get it.
19:41>> And check it.
19:43>> Yeah, let me actually get your Panadol
19:45order. This will take about a minute
19:48while I pull up your order history.
19:52Thanks for waiting. So, I checked your
19:54recent orders and I can see a few
19:57transactions from July 25th.
20:00>> Because of the sheer volumes of
20:02transactions we process,
20:04we tend to get a lot of phone calls. So,
20:06we have developed
20:08a a virtual call center where when you
20:11call,
20:12our calls are sort of pre-screened by an
20:15AI agent. It has really increased
20:18our ability to be able to serve a lot
20:20more customers. This one is like what we
20:22call like a closed-loop model. So,
20:24because all the data it needs to do its
20:26work [music]
20:27>> It sits within Airtel.
20:29>> It sits within Airtel.
20:30>> Locally built?
20:31>> Built locally. In fact, the lead
20:33engineer is sitting in Accra right now.
20:36>> Built in Ghana, trained on Ghanaian
20:38data, answering [music]
20:40Ghanaian customers in a Ghanaian accent.
20:44So, I asked him the simplest question of
20:46all. So, in a sentence, how would you
20:49define AI to a typical
20:52Airtel customer in a village? So, AI is
20:56not a thing. It is us. It's still us.
20:59There is no AI without humans.
21:02Everything AI knows is us. We gave it to
21:06it. It's our information. It's our
21:08files. It's our documents. It's the
21:10things we know. If we are biased, AI
21:12will be biased. It's almost like a
21:14knowledge base. It's a knowledge base
21:16that we all tap into to solve our
21:17problems.
21:18>> It is us. We made it. It is It is us.
21:20It's all our data. It's our files. It's
21:22our documents. It's the things we know.
21:24If we know it, we AI
21:26we can and we want to make AI know it,
21:28we'll make it know it. [music] Yeah.
21:30>> Okay. If artificial intelligence is
21:32built from our own data and paid for
21:35with our own money, then the last
21:37question is the one that decides who it
21:40is actually [music] for.
21:43Question four,
21:45who gets left behind?
21:47>> That what we said AI will not just mean
21:48artificial intelligence, it will mean
21:50advancing inequalities.
21:52And so, we need to be very minded about
21:54that. We've inherited almost 3,000 rural
21:56telephony sites that offer only 2G and
21:583G services, only voice services. You're
22:00You're You're fostering a division and
22:02an inequality because you want the rural
22:04folks to use artificial intelligence,
22:06then they need to be able to use 4G and
22:075G technology because that's where you
22:09can stream, that's where you can learn.
22:11>> The first wave or first beneficiary of
22:13AI going to be people who know how to
22:16use the technology. And that will be the
22:17large companies. So, what I see would
22:20happen is that K-shape economy where
22:24the people who are already rich, the big
22:26[music] companies, are going to get
22:28bigger
22:29first.
22:30And it will [music] take a while
22:32before it starts hitting the micro
22:34economy.
22:35>> We view AI as this very
22:38esoteric
22:39technology
22:41that everyone needs to learn very
22:43high-level things [music] to be able to
22:45appreciate.
22:46And if we do that, we we lose out on a
22:50greater part of our population because
22:52there are so many people who don't even
22:54have basic skills. [music]
22:55So, rather than thinking about all these
22:58high-level topics, why don't you think
23:00about the basics? People living rural
23:02areas, people living with disability,
23:04women who do not have the economic power
23:07to even [music] own a mobile phone.
23:09>> Perhaps, beyond the concerns about AI
23:12widening the digital divide is its
23:15potential [music]
23:16impact on something Ghana already
23:20does not have enough of.
23:22Jobs.
23:24According to the Ghana Statistical
23:25Service, national unemployment stands at
23:28around 13%.
23:31But among young Ghanaians age 15 to 35
23:35it is more than 22%.
23:38Seven out of every 10 unemployed people
23:41in this country are young.
23:44The fear is that AI will wipe out even
23:47more.
23:48>> [music]
23:48>> And that disruption
23:50is not coming. It is already here.
23:53>> Many years ago, in fact, before COVID
23:56there were business reports that
23:57companies would automatically send it to
23:59us for translation.
24:02But today companies have discovered
24:04that with artificial intelligence,
24:06ChatGPT, Gemini you know, DeepL,
24:09Claude especially uh they can simply
24:12upload a text
24:13and depending on the prompt, they have
24:15the output that they need. So, documents
24:17that were sent to us in the past
24:20gradually
24:21are no longer being sent to translators.
24:24>> And it is not only translation.
24:26>> So, conference interpreting is also
24:28being affected.
24:29>> There is also transcription. Dabbing
24:31these days so many voices online just
24:34get a right to
24:36you know, and you get your video dubbed
24:38for you. There is also subtitling. All
24:41these are our services.
24:43Uh services that language professionals
24:45you know, uh provide.
24:46>> I asked him what that cost him.
24:48>> In our area we always counting words.
24:52So, if you have 5,000 words, that means
24:54more money in your pocket. But now if a
24:56company would send you just a letter
24:58instead of sending you a whole report,
25:00it means you are getting money for just
25:02one page.
25:03>> There's this client who in the past
25:05would contact me every month for
25:07translation.
25:09>> I don't remember the last time I got a
25:10document from the client this year.
25:12>> That's exactly food off my table in that
25:15regard, which is painful. A translator
25:16in Lagos, but [music] also a graduate in
25:19Accra, a designer, a clerk, a junior
25:22writer, and a generation training for
25:25jobs that may not be there when they
25:28arrive.
25:29>> I use AI literally every day. When it
25:32comes to academics, AI is particularly
25:34helpful. You can literally send in your
25:36source material, give it specifics on
25:38how to test you, and basically it tests
25:40you within seconds. [music]
25:42>> The same student who cannot study
25:44without it is terrified of it.
25:48>> As someone who is looking forward to
25:49enter the media space as a journalist,
25:52it is quite an interesting reality to
25:55know that AI is basically doing most of
25:56the things humans do. It's quite
25:58threatening to me to know that the very
26:01things that I'm training to become in
26:04the future is already being overtaken by
26:06AI. And so, it's quite an interesting
26:08dynamic to train for what is under
26:10threat.
26:11>> And then, there are those who think the
26:13whole national project is built on sand.
26:17>> Your education system is still lacking.
26:19We still have schools under trees. We
26:21still have
26:23rolling blackouts.
26:24>> Teachers are trying to teach computer
26:25use, they're going to have to do it on a
26:26blackboard.
26:28>> To say this is how a mouse and this is
26:29how you use it. How is that child going
26:31to learn what an AI is and make use of
26:33it?
26:34>> If you don't have the practical basic
26:36set,
26:37>> how are you going to build up from
26:38there? Build your basics first. AI can
26:40enhance, but people are not using it to
26:42enhance, they're using it to replace.
26:43So, where are those jobs supposed to
26:45come from? So far,
26:47all the metrics prove that it is not
26:49working out that way.
26:50>> Supposing I had 20 clerks doing data
26:54entry roles within the economy and now I
26:57have a tool [music]
26:59that automates those processes. Maybe I
27:02don't need all the 20, so I can let go
27:0410 of them. Where would the other 10 go?
27:07They probably would end up joining the
27:08unemployment pool. On top of that,
27:11you're also churning out new graduates
27:14every year coming into the labor force.
27:17So, is a fear justified? [music]
27:19Around the world, the first wave of
27:22evidence is now coming in and it is more
27:26complicated than the headlines
27:28suggested. According to the
27:30International Staffing and Consulting
27:32Firm Robert Half, 29% of organizations
27:36that cut employees because of AI had
27:40already rehired into the very roles they
27:43had eliminated. [music] The research and
27:46advisory firm Forrester goes further
27:49predicting that 55% of executives who
27:53replaced their employees with AI will
27:56regret that decision within 18 months.
28:00And the most widely reported example of
28:03all, Klarna, the global fintech, was
28:06forced to rehire most of the 700 staff
28:11it had let go after integrating AI into
28:14its business.
28:15The disruption is real, but the story of
28:18what comes after it is still being
28:21[music] written. So, I put it to the man
28:24whose company has been through it.
28:26>> It don't take away jobs. I think in the
28:28end AI is going to create a lot more
28:30jobs. It's not going to take anybody's
28:31job. I think that it will force all of
28:33us to evolve.
28:34>> Evolve. So, what does it actually look
28:37like for a real person whose income has
28:40already been taking?
28:41>> Because with AI there are new roles for
28:43language professionals. Some people are
28:45not aware of those rules.
28:47But some of us, thankfully,
28:49are already performing those rules and I
28:51we are making money that we are not
28:52making in the past without AI.
28:54Today with AI, there is what is called
28:57annotation.
28:58Without AI, we did not have annotation.
29:01All these AI platforms are what they are
29:03because of the work of language
29:05professionals. That's why we call them
29:07large language models. I personally, I
29:10am consulting uh for an AI company
29:13as an auditor. The AI platform
29:17generates the output. My work is to
29:20audit
29:21audit the output
29:23and tell the the the AI company whether
29:26the output is good.
29:28And now, my judgment is used to train
29:30the model.
29:32That is money that I would not have made
29:34in the past. Today, I sit in my room
29:36with my laptop and internet connection
29:38and I'm making that money.
29:39>> The technology that took 75% of his
29:42business now pays him to make it better.
29:46>> So, the trade is not under threat.
29:48Professionals just have to adjust,
29:51realize that these tools are here to
29:53stay. We have to embrace them and grow
29:56with them. Evolve or die, that is what
29:58is going to happen.
29:59>> As humans, we are extremely creative. We
30:02look at the world we have created. Our
30:04creativity will far outweigh any
30:06technology that we will [music] be able
30:08to build.
30:09Um the the second thing is that
30:13we have almost enormous
30:17and endless desires as as humans. We're
30:21never satisfied.
30:22>> [music]
30:23>> The third and final thing is that
30:25companies don't compete on how many jobs
30:28you cut.
30:29>> [music]
30:29>> Every company competes on how much
30:31profit you can make, how many more
30:33people you can save. There's no company
30:35in the world that has become the world's
30:37biggest company
30:38>> [music]
30:38>> by cutting jobs and shrinking.
30:41>> His optimism is not a slogan. It rests
30:43[music] on three arguments. The argument
30:46is not that nothing changes. It is that
30:49the work changes
30:51>> [music]
30:51>> and the person stays.
30:54>> Our engineering department has been
30:56changing to engineering and
30:58productivity. Their role is
31:01um is to help our staff transition
31:04uh into this new era era of work.
31:06>> What I see here is AI
31:08is only changing
31:11the procedure in a particular
31:13profession. First, you use the tools
31:17to
31:18enhance your tasks.
31:20And then when you enhance your task, you
31:22realize that even the profession that
31:23you are in now should be changed.
31:27So with time that profession phased out,
31:29but it created a new job.
31:31If you don't take the opportunity now,
31:34you'll become redundant.
31:36>> [music]
31:37>> So the accountant who is trying to learn
31:39the tool of an AI tool
31:42in years to come will be more relevant
31:45than the colleague who said I am waiting
31:48[music] and I don't like AI tool.
31:49>> And that brings us to the end of news in
31:52brief.
31:52>> If you bring an AI to read the news, the
31:54AI is just going to read it like a bot.
31:57No emotions.
31:59Where is the feeling? Where is the
32:01empathy? [music] You have mastered over
32:02the years.
32:04So now the the question now becomes, how
32:06do I transfer the knowledge that I have
32:09to make whatever AI agent that is
32:11reading the news read properly like me?
32:14You understand? So in that sense, it
32:16creates a new job for you.
32:18>> Margaret is in no way a threat to
32:20journalism. Many people say AI is going
32:24to take our jobs, but then I say AI is
32:26just here to assist us. I I don't want
32:29any student to see Margaret as a
32:30replacement, but then look at Margaret
32:32as a model to help you improve in your
32:35journalism because if you have somebody
32:36reading the news, you got the time to
32:38chase the news out there.
32:39>> Not fewer people,
32:41different work.
32:43Which brings me back to a [music] farm
32:45in the Eastern Region
32:47and to a young man who lost almost
32:49everything in his first season.
32:52>> In the first season, we wanted to grow
32:55we were going for tomatoes. [music] And
32:57of course, what is recommended when
32:59you're farming is to do soil testing.
33:02But the first people that wanted to try
33:04soil testing with were stop offering the
33:07service when we were about to start. So
33:09we went in blind. The expectation was
33:12that we're looking at between two to
33:13three tons of tomatoes at the end of the
33:16season. But at the end of the season, I
33:18don't think we even got up to 200 kg. It
33:21wasn't even enough to break even with
33:23the cost that we put in.
33:25>> Two to three tons expected, less than
33:28200 kg harvested. A season's work and a
33:32season's money gone into the ground.
33:35>> So our next season, we decided that the
33:37first thing to do is to truly is to test
33:39the soil, to truly know what we are
33:41working with. They gave us the soil test
33:43just after two days, by the way, after
33:46taking the soil sample along with a
33:47nutrients management plan. We got to
33:49know that we did have nitrogen, but
33:51we're lacking in phosphorus. We're
33:52lacking also in potassium. And having a
33:55lack in phosphorus also contributed to
33:57our failed first season. This time, we
33:59are seeing less stunting of the plant.
34:01Just a day after transplanting, they
34:03were picking up. They were healthy, they
34:05were strong. After doing the fertilizer
34:06application like recommended, we saw the
34:09color of the leaves pick up and the
34:11plant starts to shoot up properly. So
34:13[music] following the recommendations,
34:15it's actually yielding results right now
34:17on our farm.
34:18>> The difference between a failed season
34:20and a good one was not more money or
34:23more land or more luck.
34:26It was better information
34:28delivered in two days.
34:31>> So it's in [music] two parts. The
34:32software part is supposed to help you
34:34map the land and that is going to help
34:36us be able to capture the land and be
34:38able to provide and satellite
34:40information and insights about it. The
34:42hardware is going to help us pick direct
34:45values right from the soil. I.e. the pH,
34:49the electrical conductivity, the
34:51moisture level, the temperature and
34:52humidity in there. These are the results
34:55that we're going to put together to be
34:57able to
34:58give the farmers better information.
35:01Most farmers do not test or do not even
35:04look into what is truly in their soil.
35:06Very, very few, less than 10% who have
35:09access to some form of testing through
35:11labs and all of this which are quite
35:13expensive are able to [music] go through
35:15those process. What Farmpine is trying
35:17to do is to make sure that this is
35:19accessible to everyone. Even smallholder
35:21farmers can be able to access our
35:23services. They don't need to be super
35:25rich farmers.
35:25>> Fewer than one in 10 Ghanaian farmers
35:28has ever tested their soil. This is what
35:31closing that gap looks like. [music]
35:34>> We have a lot of testimonials from major
35:37agribusinesses in the country new to see
35:4030% in terms of yields at the ends of
35:44most of their farming activities since
35:46we launched [music] last year till now.
35:48>> The main problem with AI when it comes
35:50to our local context is it's will give
35:52you information that is probably western
35:54or common to a western but it's not
35:57common here.
35:58The growing seasons are broad and the
36:00growing seasons here are completely
36:01different.
36:02>> [music]
36:02>> They are building their own
36:05internal large language models with
36:07their analytics and all of that that
36:09helps them to address specific issues
36:13related to their country.
36:15>> So, the more AI is trained for our
36:18context the better AI will be
36:19>> [music]
36:20>> for our use. So, AI is is indispensable
36:22to a young farmer as at this point.
36:24>> Built in Ghana,
36:26working in Ghana, feeding [music] Ghana.
36:30So, after everything, the disruption,
36:33the fear, the fragmented data, the
36:36missing engineers, and a farm that
36:38finally produced,
36:40what is the answer?
36:49>> [music]
36:50>> I belong to the school of thought that
36:52says that AI will create more jobs than
36:54cost jobs.
36:55>> We're hoping that on the average it will
36:57add about 5 billion a year US dollars to
36:59our economy in terms of productivity, in
37:02terms of savings that are made, in terms
37:04of, for example, our food security.
37:13>> It's been quite an experience talking to
37:15all manner of people.
37:17Several voices on the subject of
37:20artificial intelligence and the gig
37:22economy.
37:23Is it a disruption
37:25or is it an opportunity?
37:27And it's obvious to me, having heard
37:29everybody,
37:30that the consensus is overwhelming
37:35that artificial intelligence
37:37is
37:39an opportunity.
37:40An opportunity we need to embrace.
37:44And even if
37:46there will be
37:47a disruption,
37:49it is just entering.
37:51But
37:52an everlasting opportunity is indeed.
37:59>> [music]
38:06[music]
38:18>> So, welcome back. Now, to help us unpack
38:21what we have just watched, let me
38:22introduce our panelists. We have
38:24Sam George. You saw him in the
38:26documentary lay out the government
38:27policies and Minister of Communication,
38:29Digital Technology and Innovation. Sam
38:32George, thank you very much for joining
38:33us.
38:34Also, Adam Watkins, founder of AI for
38:36Ghana. Adam, I'm grateful that you could
38:38join us.
38:39Augustine A. Jay is head of engineering
38:42at Haptor. Bright Simons is vice
38:45president of Imani Africa. Daniel
38:47Quarcoo Mackey is the country director
38:49of Zipline Ghana.
38:51I'm I'm delighted that you could join
38:53us.
38:54And also joining us via Zoom is
38:56economist and technical advisor at the
38:58Finance Ministry. He joins us on Zoom
39:00from London. I'm grateful that he could
39:02join us, too. And thank you all for
39:04joining us. I want to come to the studio
39:07now for
39:08quick reactions to what we just watched
39:10there.
39:11Adam, you were not part of the people we
39:13spoke to, so I want to start with you
39:15first. Give me your your impressions. Do
39:16you agree with the central conclusion in
39:19the documentary that AI
39:21for for country like Ghana and our
39:22economy is ultimately going to be an
39:24opportunity not a disruption?
39:27>> I believe that
39:29AI in Ghana
39:31and across Africa is not an opportunity
39:34or disruption. I believe that it's the
39:37same event
39:38from two different perspectives.
39:41So, we we heard it in the documentary.
39:43You know, some people lost jobs and then
39:47they were gained back.
39:48So, take the take the company I lead in
39:50the US, right? Just as an example and
39:52then we can transition to Ghana.
39:54We took quality control of calls from
39:588 hours over three managers down to just
40:01two hours over one manager.
40:03That represented an opportunity for the
40:06one manager left because they were
40:07redeployed to coach more people. It
40:10represented disruption for the two that
40:12we laid off because we didn't need as
40:13much. And when I look at the people I
40:16work with in Ghana, I mean
40:18I think that people in Ghana are far
40:20less susceptible to disruption. And
40:23that's because of the way the, you know,
40:25MSMEs are are here predominantly within
40:27the country. And the people that I work
40:30with on the ground level, they're
40:31creating something new that was never
40:33there before. So from their perspective,
40:35it's absolutely opportunity.
40:37>> Okay, I see. That's a fascinating way to
40:39put it. So it's both at the same time.
40:41>> Both at the same
40:41>> Interesting. We'll interrogate that
40:43pretty shortly from the other panelists.
40:45>> [snorts]
40:45>> And Daniel, I want to I want to come to
40:47you.
40:48Does any of what you watch reflect the
40:50reality of what you've experienced with
40:52Zipline as it deployed technology to to
40:54serve rural communities?
40:59>> I mean, it's kind of
41:01difficult to just narrow down to the
41:04Zipline
41:06example specifically just because it
41:07kind of touches up on everything.
41:10But
41:12um
41:14if I take a step back and I look at
41:15disruption versus opportunity, right? I
41:17think disruption in itself is an
41:19opportunity. And that's what we
41:21definitely have experienced that things
41:23are possible in terms of technology
41:25delivery that you can't anticipate the
41:28outcome until you actually do it. So
41:31there's far more possible that you can
41:33envision initially.
41:35But I would also like to add that for
41:36me, the way I see it is that it is an
41:38opportunity that we cannot miss. Okay. I
41:41think it's like one of those
41:43technologies that if we miss the train
41:45it will last for
41:48decades if not centuries.
41:49>> Okay, so I'll come back and ask the
41:50question then, how do we ensure we don't
41:52miss it? Because that train is moving
41:54very, very fast. On a daily basis,
41:56things are changing. And whether we are
41:57we are we are moving fast enough. Let me
41:59stay with the industry players and I'll
42:01come to you, Augustine, now.
42:03What's your conclusion? What's your
42:05What's your reaction to what you just
42:06saw there? Because you're also deploying
42:08AI. Where Where do you stand on this
42:10question?
42:11>> So even so, uh good evening to your
42:13studio audience and then to the viewers
42:15at home and everywhere.
42:17Um I think it's it's also is is both
42:20um
42:21growth is painful.
42:23Uh growth is hard.
42:24But um
42:26the disruption is downwards, but this is
42:29forward.
42:30The opportunity is exponential.
42:33So, you heard my boss saying something
42:35about the K-shape in the in the that
42:38graph thing in economics. It's very
42:41real. That's what is going to happen.
42:43Um there's going to be a challenge.
42:45Leaders are going to rethink how they
42:48work
42:50if they really want to grow. It's going
42:52to bring that painful setback, but it's
42:54going to leap us forward at the same
42:56time, just like how a catapult would do.
42:58Let me Let me just tell a story. Over
43:00the weekend,
43:02um Alex came up with an idea that is
43:04best expressed as a mobile app for
43:07users. And it might create some jobs,
43:10too. And over the weekend, he just
43:11called a few people in Haptel, discussed
43:14with them, tried it out manually. Monday
43:17morning, we're on the call, 8:00 a.m. He
43:20had opened up his AI assistant, poured
43:23in a lot of context
43:25within the the the the framework that we
43:28operate in Haptel, our processes and all
43:30that. He came up with very beautiful
43:32prototypes. And this is a CEO, a very
43:34busy guy, right? But he was able to come
43:36up with prototypes for how the mobile
43:38app should look like. He assembled us
43:40together, the engineering team, the
43:42designers, the the sales people, and
43:46others in the meeting. And then we had a
43:49nice call. So, you you can see that
43:53we are just blurring the lines between
43:56when a product um or who can actually
44:00dream up faster.
44:02Because here's a very busy individual.
44:04He didn't need to go have a long talk
44:06with designers. And then a week later
44:09they come up with a design right there
44:10in about a few hours over a weekend. He
44:13does that with artificial intelligence.
44:15Right? He gets us in the call, engineers
44:18understand what he wants to do. They
44:20pick it up. Two days later, we have the
44:23mobile app ready as a as a prototype on
44:26the Play Stores and the App Stores.
44:28Right? The sales guys are also ready to
44:31pitch it. They're ready to get people
44:32involved to try it out. So, you can see
44:35the speed to market. You can see
44:38fast [snorts] failure if you if you may
44:40or fail forward because if the idea
44:42doesn't work, we can roll back quickly
44:44and say that, "Okay, let's go back to
44:45the drawing board and do our work well."
44:47Whereas, if you compare to previous ways
44:50of work, this idea would have taken a
44:52month because designers have to come in,
44:55uh product managers have to think
44:57through the the readiness for markets,
44:59go for competitive analysis, and all
45:01that. But now that time has been
45:03shortened. And so, it is forcing us as a
45:06company to also rethink our our whole
45:09process flow for getting out products
45:11that are really meaningful and useful
45:14for those that will use them.
45:16>> It's an interesting example you use. I
45:17mean, there in lies the tension. So,
45:19he's done it quickly.
45:21Before, you'd need more hands to get
45:23this done. It would take longer. And
45:25that's where the tension is. So, you
45:27doing quickly is good, but then of
45:28course, the tension is that it it may
45:30lead to somebody
45:32>> [snorts]
45:32>> not having as much to do or then you
45:34eventually say, "Well, instead of five
45:36people, I need two because now you can
45:38move quickly." And then three people are
45:40now looking for jobs. We'll come and
45:41explore that. Let me bring in Sam
45:42George. Sam George, you were a central
45:45character in the documentary. Now,
45:47having listened to all the views there
45:49from those who are actually using it,
45:51the the the the question that they were
45:52they were asking. And the interesting
45:54thing about documentaries is that
45:55everybody speaks different points in
45:57time, so you don't get to interact. But
45:58questions were asked. Having watched
46:00everything, does that How does that
46:03shape your perspective as a policy
46:04holder? Because everything else is based
46:06on what Ghana is attempting to do.
46:09>> Well, well Evans, thank you and good
46:11evening to my co-panelists and the
46:14studio audience and our viewers at home.
46:16Um listening to the documentary and the
46:19various um
46:21individuals who spoke, you could see
46:24the need for the work that we're doing
46:26at the ministry.
46:28Um Alex speaking from Hubtel spoke about
46:32the challenges that they had with data
46:35and access to data. And and and so I'm
46:38happy that that's something that is
46:40being worked on at the ministry at the
46:41moment. We are working on a data
46:43exchange hub. We have about 48 um data
46:47registries in the country that process
46:49so from Ghana Health Service to National
46:51Identification Authority to Ghana
46:52Statistical Service. Literally any data
46:55set in the country is processed by one
46:58data processor that doesn't speak to the
47:01other.
47:02Now, if [snorts] Hubtel wanted that
47:04information, it meant you needed to have
47:06an established connection with 48
47:07different registries.
47:09With a data exchange hub, what we're
47:10doing is we're building a single
47:12national backbone
47:14where all our 48 public registries
47:16connect seamlessly in an interoperable
47:19manner to that backbone. What it does
47:22for us is we're now going to issue
47:23standards from NIITA. So, every data
47:25processor is going to have to classify
47:27his data in a unified manner across the
47:29country. So, we're dealing with wrong
47:31classification and annotation of data in
47:34one swoop with one policy decision. Now,
47:37all data processors are still going to
47:38hold their data and be responsible for
47:39their data, but all you do is you have a
47:41single point where you query that data
47:44as a as a developer. So, you want data
47:46from the National Identification
47:48Authority, the Ghana Statistical
47:50Service, the Birth and Death Registry,
47:52and the Ghana Police Service. One single
47:54request to that exchange hub provides
47:57you that data set.
47:58We're starting as a public data
48:00registry, but we believe that it's going
48:02to be open it's going to be opened up to
48:03the private sector. We won't compel
48:04private sector to, but the incentives
48:07for private sector to take advantage of
48:09that registry exist because you can only
48:12take data out of that registry if you
48:14contribute. So, if you're a private
48:16sector player and you also hold data and
48:18you want the data that government is
48:20controlling, you need to contribute your
48:21data so that you you're part of the
48:23ecosystem.
48:24Um the question about jobs came up um
48:28very strongly and and I agree that jobs
48:31are going to I always use this analogy
48:33of um an elderly aunt of mine who was a
48:37stenographer grade three in the public
48:39service much earlier in in what when I
48:42was much younger and whenever we visited
48:44her home, her typewriter was one of the
48:46highlights of why we always wanted to
48:47visit her home because she had the
48:49carbon paper and the paper sheets in
48:51there and the duplicate behind it, but
48:53the most exciting part for me as a young
48:55kid was after typing, slapping the the
48:58return [snorts] the yeah, you know,
49:00slapping and Gen Z's won't understand
49:01what we're talking about.
49:02>> [laughter]
49:03>> But you pull on it and then that's your
49:05enter key.
49:07Then we transitioned to computers.
49:10First electronic typewriters, then
49:12computers.
49:13Now, with the electronic typewriter, you
49:15still did but this time with an enter
49:17key, but you never saved your document.
49:20Then we moved to computers where you had
49:21to save your document and know how to
49:23use the Microsoft Office Suite. Now,
49:25that aunt of mine who was a stenographer
49:27grade three in the public service, if
49:30she did not learn with Mavis Beacon as
49:32most people did that time how to use the
49:33Microsoft Office Suite, would have lost
49:35her job because
49:37by 2000, you didn't have any typewriters
49:39sitting in the public service.
49:41And so, it's about evolving and
49:43adapting.
49:44Those who evolve and adapt will survive.
49:47Those who fail to evolve and adapt will
49:49fall off. It's the hard reality of
49:50artificial intelligence.
49:52The lecturer you spoke to from Ashesi
49:54asked the question of who is building
49:57and what are you building? We have a
49:59headline figure of $270 million. $20
50:02million for seed funding a seed fund for
50:05the AI fund, but $250 million is going
50:08into building a physical structure, the
50:11compute, and the AI backbone. Because I
50:16have a problem as minister today because
50:18I I see a number of government material,
50:23what you call classified government
50:24material, being fed into ChatGPT and
50:27Claude.
50:28>> Okay.
50:30>> That for me represents a critical
50:32national
50:33risk
50:34because
50:35memos, government material, critical
50:38government sensitive data is being fed
50:40into these systems by analysts sitting
50:43in ministers' offices to process that
50:44data. Now, that we have a data residency
50:47law that we've passed, but it's useless
50:50if the government system itself is using
50:52a foreign um um um
50:55AI model to do processing, but you can't
50:57deny the public sector
51:00the speed and power of AI compute. And
51:03so, we're building what we call the
51:05Ghana Chat, which becomes our own
51:06national language processing model that
51:09allows for government to interface on
51:11our own, build our own capacity.
51:14At at start, will it be as effective as
51:16Claude and ChatGPT? Definitely not. But
51:18then the more you use it and the more
51:20you teach it, the more the machines
51:22learn, the better the system will get.
51:24Because whatever systems we're using
51:25today were built, piloted, trained, and
51:29taught on billions, if not trillions, of
51:31data sets.
51:32>> Chat is going to be a local model we
51:33built for Ghana.
51:34>> Absolutely built by Ghanaians, and fed
51:35with Ghanaian data.
51:36>> Fed with Ghanaian data. Yes. So,
51:38basically, I see all the questions that
51:40have been raised, very legitimate
51:42questions, very legitimate issues, but I
51:46believe that we are on the right track
51:48to resolving all of these issues. Like I
51:50said,
51:51we're currently going through
51:52procurement with PPA
51:56on on getting the data center, the
51:58computing infrastructure, and all the
51:59various parts of it put together. 24
52:02months from now, we should have the
52:04phase one completed with serious
52:06computing infrastructure that we can
52:08then give to our researchers, our
52:10industry, our universities who are
52:13currently buying tokens outside the
52:15country and buying capacity outside the
52:16country. It's the first step. It's a
52:18long journey. Are we going to fix all of
52:20our AI needs in two years before the
52:21Muhammad administration is off? That's
52:23wishful thinking. It's impossible. But
52:25we are taking the key foundational steps
52:28that will set this country on the right
52:30track for its AI journey.
52:32>> Okay, great initial comments. And I have
52:36Bright Simons here, vice president of
52:37Imani.
52:38Bright has a short presentation on the
52:41back of everything that has been said
52:42and what we've watched.
52:44And then once he's done, we will
52:45interrogate what he said, what the
52:46minister has said, and I'll bring my
52:48audiences into the conversation. I don't
52:50know how many out there are watching. So
52:51Bright, let's let's let's let's watch
52:52what what you've done with the the work
52:55that we've already presented, what the
52:57minister has said in the initial
52:58comments that had been put forward on on
53:00the data economic forum.
53:02>> Thank you very much, Evans. Is my voice
53:04audible?
53:04>> Clear. Yes.
53:05>> Wonderful. So like you said, I have a
53:07few even though the scene has been set
53:09by the documentary, my own sincere and
53:11remarks
53:13that I prepared. And I wanted you to
53:15take us up on this notion of teaching
53:16and learning because what makes AI
53:19different, without doubt, is the fact
53:21that it's a new system that learns.
53:24Most technologies are deterministic. You
53:25have to plan it, and then more or less
53:27what you get is what you plan.
53:29AI is grown. It It grows in the medium
53:32in which it's planted. So the way your
53:34AI works is probably different from the
53:36way this
53:37Emmanuel's AI works, primarily because
53:39you use it differently, do it
53:40differently, etc., etc.
53:43So, I'm using this notion of resets and
53:45resets because reset is the big Let's
53:47move this way.
53:48Uh reset is the big um
53:50um
53:51you know, buzzword in Ghana right now.
53:53But when you look at our policy history,
53:55it looks as if for a lot of these most
53:58critical issues, we are resetting our
54:00exams.
54:01So, you look at, you know, the power
54:02crisis that we've had going all the way
54:04back to 1984 in very successive cycles.
54:08You look at the IMF programs, bailout
54:11programs. You go all the way back to
54:121966 to 1999.
54:15Um and then you see them in periodic
54:16cycles, about 17 of them. We keep going
54:19through the same cycles.
54:23Payroll uh ghosts. I can give you a
54:26chart of how often this problem of
54:29ghosts on the payroll have come up and
54:31we've been promised final solutions and
54:33it just repeated themselves. So, our
54:35biggest problem as a country is that
54:36somehow learning is a difficulty. We are
54:39unable to learn from our problems. And
54:41the problems that emerge are not even
54:43mutating. They're the same problems, you
54:44know, just wearing a different dress
54:46from cycle to cycle.
54:48We have a term for it that I don't have
54:50time to go into which we call
54:51Katanomics. But the basic idea is that a
54:53lot of the time democracy gives you
54:55political accountability. You're able to
54:57choose your leaders. Ministers are
54:58sensitive about what they say to people
55:00cuz they know you are observing and the
55:02rest of it. But that is just one layer.
55:04The second layer is policy.
55:06And policy is the detail.
55:08And far too few citizens will always be
55:12about each policy because the policy by
55:14its very nature disaggregates. You take
55:18AI policy and all of a sudden you're
55:19talking about jobs. But then you're also
55:20talking about AI infrastructure. Then
55:22some say no, it's about energy and about
55:23this. And each time you do that, the
55:26number of people who can engage reduce
55:28in number. So, that disaggregation and
55:30aggregation effect is what makes policy
55:32accountability very different from
55:34political accountability. And in most
55:35democracies, certainly in Africa, we
55:37talk only about political accountability
55:39and not policy accountability. So, we've
55:40just heard a lot about the AI compute
55:42center that the minister is very excited
55:44about. But, we have to ask ourselves,
55:45you go back to 2006
55:47and we were building already the vision
55:50for national data centers.
55:53So, we had a tier three data center that
55:55came back later on with over 300 million
55:57dollars borrowed if you take the
55:58commercial facility and Accra facility.
56:00And we still can't agree on how much
56:01exactly, but a huge amount of money. And
56:03then you go back to the policy archives
56:05and they'll tell you that there was a
56:06time the minister will concede that
56:08NITA, which is the government's main IT
56:10agency, couldn't get access to the data
56:12centers. In fact, NITA or NITA,
56:15depending on how you pronounce it, was
56:17paying money to a private contractor and
56:19at one point it had what, barely 25
56:21cedis in its bank account and was owing
56:23about 800 million Ghana cedis. This is
56:25supposed to be the national IT brain to
56:27manage the infrastructure. Couldn't get
56:29it to work. Now, we're going to build a
56:30new AI compute center. Politically, it's
56:33amazing because that's what the people
56:35want. They want to hear that the
56:36government is making progress in solving
56:37our problems. The problem is the next
56:39layer, which is the policy layer, is
56:41about the detail. Why did the previous
56:43tier three data centers in Accra and the
56:45rest, in Kumasi and Accra, why did they
56:47fail? Have they failed? Have they not
56:49failed?
56:50Those kinds of questions require great
56:52attention to the minutia and it's
56:55fortunately not the job for the
56:57political accountability level. And I
56:59think that's the key problem that we
57:00face as a country. This inability to
57:02learn from the detail of our mistakes in
57:05the past.
57:07The belief is that AI will somehow help
57:09us to leapfrog
57:11and jump this learning process. That
57:14somehow, because of AI, we can create
57:16and do new things beyond our current
57:18capacity.
57:20Because some of the learning steps that
57:21we have to take, we may not have to take
57:22them. Because the machine somehow can
57:25generate artifacts that are perfectly
57:27made.
57:28An example [clears throat] would be a
57:29manual for discrimination or worker
57:32abuse or sexual abuse in your workplace.
57:36If you give enough description, maybe
57:38couple of paragraphs, and the AI model
57:40will generate a 100-page document. It's
57:43an artifact.
57:44That does not change
57:46the practice within your organization of
57:49implementing safe garden and ensuring
57:52that there's no abuse or discrimination.
57:53But that requires workflows to transform
57:56the workflows in your organization, not
57:57to create artifacts. So, while AI can
57:59help you to generate artifacts on
58:01demand, it does not change the internal
58:03learning and teaching. First, you have
58:05to tell the AI what you want in great
58:06detail in ways that align with your own
58:08organization. And then when the AI
58:10generates it, they have to learn whether
58:11or not what they did was useful. That
58:13learning and teaching process is
58:15extremely difficult because that is
58:16where the organizational overhead lies.
58:19And that's not what the machine does.
58:21So,
58:23I argue that we think of AI like a
58:25genie.
58:26You know, you rub the lamp, the genie
58:27pops out, you tell it what you want, and
58:29it manifests.
58:30But that's not how the AI systems that
58:32are being developed today and the ones
58:33that are the likely to make the most
58:35dramatic impact on our lives are
58:37agentic.
58:38They are organizational in nature. They
58:40do pieces bits and pieces, and you have
58:42to organize them and configure them, and
58:44working together, they try and replicate
58:46what you do. That's high management
58:48intensity. It requires a lot of
58:49understanding your own process deeply.
58:52It requires you teaching the AI.
58:54So, the AI is not a genie, it's an
58:55apprentice. It requires molding the AI
58:57to suit your organizational culture. And
58:59my argument is that without thinking of
59:01AI that way, we're going to make a big
59:03mistake because we're going to think of
59:04creating this massive wonder cave full
59:07of magical lamps where you walk in, you
59:09you know, you rub it, genies pop out,
59:11and they just do whatever you want.
59:13If you look at a lot of AI
59:14implementations in organizations today
59:16in the world,
59:17then
59:18empirical data is very emphatic. Most of
59:20them fail. AI initiatives in
59:21organizations fail. Not because the AI
59:23as a system is not extremely powerful.
59:25It's very powerful. If you define your
59:27task very well and give it discrete
59:28task, it can do amazing things. But that
59:30is rarely what you want to do.
59:32Typically, you want to change a certain
59:34process within the organization.
59:36Typically, you want to change a certain
59:37workflow within the organization. That
59:38is not something that can be generated
59:39through a simple act of fact. It
59:41requires you to change your culture. It
59:43requires you to manage your culture. It
59:44requires you to manage change. And a lot
59:46of organizations, that's where they are
59:47failing. I will argue that as a country,
59:49that is where if you're not careful, we
59:51will fail because the way we manage
59:53things, the way we talk to each other,
59:55the way we connect to each other, the
59:56way we implement policy accountability,
59:58the way we ask strong and difficult
1:00:00questions, and the way we go beyond the
1:00:02surface has been our challenge. And if
1:00:05you're not able to fix those things
1:00:06fundamentally, I don't see how AI will
1:00:07make a big difference. So, building AI
1:00:09is about a curriculum, a national
1:00:11curriculum, and a national culture.
1:00:13And in that regard, my argument is that
1:00:14if a country is struggling to learn,
1:00:17so we have flooding, we kind of throw
1:00:19our arms about, we claim we've made
1:00:21implementations of different
1:00:22implementations of change, we break a
1:00:24few things down, and then a couple of
1:00:26years again, we have another flooding.
1:00:27It's not clear what failed last time and
1:00:28who was responsible for that failure,
1:00:30and they don't lose their jobs. If
1:00:31that's the nature of our society, then
1:00:34we now have to teach AI to be more
1:00:36effective. What is the AI learning?
1:00:39You have a certain culture, and that
1:00:41culture does not necessarily learn
1:00:43effectively, and you bring an agent into
1:00:45the system, and the purpose of AI is to
1:00:47learn and imitate and align because
1:00:49something alignment in AI, that tend to
1:00:51get the AI to you know, align with the
1:00:53way you are as a person or a culture. If
1:00:55that's truly because that's how you make
1:00:56the AI safe. If the AI thinks like you,
1:00:58behaves like you, then generally it's
1:01:00not going to kill you. But if your
1:01:01culture is disorganized, and the AI is
1:01:03going to align with your culture, will
1:01:04the AI deliver outputs that are
1:01:06transformative?
1:01:07That's the philosophical question that a
1:01:08lot of people don't ask. And I will just
1:01:10argue that
1:01:11the problems that we face, and here I'm
1:01:13getting to a little bit of solutions.
1:01:15We have this issue where every year it
1:01:18takes us many, many months, sometimes
1:01:19years to to new workers into the public
1:01:21sector. We have it every single year
1:01:24new nurses, new teachers just can't be
1:01:26paid.
1:01:27Somehow you didn't know.
1:01:29We never codified it good enough. And
1:01:31the problem with AI is that if you don't
1:01:32define your problem extremely well, AI
1:01:35can't give you answers. That actually is
1:01:37the limitation of AI.
1:01:38In that sense, I argue very, very
1:01:40strongly that the inability to generate
1:01:44accountability itself is going to make
1:01:45it difficult for the data that the
1:01:46minister says he wants to connect. I'll
1:01:48use an example. Let's say all of a
1:01:49sudden, you know, the minister creates
1:01:51his regi- the minister creates his
1:01:52registry. Powerful, you know, system.
1:01:55And then you go in as a researcher and
1:01:56you want to find out galamsey related
1:01:58data linked to
1:02:00uh pollution of waters.
1:02:02First of all, somebody has to report the
1:02:03turbidity information.
1:02:05But if somebody has decided that, you
1:02:06know, it doesn't look good for the
1:02:07government, so the turbidity information
1:02:09is not going to be reported, which is
1:02:10the claim that currently environmental
1:02:12activists are making in Ghana that we
1:02:13are not reporting the turbidity
1:02:14information, so we don't know whether
1:02:15the rivers are getting better or not,
1:02:17then how are you going to use that
1:02:18connected registry to solve the problem
1:02:20of galamsey or get better or understand
1:02:22it better? Because the data is not
1:02:23there.
1:02:23And if the data is not there because
1:02:24somebody doesn't want accountability,
1:02:26then it doesn't matter how many times
1:02:28you connect, you're not going to get the
1:02:29right data that working together
1:02:31transforms the outcome.
1:02:33Last set of points.
1:02:35I do think though that the mere thought
1:02:37that you have a system that is this
1:02:38powerful and can be transformative can
1:02:40do something I call ambition buffering.
1:02:42We often go through this cycle where we
1:02:44are very gung-ho, we think we can do
1:02:46things, and then it doesn't work. AI,
1:02:48because it's system that, you know,
1:02:50doesn't really get tired, it doesn't
1:02:52suffer from our
1:02:53um social anxieties and the rest, it can
1:02:56keep up even when we're not too sure.
1:02:58And in a way that can create the basis
1:03:01and the pretext for reform, because it
1:03:03just keep going on creating new things.
1:03:05It makes experimentation cheap. It makes
1:03:07it less risky.
1:03:09I think that's one of the biggest
1:03:11opportunities if we, you know, become
1:03:13much more experimental as a country, but
1:03:14also if we track the results of the
1:03:16country.
1:03:17So in short, we have all these great
1:03:19things that the minister has talked
1:03:20talked to us about. $5 a year additional
1:03:22income.
1:03:24Um that's 5% of GDP roughly, maybe a
1:03:26little bit less now that we are 128
1:03:27billion. Got the currency improvement.
1:03:2948 registries and connecting them. He
1:03:31wants to make sure that rural urban is
1:03:33somehow aligned. Um and obviously he
1:03:36wants to build a national AI computer
1:03:37center. These are amazing things to do.
1:03:39My argument is that in each of these
1:03:42circumstance,
1:03:43you have to fundamentally think about
1:03:46the culture in your country
1:03:48that relates to performance, how you do
1:03:50things. Because AI is uniquely designed
1:03:53to try and mimic you.
1:03:55And so if you have a country where the
1:03:57data
1:03:58sets are deliberately gapping, they
1:04:01leave gaps in there because of strategic
1:04:03manipulation, then you have AI that
1:04:05hallucinates a lot because it will
1:04:07interpolate into those gaps.
1:04:10If you have a system where occasionally,
1:04:13you know, even though certain things
1:04:15have failed, we insist that we must say
1:04:18they've worked. Because in our political
1:04:20system, you know, if you say something
1:04:21has failed or something is not working,
1:04:23then you are an against.
1:04:24In that kind of system, what do you tell
1:04:25the AI emphatically about what is not
1:04:27working and what is working so that the
1:04:29AI can learn? Because what is not
1:04:30working, you say you have to tell the AI
1:04:31it's working.
1:04:32I'm a minister, I'm an an as analyst,
1:04:35and I work at the Ministry of Agric.
1:04:37And I have to be very clear and emphatic
1:04:38about what kind of data I'm feeding the
1:04:40AI system. But I also know that the
1:04:41system has been designed such that the
1:04:42academic at Legon can plug into it and
1:04:45detect whether or not our harvest are
1:04:46failing.
1:04:48Do I put in that right information?
1:04:50Do I do it in such a manner that some
1:04:51opposition person can go and collect
1:04:52that data, and having collected that
1:04:54data, embarrass my minister?
1:04:56These are difficult questions. We should
1:04:58not stay on the surface. Thank you.
1:05:00>> Mhm. Okay. Um yeah, please if you want
1:05:03to translate, please feel free to do so.
1:05:05Um let let's come back now. Uh Bright,
1:05:06thank you very much. I will now ask say
1:05:08um Sam George's thoughts.
1:05:11>> Well I think that um
1:05:13Bright makes some
1:05:15valid points, but I think that in the
1:05:18conversation about artificial
1:05:19intelligence, we need to be clear what
1:05:21is technology
1:05:23and what is
1:05:26ideology.
1:05:28A lot of the problems he raises in his
1:05:30presentation are human
1:05:34idiosyncrasies.
1:05:36Nobody is suggesting that artificial
1:05:37intelligence is coming to fix the
1:05:40cultural nuances of our people.
1:05:44We all know what the cultural
1:05:46difficulties are.
1:05:48And so, the suggestion here is not to
1:05:50say that if we roll out AI systems, the
1:05:53AI systems will fix our cultural
1:05:56nuances. So, take for example, the gaps
1:05:59in reporting data. Whether you have an
1:06:02AI system or not, if a public officer
1:06:06fails to report that data, whether it's
1:06:08an AI system or not an AI system, that's
1:06:10not right.
1:06:12So, whether you have an AI system or you
1:06:13don't have an AI system, that gap will
1:06:16exist. But, if that gap does if the data
1:06:19is reported and a business wants to use
1:06:22it, how do you then make it easier for
1:06:25that business to use it? That's where my
1:06:26job comes in as Minister for Digital
1:06:28Technology. What exists, how do I make
1:06:31it possible? If the data exists and you
1:06:34can get it through an RTI request, it
1:06:36means that that data is going to be
1:06:37accessible to any developer who needs
1:06:40that data. So, the question here is not
1:06:42about whether AI is going to cure the
1:06:46human inefficiencies.
1:06:48No. AI amplifies. AI is an accelerator.
1:06:52It it is meant to give us amplification
1:06:55for the things that we can do.
1:06:58And so, take for example, NITDA, the
1:07:00National IT Agency, which is one of 11
1:07:02agencies under my ministry that you
1:07:04referenced.
1:07:06When I took over office in February,
1:07:09the system, in fact, the reference you
1:07:11make is from a public statement I made
1:07:14about the state of NITA,
1:07:15what it had been.
1:07:17Today, what is the state of NITA? NITA
1:07:19has absolute 100% control of our data
1:07:22center.
1:07:23Because, again, it comes back to and
1:07:26that's why I agree with you, political
1:07:27will and what policy is.
1:07:30We can choose to look at what is flowery
1:07:32policy and what is hard work. And I
1:07:35mean, I've come under a lot of
1:07:38flak. I've taken a lot of flak for some
1:07:39of the policy decisions I've taken. But,
1:07:42when you look at the work we're trying
1:07:43to do with our legislative review, be it
1:07:46cybersecurity, data protection, or NITA
1:07:48itself, a lot of the things we're trying
1:07:50to fix
1:07:52or the legislations we're trying to put
1:07:53in will fix the systemic gaps and
1:07:56loopholes you speak about in our system.
1:07:59This morning, I was at Gold Board
1:08:01and I made a public statement that from
1:08:03next year, every government agency that
1:08:05doesn't have a certified data protection
1:08:07officer and a certified cyber CISO will
1:08:10get a fine from the regulator. We've
1:08:12started finding government agencies.
1:08:14I've I'm I've fined a few private sector
1:08:16agencies, but government must fix
1:08:18itself. And so, if you do not have a
1:08:21regulated system and people are allowed,
1:08:24for example, reporting of government
1:08:27data, there is no framework around it.
1:08:29>> Does the honorable minister know that
1:08:30the data protection
1:08:32center commissions register has been
1:08:34down for weeks? You can't check the
1:08:36register.
1:08:37>> No. So, again, there's a reason why that
1:08:39has happened. Because that register was
1:08:42funded by a World Bank facility that has
1:08:45ended. And so, you have a contractor who
1:08:48was paid by the World Bank, again,
1:08:51preceded us coming into government, but
1:08:52the way that contract was signed, if a
1:08:55contractor offers a service to a
1:08:57government agency, the ownership of that
1:08:59registry should not reside with the
1:09:01contractor. And so just yesterday I
1:09:03issued a directive to the company. We've
1:09:05given the company 14 days to return all
1:09:08assets of the registry to data
1:09:11protection or the economic and organized
1:09:13crime office will begin an investigation
1:09:14into the committee that company. So the
1:09:16reason why it is down is because of the
1:09:18modus or the mode in which the contract
1:09:21for that arrangement was made in the
1:09:24past.
1:09:24>> What has the commission informed us?
1:09:27>> So the commission is so the commission
1:09:29would not come to the public and say we
1:09:31have a problem with our service provider
1:09:34because the previous government had done
1:09:35that. The commission will write to his
1:09:36minister.
1:09:37And the minister will take action, which
1:09:39is what I'm telling you I took
1:09:40yesterday. Where we wrote and that's why
1:09:42I've said when they brought me the
1:09:43details, I wrote to the company and
1:09:44we've written to them and given them 14
1:09:46days. We said, "Look, you had a
1:09:47contract. Your contract was to build a
1:09:49registry for the data protection
1:09:51commission."
1:09:52Now, whatever is contained in that
1:09:54registry is not the property of the
1:09:56company. You were paid for it.
1:09:58Your contract has elapsed. You say you
1:10:00want a renewal. I have the right to
1:10:02choose to continue to have a maintenance
1:10:04service with you contract with you or
1:10:06not. If I choose to use a different
1:10:07service provider, you cannot say that
1:10:09you are It's just like what happened
1:10:11with the e-health.
1:10:13E-health. You understand me? It's like
1:10:15what happened with the data center. So
1:10:17when you inherit some of this because
1:10:18the data center was built in 2015 by the
1:10:21previous Muhammad administration, a
1:10:23privatization contract was signed that
1:10:25took all of government's $300 million
1:10:27investment and put under a private
1:10:29sector company.
1:10:30>> But would you be surprised if I told you
1:10:32that the new attempt to replace the
1:10:34e-health system with a government owned
1:10:36system hasn't fixed the fundamental
1:10:38issue of vendor locking? Would you be
1:10:39surprised if I told you that?
1:10:40>> Well, when that project was going to be
1:10:42done, I wrote to the Minister of Health
1:10:44and demanded that whatever contract
1:10:46agreement was going to be signed will be
1:10:48brought to us. We It was referred to
1:10:50NITA and data protection and we changed
1:10:52the terms of the contract. So except the
1:10:55terms of the contract as mandated by
1:10:57NITA. And I'm sure you've seen that
1:10:59there's a cabinet directive to all
1:11:00government agencies for a technical
1:11:02clearance from NITA for all government
1:11:03IT procurements.
1:11:05>> Mhm.
1:11:05>> So, right now, every government IT
1:11:06procurement comes to NITA for us to look
1:11:08at the terms of it. It goes to NITA
1:11:10cybersecurity and data protection. And
1:11:12so, in that particular LIMS arrangement
1:11:14by the Ministry of Health, we made
1:11:16changes to the data ownership rights in
1:11:19that contract. So, if I I I cannot speak
1:11:22to what's going to roll out Publican.
1:11:23They brought that to us. The original
1:11:25contract for Publican had the data that
1:11:28was generated sitting outside of Ghana.
1:11:30>> Mhm.
1:11:30>> We insisted that that should be changed.
1:11:32We all can testify to what Publican is
1:11:34doing for us. $300 million a month in
1:11:37increased tax revenues. The data is
1:11:39resident here, and we're keeping it
1:11:41here.
1:11:41>> way, that's an AI model for uh
1:11:44revenue collection.
1:11:44>> Revenue collection. So, so, you see,
1:11:46Bright, like I said in my initial
1:11:48submission, I will not sit here and say
1:11:49that with
1:11:51a one single policy directive, you can
1:11:54fix the entire ecosystem. But, we'll
1:11:56keep fixing this one step at a time. And
1:11:58this is driven by the national AI
1:11:59policy, which is a 10-year plan, which
1:12:02has measurable goals. Just this morning,
1:12:04the presidency wrote back to me with my
1:12:07after I sent a request, cuz if you look
1:12:08at the strategy, we talk about having a
1:12:10responsible AI office. The presidency
1:12:12has given me authorization to establish
1:12:14that office and establish the AI
1:12:15council, which is going to have private
1:12:17sector and government sector players sit
1:12:19on that AI council to govern
1:12:20government's deployment of AI. So, we
1:12:23are building the blocks. What we are
1:12:25doing today, with give or take a paltry
1:12:28budget of $270 million, is
1:12:31is less than
1:12:33one private company's investment in a
1:12:35month
1:12:37in AI. So, we don't have necessarily
1:12:40what the financial muscle to do all the
1:12:42things we want to do. But, the question
1:12:43is, do we have a vision of where we are
1:12:46going? Do we have a road map that is
1:12:48testable,
1:12:50auditable, and workable?
1:12:51>> Mhm.
1:12:52>> Can we commit to it such that whether
1:12:54some judge sits there
1:12:56or it is another person in the NDC or
1:12:59it's even a different political party,
1:13:01there is this roadmap that we all signed
1:13:03onto and committed to. And I saw
1:13:05something in your in your in your slides
1:13:07which I agree 100% with.
1:13:10If a minister is trained in on AI, which
1:13:13we have done,
1:13:14and his directors and the technical
1:13:16staff under him don't understand what it
1:13:18it is that he's doing, we've done
1:13:20nothing. And that's why right now as we
1:13:22speak, we've trained all the ministers,
1:13:24we've trained all the chief directors,
1:13:25we're training directors and deputy
1:13:26directors at the moment. The idea is to
1:13:28train every single person. We have We've
1:13:31established AI focal persons in every
1:13:32ministry. So, you you are basically
1:13:35working the paradigm shift you talk
1:13:37about, but it is one step at a time. I
1:13:40wish there was a magic wand that will
1:13:41fix all of the human inefficiencies with
1:13:44artificial intelligence. But again, that
1:13:46would be wishful thinking. So, as we fix
1:13:49the human inefficiencies with good
1:13:51governance practices, we must make sure
1:13:53that the people like
1:13:56um Alex and and Ghanaian private
1:13:59industry that need access to existing
1:14:02data. Because yes, we will have
1:14:03inefficiencies, but we have a treasure
1:14:05trove of of of information as well
1:14:07that's sitting there. For example, the
1:14:08work that they did with ECG was simply
1:14:11based on the amount of data that they
1:14:13had access to at ECG. But if they were
1:14:16to replicate that with Ghana Water, they
1:14:18would have to go through the same cycle
1:14:20again to generate that kind of data set.
1:14:23But if through the Utilities Regulatory
1:14:24Commission, all of these utilities
1:14:26submit their their their their data to
1:14:28that to their own registry which already
1:14:30exists. These registries exist already.
1:14:33Then you make it easy for them as a tech
1:14:35developer to access that. So, we will
1:14:37continue to work
1:14:39with as government, but largely with
1:14:41private sector. And I made that point in
1:14:43in the documentary that this is not a
1:14:45journey for just government alone. But I
1:14:47think that we must have a more
1:14:50optimistic approach to this.
1:14:53Pessimism is is a Ghanaian thing
1:14:56where we think that everything will
1:14:58fail. But, I believe that if for once we
1:15:01we put aside all of our differences and
1:15:04point out where our strengths are and
1:15:06what we can do together, leverage on the
1:15:08successes and strengths of our private
1:15:10sector
1:15:11and work collaboratively from
1:15:13government, we will be able to leap from
1:15:15because financial inclusion,
1:15:17which is in your in your presentation,
1:15:20was not overnight by private sector.
1:15:22There were private sector players who
1:15:24were just waiting for the right opening
1:15:27to blow up our financial sector space.
1:15:29When things like interoperability
1:15:31clicked in for them, which then took it
1:15:34dropped the bar for them as financial
1:15:36sector players, what did you see? You
1:15:38saw the boom. So, one right government
1:15:40policy,
1:15:41private sector on top of that, you have
1:15:43a boom.
1:15:43>> move. I I want to I want to bring others
1:15:45into Wayne. I think one of the thing
1:15:46that is clear to me from the bright
1:15:48presentation what Minister said is that
1:15:50a lot of fundamentals we need to fix
1:15:52first before we begin to build on it.
1:15:55Um if we don't fix the fundamentals,
1:15:57something would would happen that would
1:15:59lead to a lot of investment that doesn't
1:16:01lead to the outcomes. And and Adam, I'll
1:16:02come to you because you you consulted
1:16:04this. So, looking at a country like
1:16:06Ghana, what will it actually take to
1:16:08build an AI solution that works for us?
1:16:12Considering everything of that bright
1:16:13has just told us that the Minister has
1:16:14said we are trying to fix. What would it
1:16:16take to to build one that works for us?
1:16:17>> Yeah, thank you, Evans. First of all,
1:16:19thank you for having me. Um and then
1:16:20thank you for sitting me in between the
1:16:22honorable Minister
1:16:23>> [laughter]
1:16:24>> and and Bright because I've been
1:16:25following them both. So, to be sitting
1:16:27here in between them is like I'm I'm a
1:16:29front row fanboy. So, thank you both for
1:16:32uh for that for that dialogue. Um look,
1:16:34I
1:16:34look, these are all great conversations
1:16:36to have, right? Macro, big policy
1:16:39things. It It's not It's not my
1:16:40specialty. I don't I'm not a builder. Um
1:16:42I'm I'm not a a developer. I'm an
1:16:44applied AI practitioner and I'm an
1:16:46applied AI consultant. And the people
1:16:49that I work with on the ground, right?
1:16:50I'm talking about Uber drivers. I'm
1:16:52talking about the momo stand. I'm
1:16:54talking about the the person sitting in
1:16:57a bank or even, you know, in some of the
1:17:00state-owned businesses, right? My
1:17:02concern because as I said earlier that
1:17:04opportunity and disruption are the same
1:17:05event that have a different perspective.
1:17:07There's one disruption that is real,
1:17:11which is exclusion. And that's my
1:17:13concern and that's where AI for Ghana
1:17:15fits in. And and it's the exclusion of
1:17:17individuals. I'm not talking about
1:17:19informal versus formal economy. I'm
1:17:21talking about as we develop and have
1:17:23these conversations about these
1:17:24wonderful things that that, you know,
1:17:26the the ministry and the government and
1:17:28you know, everybody on the esteemed
1:17:29panel is talking about over the next 3 5
1:17:3210 years.
1:17:33My concern and where my focus is is on
1:17:35the individual and what they can do
1:17:37today with the phone in their hand and
1:17:40for free. And I'll I'll I love the
1:17:42honorable minister's story about the the
1:17:44stenographer, was it your hand? I love
1:17:46that because, [snorts]
1:17:47you know, you can go back to the
1:17:49industrial revolution and talk about,
1:17:50you know, how people were scared about
1:17:52the lightbulb or the combustible engine
1:17:53and and the opportunities that came from
1:17:55that. I'm sure that, you know, some some
1:17:57candle makers lost their job when the
1:17:58lightbulb came around or you can look
1:17:59even more recent to, you know, the momo
1:18:01vendors that, you know, now to be a
1:18:03banker on the street, you know, you need
1:18:05a cell phone and you know,
1:18:08a shipping container, right? So, it's it
1:18:10you were in a moment where all these
1:18:12opportunities are coming about and I'll
1:18:14give you a couple of brief examples.
1:18:16One person I work with,
1:18:18a content creator, Francisca, you know,
1:18:20she works a 9-5. She has her own job.
1:18:24She had aspirations to do more, right?
1:18:27She was able to create enough content
1:18:30and this was using AI. You know, with
1:18:32AI. She was using AI and it was taking
1:18:34her a full day spread across a week in
1:18:37her mornings and in her evenings to
1:18:39create content for one client. After one
1:18:41of our sessions, after organizing her
1:18:42tool, she got that 8 hours of work down
1:18:45to 30 minutes. So, is it disruption or
1:18:47is it opportunity? Well, what she did
1:18:49was she redeployed her time, and now
1:18:51she's servicing three clients. So, where
1:18:52she had one, now she has three.
1:18:54Fatau, he's an air conditioning uh
1:18:56technician. In his slowest month, I'm
1:18:58like, "Hey, what's going on, man? You
1:18:59know, how's business doing?" He said,
1:19:00"Oh, it's slow. It's the rainy season.
1:19:02Business isn't coming in." We put
1:19:04together a game plan that had him
1:19:07uh use a prompt to go through his past
1:19:09contacts. By sending 10 messages to 10
1:19:12just 10 of his clients, he got six
1:19:14responses. He got four jobs that created
1:19:162,000 CDs in on top of his month, his
1:19:19slowest month. Right? The one that's my
1:19:21favorite that I'll end on is Ibrahim.
1:19:23Ibrahim works at a gate, opening a gate.
1:19:26He came from the central region. He's
1:19:28Fanti speaker. His uh his uh mother had
1:19:31passed away. His father is sick. You
1:19:32know, like so many other Ghanaians, he's
1:19:34the eldest. So, everything falls on him.
1:19:36Um he he opens a gate for a living. Now,
1:19:39this man, right, with me by his side,
1:19:41we've walked to a place where he has
1:19:43uh learned coding, right? Uh through his
1:19:46chat, okay, through his AI tool, he's
1:19:48enrolled himself in the 1 million coders
1:19:50program. And the best part on it, on top
1:19:52of that, he's brought others into it.
1:19:54Cuz I I saw recently the honorable
1:19:55minister asked for people to complete
1:19:57the program and to bring others. He's uh
1:19:59100% on the first two. He's 87% on the
1:20:01third. And he just went to a momo vendor
1:20:03and he got them enrolled in the program
1:20:05and set up on his chat GPT a project on
1:20:07how to train himself how to do that. At
1:20:09the same time, that young man who worked
1:20:10at a gate is now a paid applied AI
1:20:13consultant for AI for Ghana.
1:20:14>> I should know.
1:20:15>> So, you tell me, is that disruption or
1:20:17is that opportunity?
1:20:18>> Uh definitely an opportunity. But let me
1:20:19prove that a bit more. So, that's one of
1:20:21the uh government policies that you just
1:20:23testified is working because this person
1:20:26tapped into 1 million coders that I know
1:20:28um some George has been very loud about.
1:20:30Some George, so I guess we'll come to
1:20:31that and say, how did that benefit and
1:20:33trickle down to more people. But, which
1:20:35AI models were three individuals and I
1:20:38like your examples because I want this
1:20:40to be about the real people.
1:20:41>> Yeah.
1:20:41>> It's the real about the real people.
1:20:42Because if we don't fix get AI that
1:20:44benefits most people, we possibly as
1:20:47Alex said in the documentary,
1:20:49a top few companies will begin to sort
1:20:51of enjoy the vast majority of us will
1:20:54never see the benefit.
1:20:55>> Mhm.
1:20:55>> What AI models were they using? Are
1:20:57these developed in Ghana or these are
1:20:59external models that we all tap into?
1:21:01>> Yeah. So, if you allow me the liberty of
1:21:03just saying first that you know, you may
1:21:05see a brony here, right? But, I was
1:21:07married here in 2015.
1:21:10My my wife is Ghanaian. My mother-in-law
1:21:12is Ghanaian. My three children have
1:21:14Ghanaian blood. Ghana's future is my
1:21:16future. So, I want to say that first of
1:21:18all. I'm not some outsider who came here
1:21:19just for some little weekend getaway.
1:21:21I'm here to stay. So, the things I do
1:21:24are for Ghana's future because they're
1:21:25my future. I want to say that first. The
1:21:27tools we used are the western ones
1:21:28because they're the ones that work. But,
1:21:30if you look, we just published a video
1:21:31this past Sunday on AI for Ghana that
1:21:33speaks about that gap between the native
1:21:36languages, the native language models,
1:21:38and the western ones. I am an ally of
1:21:40them being built. I'm not a developer. I
1:21:42don't know how to develop them. But, as
1:21:44soon as there is one that can speak, you
1:21:47know, let me back up. In this video, I
1:21:50went to a corner. It's not a study, you
1:21:52know, it wasn't a large enough sample,
1:21:54but I spoke to eight people, okay? They
1:21:55all spoke tree. And of the eight people
1:21:57who spoke tree, only four of them could
1:22:00read and write tree. The other four
1:22:02could speak English some to some degree
1:22:05or write English to some degree. So,
1:22:07their access to the tool was their
1:22:09voice. That's the way they get into the
1:22:11tool, but the tools right now, okay, the
1:22:13ones that that I've experimented with
1:22:15and used, they're not ready to be able
1:22:17to hear somebody on the street and
1:22:18understand them at the first time and do
1:22:20the work. And the moment that that tool
1:22:22is available will be the tool that we
1:22:23use and we teach all the time.
1:22:24>> Okay. Is that consensus on the table
1:22:27that is great and after the example is a
1:22:29good one. I'll get some coming to you.
1:22:31Alex said in the video you tried to
1:22:33build your own.
1:22:34>> Right.
1:22:34>> You tried it, you rolled in, you pulled
1:22:36in the money, but after a while you
1:22:38abandoned it and then you sort of buying
1:22:40tokens and he's telling you
1:22:43he's using foreign ones to help
1:22:45Ghanaians because it works. And on the
1:22:47day we get one for ourselves, we'll jump
1:22:49on it. Minister says we're trying to
1:22:51build that.
1:22:52Bright says you can build it but you
1:22:54have to fix the fundamentals. For you,
1:22:57draw [clears throat] from that
1:22:58experience for me when you when you
1:22:59abandoned it. Which will better serve
1:23:03your customers? One that you sort of
1:23:06commercially purchased, built from
1:23:08abroad with data from abroad, or one
1:23:10that we invest to build locally, by
1:23:12Ghanaians, for Ghanaians, built on
1:23:13Ghanaian data that understands it.
1:23:15Bright says it's very important. He says
1:23:16it's designed to mimic you. Right? Which
1:23:19one for from a businessman's point of
1:23:22view, you run a business, have
1:23:22customers, will better serve your
1:23:24customers?
1:23:25>> So, even if you don't
1:23:26>> I've been trying it for a while.
1:23:27>> Right. It depends on the context and
1:23:30then um
1:23:32let me put it this way as an example.
1:23:36Let's say you
1:23:38you want to hire a chef, right?
1:23:41Would you Would you rather go in for an
1:23:45Italian chef to do your local fufu
1:23:49and groundnut soup
1:23:51or you would go in for an equally
1:23:54competent Ghanaian chef
1:23:56who has lived here
1:23:58for more than I mean their whole life.
1:24:01They've been to farms, they know the the
1:24:03feel and taste and the texture and the
1:24:07knowledge and experience behind that
1:24:09meal that you want.
1:24:11Right? We all want the good meals.
1:24:14But unfortunately, sometimes the
1:24:17resources we need
1:24:19to get that prepared for us it's not
1:24:22readily available.
1:24:24>> Mhm.
1:24:24>> Why? Because
1:24:26um there is inequality. Um there is um
1:24:31access to those resources, there's cost.
1:24:33And there's time of develop time to
1:24:36develop those kind of expertise that you
1:24:38need.
1:24:39Um and so you need to take a step back
1:24:42and then make that
1:24:45decision
1:24:46that okay
1:24:48>> Obviously, the value here is that you
1:24:50still want to satisfy your hunger.
1:24:51You're hungry.
1:24:52>> Yeah.
1:24:52>> Right? So, in the meantime, you could go
1:24:56in for
1:24:58that chef that can get you something to
1:25:01satisfy your cravings, but in the long
1:25:04term you should have a plan that okay,
1:25:07I'm going to do ABC, but I can't do it
1:25:09alone.
1:25:10Even with all the money I have, I cannot
1:25:12do it alone. Let me go contact Busia
1:25:15Pinyin and his Busia Pinyin and all the
1:25:17family people in this case, which is the
1:25:18government. Right? Because they can
1:25:21break that ice.
1:25:23Right? At that level. They can make it
1:25:25happen at that level. And then now we
1:25:27can feed on top of the platforms that
1:25:30they will build. So, that's the same um
1:25:33situation that we were faced then.
1:25:36Right? We We hired very competent
1:25:38Ghanaians. I mean, some of them were
1:25:40even hired on the spot. They were
1:25:42brilliant. Um somebody has said talent
1:25:45is is um equally distributed, but not
1:25:47the resources, right? Not the
1:25:51So, it it it it appears that we were
1:25:54stuck in advancing
1:25:57the vision that we had
1:25:59because resources were lacking. It
1:26:01doesn't make sense to put in all of our
1:26:04money. We're just We cannot compete at
1:26:06that level. Right? So, these so-called
1:26:09frontier companies, they had gone
1:26:12terribly wild
1:26:14and they've been able to develop these
1:26:16beautiful systems, these intelligent
1:26:18things, systems, and scale it down and
1:26:22distribute it at a very cheap price.
1:26:25Cheap as the company's, you know,
1:26:27purview, right? $200 a month, $100 a
1:26:30month is it's not really anything. But
1:26:33the cost that would have come to us if
1:26:35we had done the same thing,
1:26:38I'm sure we'll be on our knees by now.
1:26:39But this is where
1:26:42government comes in.
1:26:44to
1:26:45create that platform.
1:26:47But I mean, with what right Simo also
1:26:51said, we we have to look at the the
1:26:54other sides, right?
1:26:57What are the building blocks behind
1:26:59getting these AI systems homegrown?
1:27:02We need to critically look at the data,
1:27:04how it's fragmented, how we have to
1:27:06structure it. We need to do some
1:27:08homework.
1:27:09So when we get that done properly,
1:27:11then other companies like ourselves and
1:27:15the very But I was going to say
1:27:16>> say, which of these models are you going
1:27:17to prefer? A locally built uh
1:27:19>> Certainly, if it was if it was that
1:27:21available, certainly. We have very
1:27:23bright minds in Ghana. We have it. These
1:27:25AI systems are developed by human
1:27:27beings, ordinary human beings like me
1:27:29and you. They went to school. Yes, they
1:27:31pay more attention, they're more
1:27:33disciplined. They are they know their
1:27:35stuff.
1:27:36We also know our stuff. If the
1:27:38opportunity is there, we can do it here,
1:27:39too. It is it is it is really a
1:27:42blessing.
1:27:42>> want to quickly hear his example because
1:27:43he's an industrial person. What what
1:27:44show your experience on this because the
1:27:46conversation right now is about um how
1:27:48do we build one that works for us? Um
1:27:53I guess the question is can we continue
1:27:55relying on the models that we didn't
1:27:58build, we didn't sort of uh teach it?
1:28:01Righty saying um AI models and I I took
1:28:03a shot of this cuz I It It says a
1:28:06country
1:28:07Can a country that struggles to learn
1:28:09become a good teacher? And you have to
1:28:11teach your models.
1:28:13What's your experience at at Zipline?
1:28:15What what have you done it?
1:28:18>> I mean
1:28:19maybe I'll give you a little bit of
1:28:20context also how I am looking at it.
1:28:22Because in terms of thinking of about
1:28:24these things on a macro level, I think
1:28:26we have
1:28:28better suited people here, right? But
1:28:30I'm at heart I'm an entrepreneur. Aside
1:28:32of Zipline, I invest into very early
1:28:34stage startups.
1:28:35And
1:28:37that's the core focus of what I do. It's
1:28:39like built from very small.
1:28:42And when you look at that all the time
1:28:46when you want to build something, let's
1:28:48say locally
1:28:49the source is to go really small and
1:28:51have a differentiation. So ultimately
1:28:53you look at his example with the cook,
1:28:56right?
1:28:57And that's the only way that I see that
1:28:59you can build competitive advantage. And
1:29:01when you go
1:29:03too high up, of course, the public
1:29:04sector is different, but generally
1:29:05speaking if you go too high up, what
1:29:07happens is
1:29:08you will
1:29:11I'm trying not to be too radical, but
1:29:12you will fail.
1:29:14And the only way you can compete because
1:29:17it's the same for a startup and it's the
1:29:18same maybe if you compare Ghana to
1:29:20another country, you have a certain
1:29:21disadvantage.
1:29:23You cannot just jump and go and compete
1:29:26with somebody who has all the
1:29:27advantages. You have to find your source
1:29:30of differentiation, go small, and solve
1:29:32something very narrowly, and then scale
1:29:35up. When you scale up, what will happen
1:29:37to you is that your solution will be
1:29:39differentiated from other people and
1:29:41they cannot come in and compete. I mean,
1:29:42I've invested into a logistics company
1:29:44that is
1:29:45um of course, tech-enabled logistics
1:29:48company in Ghana. It's going pretty
1:29:50well. What happens to that company when
1:29:52Amazon looks it's a fulfillment company.
1:29:54When Amazon wants to enter the market,
1:29:56if they do it the right way, Amazon
1:29:58cannot compete in this market because
1:30:00there are certain things that work even
1:30:02in the informal sector that you have to
1:30:04leverage the technology, integrate, and
1:30:07make that a core.
1:30:08That being said, of course, there is
1:30:10certain problems that that's how I see
1:30:11it. It's like where of course the
1:30:13government has to come in
1:30:14and provide a basis.
1:30:16And I think
1:30:17the minister mentioned that this is a
1:30:20collective effort. And I think what
1:30:22matters to me is that I see that there
1:30:24is a strategy that is
1:30:2610 years out, that there is ambition.
1:30:29When I read it as an
1:30:30entrepreneur, I saw that there should be
1:30:34I think 10 unicorns by 2035.
1:30:38I mean, in my estimation
1:30:40we can have more than that.
1:30:41>> What is a unicorn by the way? I mean,
1:30:43for me and mortals like
1:30:44>> Is it Is it a
1:30:45Is it a billion-dollar company held in a
1:30:47in a private market, right?
1:30:50A billion-dollar company. But I think
1:30:52the the
1:30:52>> So I was saying you were saying we we
1:30:53want 10 unicorns.
1:30:55>> Yes, as per the strategy by 2035.
1:30:58>> 10 Ghanaian
1:30:59>> com- companies
1:31:00>> companies
1:31:01>> that are worth a billion dollars. A
1:31:02billion dollars by 20
1:31:04>> 2035.
1:31:06>> Ooh.
1:31:07>> I think it's ambitious, but I think it
1:31:08can be done. But I think the one thing
1:31:10that I heard
1:31:10>> Entrepreneurs say it can be done. Who am
1:31:12I to say otherwise?
1:31:13I mean
1:31:14>> I mean, I heard the CEO the Hop CEO
1:31:16saying that the big companies will
1:31:17benefit. I think the biggest companies
1:31:19that will exist are probably not yet
1:31:21founded even by even in Ghana within
1:31:232035.
1:31:24And I think that's where it comes in.
1:31:26For me as an entrepreneur is is like
1:31:29what do we do with the talent? I mean,
1:31:30there is the broad
1:31:32>> By the way, billion dollars you're
1:31:33talking about billion cities.
1:31:34>> Yes. Yes.
1:31:35>> Okay.
1:31:36>> Yes. I mean, you have to think around
1:31:38it. Yes, we see all these companies, but
1:31:40there will be new companies let's say
1:31:42even a lawyer a lawyer will come in
1:31:44Ghanaian lawyer
1:31:46AI or digital native grew up with AI is
1:31:50building his company from day one with
1:31:53AI embedded in his business model.
1:31:57It It will not be easy for the top law
1:32:01firm in the country to switch and
1:32:03compete. I mean, there will be certain
1:32:04advantages that they have. So,
1:32:07I think that's a challenge and I guess
1:32:09when I look at the strategy, what is
1:32:11trying to be done is to
1:32:14give kind of the the systems, the
1:32:17requirement, the data integrity at a
1:32:19broad level that people can tap into and
1:32:22leverage. And I think the same goes with
1:32:251 million coders, right? We need the
1:32:26talent, we need the broad base, but then
1:32:28we also need the people at the very top,
1:32:30like real disruptive entrepreneurs can
1:32:32only be a few. And
1:32:35at the end of the day,
1:32:37that's how I am looking at it is like
1:32:39how can we support the people that can
1:32:42build those disruptive companies? How do
1:32:44we ensure they don't go abroad? How do
1:32:47we ensure they don't choose another
1:32:49another career because maybe they don't
1:32:51see the opportunities within our within
1:32:54our market. And I think the it's an era
1:32:57of the thinkers. The people that really
1:32:59can think will be able to do so much.
1:33:03So, I think the gap will not necessarily
1:33:06be in who is
1:33:08big now, but the gap will be in who has
1:33:12the actual capacity to be creative and
1:33:15think. And I think that's where
1:33:18if I'm very honest, that's where I see
1:33:20the challenge in terms of how do we
1:33:22ensure that the young ones
1:33:25have that kind of capacity and we push
1:33:27them.
1:33:27>> I see. Right.
1:33:29So, I just want to touch on a few points
1:33:31that have been made that I think are
1:33:32critical.
1:33:33One is this metaphor of the cook. It's
1:33:36incredible metaphor. So, thanks for
1:33:37sharing it with us.
1:33:39And I like to be empirical. I just like
1:33:41to, you know, look at data and really
1:33:42observe as opposed to make assumptions.
1:33:45If you went around Ghana today and went
1:33:47to all the five-star and four-star
1:33:48hotels, 80% of them are foreign chefs.
1:33:51The reason being that it's not all the
1:33:53people who can afford those places who
1:33:55are Ghanaian. And what And even when
1:33:56they are Ghanaian, they want
1:33:58uh the the kontomire to be the specific
1:34:00village type of kontomire. Right? They
1:34:03develop fusion tastes because they are
1:34:05paying 300 cedis for the same palava
1:34:07sauce that they can get 20 cedis or 30
1:34:09cedis somewhere else. They don't just
1:34:10want normal palava sauce. They want
1:34:12something else.
1:34:13If you go down your, you know,
1:34:15neighborhood, you'll find a chop bar
1:34:16where the chefs are all Ghanaian.
1:34:19So, you have to ask yourself, what is
1:34:20your AI compute for?
1:34:22Are you building it so that Haptel can
1:34:24become an African beating company?
1:34:28Or you are beating it You are building
1:34:29it so that Haptel can serve a few momo
1:34:31agents in Ghana.
1:34:33If Haptel wants to serve global
1:34:35companies who have incredible
1:34:37performance standards operating out of
1:34:39South Africa or Kenya, then Haptel wants
1:34:41the best performing data center
1:34:42possible.
1:34:43And if that's AWS in Fresno, it's AWS in
1:34:46Fresno.
1:34:47These realities are harsh, they are
1:34:49cold, and you need people like me to say
1:34:51them because a lot of people are not
1:34:52comfortable rocking the boat. So, that
1:34:54is the the evolutionary role that people
1:34:56like me play. So, even though if the
1:34:58minister is here, it's important that
1:34:59people like me say to his face and say
1:35:01if you want to build an AI compute
1:35:02center in Ghana, here are the
1:35:04constraints. Don't hide from the
1:35:06constraints. One, power.
1:35:08Nobody builds an AI center unless the
1:35:11energy say, you know, the price of
1:35:13energy is roughly 3 cents to 5 cents.
1:35:16So, if Ghana does it from 10 cents to 12
1:35:18cents, that AI data computing center
1:35:21will not be efficient. It's as simple as
1:35:22that. You have to charge more.
1:35:24If it doesn't charge more, the
1:35:25government will have to subsidize it.
1:35:27And then we come back to the usual
1:35:28thing, "Oh, ECG doesn't work. Oh, what
1:35:30Ghana Water doesn't work. When are we
1:35:31going to privatize it?" So, my argument
1:35:33is let's learn from history.
1:35:35Let's do these things analytically, open
1:35:37up,
1:35:38ask think tanks to submit memos, debate
1:35:41it rigorously, finally choose something
1:35:43that is well
1:35:45grounded and robust. In this country, we
1:35:47don't like that. We want all of us to
1:35:49praise the minister, to praise the
1:35:50ministry. You say, "Oh, this is so
1:35:52brilliant." It doesn't work. In serious
1:35:54countries, you get serious debate about
1:35:56everything.
1:35:57And this is even in countries that we
1:35:59consider non-democratic like China and
1:36:00Singapore. Go and look at their
1:36:01journals. How much they take ministers
1:36:03on and question everything and they they
1:36:04do the numbers.
1:36:06Tell me you need water.
1:36:07This is a country where we had a
1:36:09desalination plant. We couldn't manage
1:36:10the desalination plant. We got into big
1:36:12trouble, etc. The amount of water you
1:36:14need for a big data center of that sort
1:36:16is unbelievable.
1:36:17Even in terms of the power itself, even
1:36:20if you had the pricing rights, the
1:36:23actual availability of power in a
1:36:24country where we were struggling with,
1:36:26you know, so-called capacity issues,
1:36:28it's important. You have to measure that
1:36:29as well.
1:36:30Thirdly, you have a lot of policy
1:36:32history to learn from.
1:36:34This country tried to create email
1:36:36services for government, built in Ghana.
1:36:39We tried to create payroll software uh
1:36:42uh built in Ghana. My one of my best
1:36:43friends ever, Herman Chinery-Hesse, he
1:36:45fought wars in this country. He tried to
1:36:48get us to use our own local software for
1:36:50payroll management, email services, and
1:36:52the rest. And he failed.
1:36:55So, the question is why did Herman fail?
1:36:57And why would the honorable the right
1:36:58honorable Sam George succeed? I want to
1:37:00understand that dynamic. What has
1:37:02changed? What is new? What, you know,
1:37:05what what new ideas has completely
1:37:08transformed that ecosystem? Okay. If we
1:37:10don't get that, [clears throat] to me,
1:37:11I'm not convinced. That's the problem.
1:37:13It's not that I'm pessimistic. I'm It's
1:37:14that I'm empirical. I'm pragmatic and
1:37:17I'm realist. Last point, which I think
1:37:19is also important, is that we talk a lot
1:37:20about, you know, the disruption element
1:37:22and uh the opportunity element.
1:37:25In the same community,
1:37:27and from time to time, there will be
1:37:30changes.
1:37:31You know, it's anisotropic.
1:37:33So, what happened with communication
1:37:35centers in this country was that if you
1:37:37went to communication centers in the old
1:37:38days, where you went to go there, you
1:37:39pay money, you made a call, and they
1:37:41gave it to you. A lot of them were lower
1:37:43middle class people.
1:37:45When that was disrupted by mobile
1:37:47phones,
1:37:48all the lower middle class segment, they
1:37:50didn't get jobs in a similar area. Those
1:37:53just migrated to the lower and working
1:37:54classes through MoMo agencies.
1:37:57So, when you go to circle and you see a
1:37:59MoMo agent, 85% of the time, they don't
1:38:02belong to the middle class.
1:38:03But, that was not the case with
1:38:04communication centers. So, we need to
1:38:06look at it not just in terms of view of,
1:38:08you know, who is suffering and the rest.
1:38:08Just look at it in terms of class
1:38:10differences. If the middle class in this
1:38:12country is threatened because their
1:38:13little accounting shops and the rest are
1:38:14being disrupted by AI more so than the
1:38:17people that are doing MoMo agency on the
1:38:19street, you have a different political
1:38:20problem than if it's those doing MoMo
1:38:22agency on the street who are disrupted.
1:38:25>> So, uh Sammy you know, come to you. I
1:38:27definitely. I want to take a quick
1:38:28break, but I think you laid out the
1:38:30foundation when I when I return after
1:38:31Sammy George had had a bite at the big
1:38:34questions that you asked. Is then to
1:38:36throw out the question then and and uh
1:38:38we have a a very strong AI skeptic in
1:38:41the audience who was in the documentary
1:38:43echo some of the sentiments.
1:38:45So, we have basic problems as a country.
1:38:48Um he's mentioned some uh bright, I
1:38:51mean, electricity, water, etc.
1:38:55So, why should Ghana invest in AI when
1:38:57there are more basic development needs?
1:39:01It's a question that has been asked
1:39:02legitimately. Should we be spending 270
1:39:05million
1:39:07on that
1:39:08when the fundamentals must be fixed
1:39:10first? Or should we focus on addressing
1:39:13these big challenges before we build it?
1:39:16>> Actually, you need it for that. That's
1:39:17the point I'm making. You need the power
1:39:19>> for that. So, if you don't if you don't
1:39:21fix the power and the water,
1:39:22>> you can't do the AI
1:39:23chip. You have to do it expensively.
1:39:25>> That that I guess comes to the question.
1:39:26So, and that was asked of me. I mean,
1:39:28people when we say, "Well, why are you
1:39:29going to talk about AI? Why are you even
1:39:30spending money in the way we haven't
1:39:32fixed the basic developmental challenges
1:39:34that we have like electricity and
1:39:36water?" The minister will attempt it and
1:39:38then I'll get my guests to also weigh in
1:39:40because the business community are
1:39:42represented here. What would they want
1:39:43to see
1:39:45um as far as that is concerned to have
1:39:47something that is local, but they can
1:39:50also tap into that works for them and
1:39:52all of us. And then my audience is also
1:39:54weighing stay with us.
1:40:04>> Today, I ride [music] one. Tomorrow, I
1:40:07hope to own 10.
1:40:09Today, I supply a few shops. Tomorrow,
1:40:12[music] I hope to supply the whole city.
1:40:14Today, I take you this far. Tomorrow, I
1:40:17hope he goes much further.
1:40:19>> Boxer is more than something you [music]
1:40:20ride. It helps you keep moving towards
1:40:23what matters because when you move
1:40:25forward, Ghana [music] moves forward.
1:40:27Buy Bajaj Boxer.
1:40:28Moving you forward exclusively from
1:40:31Somoco Ghana with rider training at the
1:40:32Somoco Riding Academy.
1:40:36>> When malaria shows up, it doesn't knock.
1:40:40It just takes over.
1:40:52>> This fever is too much.
1:40:53>> When malaria comes, respond with
1:40:56Malatou.
1:40:58>> [music]
1:41:00>> Fight malaria with Malatou tablets or
1:41:03syrup [music] from Ernest Chemist.
1:41:06Test for malaria first before taking
1:41:08Malatou. If symptoms persist for more
1:41:11than 48 hours, see a doctor. [music]
1:41:13This advert is FDA approved.
1:41:18>> Upgrade your print game with Konica
1:41:20Minolta, proudly powered by [music]
1:41:22Skysat Technologies. From all-in-one
1:41:24machines that print, scan, and copy to
1:41:27professional systems [music] and
1:41:28heavy-duty production machines, we
1:41:30deliver unmatched performance. Right
1:41:33now, Skysat has an unbeatable offer. One
1:41:36entire year of free services on all
1:41:39Konica Minolta machines. That's 12
1:41:41months of expert maintenance for new and
1:41:44existing customers. [music]
1:41:46Visit our new VIP print center at
1:41:48Pegasus Place, Airport Residential Area
1:41:51for a live demonstration. Call us on
1:41:53050-164-5180
1:41:55[music]
1:41:57or 0302-798-829.
1:42:00Again, it's 050-164-5180 [music]
1:42:04or 0302-798-829.
1:42:08Skysat Technologies, performance that
1:42:10delivers every time.
1:42:35>> [music]
1:42:48[music]
1:42:54[music]
1:42:59>> Love [music] is patient.
1:43:01Love is kind.
1:43:03And today, as your pastor, I love you.
1:43:05>> [music]
1:43:06>> Love is patient.
1:43:08Love is
1:43:12>> [music]
1:43:19[music]
1:43:22>> This wedding can't
1:43:24end with a kiss.
1:43:27Yeah, but fun of what?
1:43:30>> [music]
1:43:30>> They will start having one to have.
1:43:37>> [music]
1:43:49[music]
1:44:02[music]
1:44:05>> A taste of Ghana.
1:44:09This advert is FDA approved.
1:44:14>> [music]
1:44:20>> From the heart of Africa we [music]
1:44:22stand.
1:44:25Building dreams across the land.
1:44:27>> [music]
1:44:29>> Every step we take is a sign. Helping
1:44:32you rise, taking you higher.
1:44:34>> [music]
1:44:39>> Strong as [music] a group, close as a
1:44:41partner. Bank of Africa lifting you
1:44:43higher. African bank with global [music]
1:44:46reach. We always make your future glow.
1:44:48Strong as a group, close as a partner.
1:44:51Bank of Africa lifting you higher.
1:44:53African bank with global reach. We
1:44:56always [music] make your future glow.
1:45:01We go there for you.
1:45:05>> [music]
1:45:07>> Strong as a group, close as a partner.
1:45:09Bank of Africa lifting you higher.
1:45:12>> Bank of Africa, as strong as a group, as
1:45:15close as a partner.
1:45:18>> [music]
1:45:33[music]
1:45:39[music]
1:46:02>> Good conduct, great life is brought to
1:46:04you by
1:46:06>> Tastes great.
1:46:07>> More cubes inside. Keywogari soaking
1:46:10mix, a taste of Ghana.
1:46:15>> [music]
1:46:18>> Drive EV Ghana was built on a simple
1:46:20idea, that driving can be better. It can
1:46:23be cleaner, smarter, and more refined.
1:46:26Today, we offer a range of electric
1:46:28vehicles that meet you right where you
1:46:30are. From luxury models that deliver
1:46:33power and prestige to everyday vehicles
1:46:36designed for practical, intelligent
1:46:38mobility.
1:46:46>> [music]
1:46:53>> But we didn't stop at vehicles.
1:46:55We built X Charge EV network to ensure
1:46:57charging is accessible, reliable, and
1:47:00efficient.
1:47:02>> [music]
1:47:03>> And a dedicated service center designed
1:47:06specifically for electric vehicles,
1:47:08because expertise matters.
1:47:11This is how the future becomes reality.
1:47:14Drive EV Ghana, we're leading the
1:47:16charge.
1:47:18>> [music]
1:47:23[music and singing]
1:47:35[music]
1:47:50[music]
1:48:02>> To us friends set this
1:48:03>> [music]
1:48:04>> to us fashion lovers,
1:48:06to us who flex with beautiful, unique,
1:48:09and attractive [music] fragrance.
1:48:11To us who sing of our magnificent life.
1:48:15Daring to feel different today [music]
1:48:17means impacting a pleasant scent.
1:48:20For tomorrow's dawn, feel good, feel
1:48:23strong. [music] Together, let's break
1:48:25these rules and set new ones with Slam.
1:48:28Together, let's break these rules and
1:48:30set new ones with Slam.
1:48:58>> And thank you very much for staying with
1:49:01us here on the Digital Economy Forum.
1:49:03We're having a conversation about AI and
1:49:06the Ghana economy.
1:49:08Disruption or opportunity? If you missed
1:49:11uh anything, you've missed the premiere
1:49:14of a documentary that opened up with big
1:49:16questions, big solutions, and set a tone
1:49:19for for this brilliant
1:49:22discussion that we've had so far. We've
1:49:24had the sector minister Sam George with
1:49:27us in the studio, who's laid out the
1:49:29government's vision and policy. We've
1:49:31had the rest of the audience also
1:49:34weighing on that bright presentation.
1:49:37And we've also had the industry reps
1:49:40tell us from their perspectives
1:49:42the experiences and what they believe it
1:49:44took to take to succeed and build one
1:49:47that works for Ghana. And
1:49:50Adam has brought very unique perspective
1:49:52from the very micro level, real-life
1:49:55beneficiaries of a functioning AI
1:49:58system.
1:49:59And we have a lively studio audience I
1:50:01want to get to
1:50:03pretty shortly. And I know many of you
1:50:05still have questions, so share them. Go
1:50:08on our many social media platforms and
1:50:10let's hear from you. Before we went for
1:50:12the break, there were there were a few
1:50:14additional questions that Bright had
1:50:16raised. And those questions came up in
1:50:18the documentary, but he just sort of
1:50:20amplified it. Um questions about power.
1:50:24I mean, AI compute centers
1:50:26they just
1:50:28drink water. That's another one. Um and
1:50:31you can see there are even conversations
1:50:32about building data centers in the sea
1:50:35because obviously you have water there,
1:50:37but of course it's salty. People are
1:50:38trying to figure that out.
1:50:39The power conversation he also threw up
1:50:41a big we don't have a reliable power
1:50:43infrastructure and can't we run it?
1:50:45There are huge basic questions that we
1:50:47need to fix first, he believes, because
1:50:50the AI would be dependent on those. And
1:50:52one big question I was asked
1:50:54when we were doing this was why are we
1:50:56spending money on building AI compute
1:50:59center when your basic problems,
1:51:02developmental challenges have not been
1:51:03addressed just yet. And I want to bring
1:51:06Sam George in
1:51:07to attempt to answer those questions
1:51:09that were thrown before we went for the
1:51:10break. And then I'll get my audiences to
1:51:13also start weighing in with their
1:51:14contributions and questions.
1:51:16Um my uh panel will also do so so
1:51:19because this is going to be an
1:51:20interaction. So, I'm not going to be
1:51:22asking too many questions. So, just let
1:51:23that flow and let's interact. Uh Sam
1:51:25George, you wanted to chip in.
1:51:27>> Yes. So, um Bright asked uh made a
1:51:30number of comments. Let me very clear. I
1:51:33don't expect people to cheer the
1:51:34minister. No, far from that. I I welcome
1:51:38constructive criticisms and debates.
1:51:41But, what I don't What I will not
1:51:43subscribe to is as becoming a nation of
1:51:45talkers and not doers. It's very easy
1:51:47for us to always want rigorous debates
1:51:49that lead us know everybody has an
1:51:51opinion. The role of leadership is to
1:51:54take a decision, own that decision,
1:51:56execute it, and be held responsible for
1:51:59it. If we want If we want If we want
1:52:01everybody to have an opinion, then
1:52:03there's no need for leaders. Leaders are
1:52:05put in office to take decisions, own
1:52:08them, face the consequences of their
1:52:11decisions. And so, for me, I I make the
1:52:14decisions I make, I'm confident that I
1:52:16will hold my own any day on the
1:52:18decisions I make. And I'm open for
1:52:20criticism.
1:52:22The second question of power,
1:52:25then we should not have an
1:52:27industrialization drive as well.
1:52:30Factories all take from the power grid.
1:52:33So, that argument is moot.
1:52:35As a country, you know where you want to
1:52:37go, you plan for it. So, I didn't just
1:52:39wake up and say I'm building a data
1:52:41center. It went to cabinet. There's a
1:52:43minister for energy who sits in cabinet,
1:52:45who's made the right investments. Along
1:52:48the same lines of us talking about
1:52:50building a national data center, you've
1:52:51heard about the 1,200 MW
1:52:54gas plants that's coming up
1:52:56from government. You've seen the number
1:52:57of IPPs that are coming on stream.
1:52:59Today, we don't have a generation
1:53:01capacity problem.
1:53:03That's not That's not a problem in
1:53:04Ghana. And so, the Ministry of Energy
1:53:07has seen what the power demand for our
1:53:09project is and they're planning
1:53:11alongside it. We're not just plugging in
1:53:14immediately tomorrow and saying that
1:53:15we're ready disrupt the grid. Same with
1:53:18water. So, there are plans for it. Even
1:53:20the sighting of the location is
1:53:23strategic.
1:53:24Okay? Because we're looking at the power
1:53:26grid, how it's been designed, where's
1:53:28the what's closest water supply, all of
1:53:31those. As we speak right now, the
1:53:32location of this has an IFC a $200
1:53:35million IFC
1:53:36uh private sector
1:53:37>> center.
1:53:38>> Yeah. There's a There's a $200 million
1:53:40IFC-sponsored
1:53:41uh private sector water treatment
1:53:43solution that is being done there to
1:53:45produce 200 million gallons of water.
1:53:48So,
1:53:49it's an integrated plan. It's not just
1:53:51um an off-the-cuff thing. There are
1:53:53works towards it. And like I keep
1:53:54saying, it's
1:53:57Allow us to Allow us to experiment with
1:53:59our suggestions.
1:54:01Or with our with our plan. Suggestions
1:54:03are welcome. Criticisms are welcome.
1:54:05But, if we say that we want to
1:54:07take all of our boxes before we start,
1:54:10then we're going nowhere. The The ship
1:54:12is going to sail before we move. The
1:54:14question of who is actually going to use
1:54:17it?
1:54:19I think that we've got businessmen here.
1:54:21And that answer has been given here.
1:54:24Give them the solution that meets their
1:54:27needs and they'll fix it. You know, it's
1:54:30it's
1:54:32It's easy to use the the analogy of the
1:54:35of the phone centers.
1:54:38But, you see, when mobile phones also
1:54:40came in,
1:54:41the decentralization
1:54:43or dereg- deregulation of that space
1:54:46allow mobile phones came in was a single
1:54:48government policy. One single government
1:54:50policy that look, let's break PNT up,
1:54:53the liberalization of the of the
1:54:55ecosystem.
1:54:56It was not government that brought in
1:54:57the mobile phones.
1:54:59And nobody was going to be able to bring
1:55:00in mobile phones into Ghana.
1:55:02At the time we started having mobile
1:55:04phones here 1992 and 93, nobody would
1:55:08have thought about a business in
1:55:09bringing mobile phones if there was no
1:55:12policy framework that underpinned the
1:55:15use of cellular phones and giving and
1:55:17putting in place spectrum for mobile
1:55:19telephony. Because at that time the
1:55:21policy and regulatory framework only
1:55:23allowed fixed lines.
1:55:24So, it always comes back I keep saying
1:55:27that look, government ought to be an
1:55:28enabler.
1:55:30The private sector will grow. How many
1:55:32people does government employ in this
1:55:33country? Less than a million.
1:55:35What's our workforce? Almost 18 million.
1:55:38The biggest employer is not government.
1:55:41The biggest employer is the private
1:55:42sector. But what the private sector
1:55:44needs is the right incentives to expand
1:55:47and employ.
1:55:48Now, and that's what um
1:55:51uh E- Emmanuel, right?
1:55:53>> Yeah, Emmanuel.
1:55:53>> E- Emmanuel was talking about that what
1:55:55they need is for government to break
1:55:57that ice.
1:55:59Break the boundary. Allow them to go.
1:56:02You don't expect I wouldn't expect any
1:56:04of the private sector players here to
1:56:06say that they're building their own AI
1:56:07compute center. Cuz then it makes no
1:56:09sense. But that's what government ought
1:56:10to do. And if it comes to part today,
1:56:12we've got a littering and a sprinkling
1:56:14of data centers all over this country.
1:56:17In fact, you have a tier four data
1:56:18center Onix running out of Ghana.
1:56:21They have power.
1:56:23They have water.
1:56:24So, if Onix, which is a private
1:56:26enterprise, is able to run a tier four
1:56:28data center
1:56:29and the country hasn't shut down, it's
1:56:31not in darkness,
1:56:32that's not the reason why water is not
1:56:33getting to parts of the country, why
1:56:35should a national AI center be the
1:56:38reason why
1:56:40we would have a power crisis or why we
1:56:42should not be able to go in there. We
1:56:44need a critical Look, the amount of And
1:56:47And I've done
1:56:48a few tours last week. I was in KNUST, I
1:56:51was in the University of Health and
1:56:52Allied Sciences.
1:56:53If you see the research that's going on
1:56:55there and the need for compute power in
1:56:58our universities, which are supposed to
1:56:59be in the brain hubs of our country. And
1:57:02the fact that they're having to rely on
1:57:03tokens from outside because they don't
1:57:05even have computing infrastructure
1:57:06locally that they can use. The closest
1:57:08they use is I had to sign on to the MATE
1:57:10plan in the interim so that we could
1:57:12give our universities some access to
1:57:15reduced cost computing infrastructure
1:57:17from a supercomputer sitting in Italy.
1:57:20The question is why don't I create those
1:57:21supercomputer opportunities here for
1:57:23them, for the researchers in our
1:57:24university? If we say we want to fix all
1:57:27of our problems
1:57:29before we get into the world of
1:57:30artificial intelligence, that for me
1:57:33will be very regressive thinking because
1:57:35there is no country, even when you go to
1:57:37the United States, which is the world
1:57:39leader in artificial intelligence, about
1:57:4175% of computing infrastructure is in
1:57:42the United States. You've got huge
1:57:44swaths of America where they're dealing
1:57:46with with with uh uh homelessness. Huge
1:57:50swaths where they're dealing with
1:57:51unemployment. So, if you're talking
1:57:53about the social inequalities and
1:57:55inefficiencies of a system, there is no
1:57:58country in the world that has solved all
1:58:00of its basic requirements. There are
1:58:03parts of the US today that are still
1:58:05running on dial-up internet.
1:58:08There are parts of the US today that
1:58:10still do not have 4G connectivity, but
1:58:12that has not stopped them from spending
1:58:13trillions in building AI capacity. So,
1:58:15to to to suggest that until we fix all
1:58:19our schools under trees, until we have
1:58:21water flowing in all of our pipes, let's
1:58:24sit down and put our hands in between
1:58:25our legs and debate
1:58:28and not go into that world where we can
1:58:30begin to give the likes of our business
1:58:32community sitting here
1:58:34that push to begin to create the
1:58:36unicorns, to grow those businesses into
1:58:38unicorns. And you see you you you were
1:58:39shocked at the unicorn thing.
1:58:41>> Yeah, $1 billion.
1:58:42>> Yeah. But we want 10 of them in 9 years
1:58:44because we
1:58:45>> It's all all it took all it took was for
1:58:47a Nigerian, in fact, the two of them,
1:58:49Flutterwave OPay, right?
1:58:51O- O- OPay is a is a unicorn. Simply
1:58:54because of the kind of investment, and
1:58:56that investment does not necessarily
1:58:58have to come from within the country.
1:59:00>> Mhm.
1:59:01>> You can attract that from outside, but
1:59:03that investment is only attracted when
1:59:05they see that the framework exists in
1:59:07country to support that business.
1:59:09>> What I want to do now I come for and I
1:59:11know I know Bright is is uh is dying to
1:59:13come back in.
1:59:13>> 20 seconds.
1:59:14>> Uh uh
1:59:15forgive me. I I want to hear from I want
1:59:17to hear from um those on the ground who
1:59:20who minister is trying to sort of uh put
1:59:22in place policies for, and I've also had
1:59:24the debate in the studio. I I believe we
1:59:26have a farmer, do we? Or somebody who
1:59:28works on the farm.
1:59:30Um
1:59:30We're not yet. Okay, we'll get at that.
1:59:32There's some questions, and I think I
1:59:34have Michael. And Michael has some
1:59:36interesting thoughts on this. Bright,
1:59:37I'll come back, and I'll have all of you
1:59:39weigh in one more time. But let me hear
1:59:40from my audience. Yes, Michael, please.
1:59:42>> Okay.
1:59:42>> Michael Adjete.
1:59:43>> Yes, sir.
1:59:44>> He's an AI skeptic and tech and
1:59:46enthusiast. If you if you can stand up
1:59:47for me so that we can the the cameras
1:59:48can can capture you.
1:59:50>> So, I have two broad questions. The
1:59:52first one is now just about a month ago
1:59:56it was revealed that Nvidia was working
1:59:58with Amazon to aid the shortfall of
2:00:01hardware on the global market. Now,
2:00:03given how much uh IT space, especially
2:00:07towards AI, has been
2:00:09pretty much blocked up to 2030 and
2:00:12beyond, how much effect or what do you
2:00:15expect the 250 million dollars to do
2:00:18because
2:00:19this is not in isolation. We're not
2:00:21doing this solely for ourselves. We are
2:00:23competing pretty much with the rest of
2:00:25the world. And we are already on the
2:00:27back foot, and they're accelerating away
2:00:29from us. So, what is that going to do
2:00:30because when you want the private sector
2:00:33to make use of that compute,
2:00:35they're not going to be you saying, "Oh,
2:00:37because it's in Ghana." The only
2:00:38advantage for it being in Ghana is now
2:00:40they have access to our own local data
2:00:43sources. But beyond that, when it comes
2:00:44to compute power,
2:00:46there's there'll be little else to say
2:00:48that well I will use the data here but
2:00:50then I'll still have to put the user as
2:00:53my tokens for the calculations
2:00:55elsewhere. So what are you hoping to for
2:00:59that 250 million dollars to do in that
2:01:02respect? Now the next question I also
2:01:03have is what broadly is our plan for
2:01:07defense?
2:01:08Because again, we are not working in
2:01:10isolation. We are literally in a global
2:01:12forest. And
2:01:14AIs have gone rogue before and they will
2:01:17go again. Just today OpenAI admitted
2:01:20that some of their AIs were disabling
2:01:22their own safety protocols and lying
2:01:24that
2:01:25to their supervisors that that had
2:01:27happened.
2:01:28And that is not even the main point of
2:01:31the question because
2:01:33AI being a tool
2:01:35is also can be akin to a child being
2:01:39given a weapon.
2:01:41Because now
2:01:42bad actors can and have made use of that
2:01:45ability to do all sorts of terrible
2:01:48Today, this morning there have been
2:01:50three massive cyber infrastructure
2:01:52attacks, one of which was
2:01:54done with OpenAI and has now rendered
2:01:58nearly 80% of websites and almost every
2:02:01device on the internet vulnerable to
2:02:03attack.
2:02:03>> To the point where they need to be
2:02:04patched.
2:02:05>> That's the one. To the point where the
2:02:07NSA
2:02:08today said that all major US companies
2:02:11that have infrastructure online should
2:02:13take them off until they are patched.
2:02:15That is the world that this policy is
2:02:18going to be competing. That's where we
2:02:19are going to be in. Already we are under
2:02:21attack attack now and not even
2:02:23necessarily because it's aimed at Ghana
2:02:25but we are in that space. And now what
2:02:28are we doing because fine, we are
2:02:29entering the space as we need to but
2:02:31what is our defense?
2:02:32>> Yeah. Two quick very important
2:02:34questions. And before I give the
2:02:35microphone to I I believe I also have
2:02:37Luke. Luke is
2:02:39he's he's lived the uh opportunity. But
2:02:42before
2:02:43Michael, just a quick one. Um it I know
2:02:46in you articulated views in the when we
2:02:48spoke to you before, very close to what
2:02:51Bright had articulated. You've heard now
2:02:53the minister address some of them.
2:02:55>> Yes.
2:02:55>> You convinced?
2:02:56>> Not really, to be honest, but I'll That
2:02:59was my my aspect. I may come back again
2:03:01and then emphasize on that one more
2:03:03time, but I'm not really convinced
2:03:05because there's a foundation to be laid
2:03:07and it is at best it is wet, if not
2:03:11partially done, before we are now going
2:03:13to build a full world competing
2:03:15structure upon it. We are not building
2:03:17it for our own sake.
2:03:18It is in competition against the rest of
2:03:21the world who have a huge head start. We
2:03:24need something stronger to stand on.
2:03:26>> But we must start somewhere.
2:03:27>> We must start somewhere. I'm not saying
2:03:28we shouldn't start and we shouldn't not
2:03:30do anything, [clears throat] but we need
2:03:31something strong to stand on.
2:03:33>> Okay.
2:03:33>> Strong being the key word because it's
2:03:36We can't You can't fix a foundation
2:03:38that's broken. Once it's
2:03:39You can't fix a bad foundation. Or your
2:03:42building will collapse before once you
2:03:43put too much weight on it. So, we need
2:03:45something strong to stand. I'm not
2:03:46saying fix all the problems, but fix the
2:03:49core problems that allow us to have that
2:03:50strong foundation to then go forward.
2:03:52That's basically what
2:03:53>> Okay, Michael.
2:03:54>> I'm hearing Thank you very much. Uh and
2:03:55and Luke now, Luke, you you live the
2:03:57disruption. You've also live the
2:03:58opportunity. Uh so, you embody both.
2:04:01Speak to me having heard the
2:04:03conversation that is panel so. I don't
2:04:04know where where where you stand on the
2:04:07on the conversation and traction we had
2:04:09uh on this side of the of the room.
2:04:11>> Okay, and thank you very much, Evans,
2:04:13and um thank you all the panelists.
2:04:15>> And Luke, by the way, is the manager of
2:04:17translation office at the University of
2:04:18Ghana.
2:04:19>> That's right.
2:04:20>> Yeah.
2:04:21>> So, um
2:04:22you know, AI, everyone is excited,
2:04:24right? Everyone is excited.
2:04:26Um as a translator and someone training
2:04:29translators at the University of Ghana,
2:04:32this is big. This is very important.
2:04:34Because people ask our students, "Why
2:04:37are you still studying uh translation
2:04:39when AI can do the work that translators
2:04:41are doing?
2:04:43And um Minister, you need to know that
2:04:46some months ago, I decided to visit the
2:04:49website of the NCA.
2:04:52And I saw
2:04:53that the text in French apparently was
2:04:57from
2:04:57maybe an AI platform.
2:05:00I thought it was from an AI platform
2:05:02because I saw some errors in it.
2:05:05So, I sent an email to the
2:05:07you know, authority, and I said, "Look,
2:05:09I am a translator. This is what I've
2:05:11seen on your platform. I think that you
2:05:13are using something probably to generate
2:05:16this. In case you are, engage
2:05:18professionals.
2:05:20>> Very
2:05:20>> What are you doing about this?
2:05:22Are you thinking of language
2:05:24professionals? Because whoever is
2:05:26talking about AI is talking about
2:05:28languages. That's why we talk about
2:05:30large language models or small language
2:05:32models.
2:05:33But it's like the language professionals
2:05:35are not really involved in this.
2:05:37Right. Are you talking to the schools
2:05:39that are running language programs like
2:05:42translators?
2:05:43I've been trained every day. Are you
2:05:45thinking of us?
2:05:46As students, English students. Are you
2:05:50using using them for the audit
2:05:51exercises? You know, I mentioned the
2:05:54annotation projects in the doc-
2:05:56documentary.
2:05:58Are you thinking of
2:05:59language students?
2:06:01If you are thinking of AI, you must be
2:06:03thinking of languages.
2:06:06And students are very important. The
2:06:08people training these students are also
2:06:10very important. Professionals out there
2:06:14running companies, you know, that render
2:06:16these services, are also very important.
2:06:19Are you collaborating with them?
2:06:21Or government agencies are just happy
2:06:22that ChatGPT can translate.
2:06:25And you expressed a concern, Minister,
2:06:27about um the use of these platforms for
2:06:31processing of sensitive government
2:06:33information.
2:06:34But I can tell you
2:06:36that there are ministries
2:06:38that are using these platforms to
2:06:40translate sensitive government
2:06:42information without even realizing the
2:06:44danger involved.
2:06:46Right? But are you engaging
2:06:48professionals?
2:06:49>> Mhm.
2:06:49>> You mentioned the use of focal persons,
2:06:52AI focal persons. Right. What of
2:06:56language professionals who'd be work
2:06:58working, you know, jointly with these
2:07:00professionals to ensure that even as
2:07:03they use AI platforms for the
2:07:05translation of materials, they would be
2:07:07doing so in a responsible manner. Right?
2:07:10That's the question I have to ask.
2:07:11>> Yeah, and and just a quick one for you.
2:07:13So, because you um you currently consult
2:07:17for some of the models abroad.
2:07:19Government is trying to localize and
2:07:20build one locally. Definitely they would
2:07:22need um you know
2:07:24services like yours.
2:07:26If you had the debate that happened
2:07:27earlier, you would stand to benefit um
2:07:30if we are building one locally because
2:07:32then of course your input is directly
2:07:33here.
2:07:34Uh we have challenges as a country. Just
2:07:36want to hear your perspective on what
2:07:38should we do
2:07:39um
2:07:40in the interim whilst we fix some of the
2:07:43challenges that we have?
2:07:45What would you What would you recommend?
2:07:47And I just want to hear you know,
2:07:48somebody who's living this on a daily
2:07:49basis who stands to benefit's point of
2:07:51view on the debate.
2:07:52>> Yeah, so um
2:07:54>> Very briefly.
2:07:55>> I I I think I've already mentioned the
2:07:57need to get closer to these institutions
2:08:00that are
2:08:01training have programs and are training
2:08:03students, right?
2:08:04>> No, not even that. I'm I'm just talking
2:08:06about the the basics that the argument
2:08:08goes might be fixed first before we we
2:08:11putting the the infrastructure for AI um
2:08:15etc. Would What would you What What's
2:08:18your argument? What's your What's your
2:08:19What do you What do you think about
2:08:20that?
2:08:20>> I'm really trying to get this question,
2:08:22but
2:08:23>> You you know, Bright makes a point and
2:08:25you your colleague also makes a point
2:08:26that we have so many big challenges.
2:08:28There's electricity, there's water, etc.
2:08:31that we need to address. Big basic
2:08:33questions. Address that first. Try and
2:08:35fix that fundamental before you build on
2:08:38top of it the complex AI models and the
2:08:40computer center.
2:08:41>> Okay, I think that the minister has
2:08:43mentioned that it would not be possible
2:08:44to fix everything before we start.
2:08:47Right?
2:08:49Awareness creation is very important.
2:08:51AI is good, but responsible AI is more
2:08:54important.
2:08:56Right? Responsible use of AI is what we
2:08:58are talking about here.
2:08:59That even as we try to embrace AI,
2:09:02let's teach our people how to use it and
2:09:04use it well.
2:09:06I think it's part of the fundamentals
2:09:07that as we try to embrace this good
2:09:10thing that we think will bring about
2:09:12development. I mean, it's it's
2:09:14productivity we are all talking about.
2:09:16People should be educated. They need to
2:09:18know that, okay, if you have a birth
2:09:20certificate for [clears throat] you
2:09:21yourself that you want to translate,
2:09:24you can use it that you can use a
2:09:25platform, but you must be careful. It is
2:09:27your information you are putting online.
2:09:29>> Okay, great. I I want to come I'll come
2:09:31back there are more questions. I'll come
2:09:33back to the to the panel.
2:09:35Um and I'll come to Bright shortly, but
2:09:37Minister, there are a few questions for
2:09:38you about defense.
2:09:39>> I'll just jump on them.
2:09:40>> Yeah.
2:09:40>> Yeah.
2:09:40>> The defense
2:09:41>> I think that I think that's the the
2:09:44first point made by
2:09:46uh the interpreter from University of
2:09:47Ghana actually addresses the second
2:09:50question that was asked by the first
2:09:52questioner because he's talking about
2:09:54the fact that there is a lacuna today in
2:09:58the quality of service when it comes for
2:10:00example interpretation. And for me,
2:10:02that's a dividend that exists. You can
2:10:05offer a service to government agencies.
2:10:06Thanks for drawing my attention to that
2:10:08of the NCA. I'll raise it with the
2:10:09director general. But it means that
2:10:11there is a service that is needed not
2:10:14just at the NCA, but at a number of
2:10:16government agencies. Those are jobs that
2:10:17are going to be created. Those are jobs
2:10:19that you're training people for. And at
2:10:21the University of Ghana, got a project
2:10:22with Google working with Professor Isaac
2:10:24Yeboah.
2:10:25I don't I don't know if you know
2:10:26Professor Isaac Yeboah, but building
2:10:285,000 hours of content because of what
2:10:32you raised, where you don't want to do
2:10:34just text, you want to do voice.
2:10:37Cuz many of the people who speak the
2:10:38local language cannot read and write it.
2:10:39They can only access it using voice.
2:10:41We're building 5,000 hours in seven
2:10:43Ghanaian languages. So, that's 35,000
2:10:46hours. That's data set that's going to
2:10:47be publicly open. That's data sets that
2:10:49will be used to train the models when we
2:10:51build the AI system. And so, that
2:10:53addresses the question of whether we
2:10:55really need to actually be investing in
2:10:58these technologies now because that
2:11:00problem exist. And I'm happy that the
2:11:03the first gentleman did not say that
2:11:04when you asked him he said, "It's not
2:11:06about not having the systems." He wants
2:11:08stronger systems.
2:11:11You had a In the documentary, you had
2:11:13other African countries are spending 300
2:11:14billion.
2:11:15>> Yeah.
2:11:17>> It's the question of how much resources
2:11:19available to us to work with. We have
2:11:21250 million. And we're going to make the
2:11:22best use of that. So, yes, I'm happy
2:11:25your question is not saying that we
2:11:27don't need the systems. You think we
2:11:28need bigger systems. Do we have to wait
2:11:30and sit till we have 300 billion as a
2:11:32country to invest in artificial
2:11:33intelligence?
2:11:35Maybe a skeptic like you will say to me
2:11:36that use that money to build a a a
2:11:38desalination plant
2:11:40and not not invest in AI.
2:11:42So,
2:11:43if I'm using 250 million today to build
2:11:46some existing capacity because you're
2:11:48asking the question of defense.
2:11:50Yes, defense is a critical part, but
2:11:52when you take our national AI strategy
2:11:53for example, you realize that we're very
2:11:56quiet. And I'm sure you've read it.
2:11:57We're very quiet on the military use of
2:11:59artificial intelligence.
2:12:01There is work that's happening in that
2:12:04space in the military sector. Um but
2:12:06we're careful to disaggregate our
2:12:09civilian use of artificial intelligence
2:12:11and what military use of artificial
2:12:13intelligence is. But you see, building
2:12:15our own internal capabilities as a
2:12:17country is highlighted by today's
2:12:21uh incident that you talk about where
2:12:22PHP-based platforms have all been been
2:12:26affected, basically. And that's what the
2:12:27NSA and the FBI cyber unit have have
2:12:30announced to all US government agencies.
2:12:32We need to have some level of
2:12:34resilience. But if our focus is on what
2:12:36you suggest, dealing with the basics,
2:12:39then all of our existing platforms that
2:12:42are there, whether you like it or not,
2:12:44we're we're deploying security systems
2:12:46that are online, we're deploying
2:12:48agricultural systems that are online,
2:12:50we're deploying health care systems that
2:12:51are online. If we do not build our own
2:12:54local capabilities, because we're not an
2:12:56island in isolation, global events will
2:13:00overtake us and overrun us. And I think
2:13:02it would be irresponsible on the part of
2:13:03leadership of this country to sleep on
2:13:06the job and not put in place at least
2:13:08the basic defense mechanisms and
2:13:11capabilities to ensure the
2:13:13sustainability of our systems. So, yes,
2:13:15I hear you as a skeptic. I hear you
2:13:16about the myriad of challenges we need,
2:13:18but the challenge of being able to be
2:13:20resilient as a country to withstand
2:13:23cyber attacks that are amplified by AI
2:13:25agentic tools today is also a basic
2:13:29necessity in the global age.
2:13:30>> Yeah.
2:13:31>> Right.
2:13:32>> Now, I think my arguments have been
2:13:34caricatured. And there's a danger that
2:13:36somebody listening and not paying
2:13:37attention will think the minister the
2:13:39right honorable minister is responding
2:13:40to me and it's not right
2:13:40>> It's not right honorable, that's only
2:13:42the speaker.
2:13:42>> Yes, but you know, in Britain, senior
2:13:44ministers are right
2:13:45>> I've heard you. This is Ghana, we have
2:13:47honorable.
2:13:48>> So,
2:13:49>> Or better still, left honorable.
2:13:50>> Okay, man sir. You like man sir?
2:13:51>> Man sir is better.
2:13:52>> So, what I'm trying to explain is not
2:13:54that, you know, we have problems, solve
2:13:56them all first
2:13:57or onto all the problems solve them
2:13:59building. No, I'm saying that build the
2:14:00AI with a lot of thoughts going to
2:14:03prioritization, sequencing, which
2:14:06elements must come first, which must be
2:14:08deprioritized now because of the issues,
2:14:09etc., etc. For instance, there are two
2:14:11big categories. You have training, you
2:14:12have inference.
2:14:14Training is actually building the model,
2:14:15right? Inference is basically using it.
2:14:18There are millions of things we could be
2:14:20using AI to do more effectively
2:14:22that we are not doing.
2:14:24So, to talk about building one is a
2:14:26clear trade-off between being very good
2:14:28at inference as a country, signing
2:14:30special agreements with the frontier
2:14:32labs, whether in China, the US, France,
2:14:34or whatever, so that you can still
2:14:36access them, but you do so on your own
2:14:38terms because like India has done with
2:14:40Mistral and others, you create
2:14:41conditions which your own local users
2:14:43are able to build small models that talk
2:14:46to like what they are
2:14:47There's a range of options. Now, if we
2:14:50were having a conversation 10 years ago
2:14:52on the sugar factory,
2:14:54you would see the politician, if there
2:14:56was this discussion about the sugar
2:14:57factory, you would see the same degree
2:14:59of confidence in the politician, right?
2:15:01His vision. That's what we call
2:15:02catonomics. Politics assembles. It
2:15:05mobilizes. It has these big banners,
2:15:08industrialization, national sovereignty,
2:15:10and it mobilizes us. Policy breaks
2:15:13things apart, and then you start to see
2:15:14the confusion. Let me give you a very
2:15:15good example right here.
2:15:17So, the minister said, "Oh, we already
2:15:18have tier four data centers."
2:15:20If you don't have the detail, you don't
2:15:21really analyze it, and this is not what
2:15:23you do, you say, "Of course, if you
2:15:24already have tier four data centers."
2:15:25Like, what is this guy talking about?
2:15:27The truth is that current traditional
2:15:29data centers use CPUs.
2:15:31AI data centers use GPUs. There's an
2:15:33engineer here who will try and explain
2:15:35what the difference is. The scaling laws
2:15:37operate dramatically differently. I'll
2:15:39give you an example. In terms of
2:15:40connectivity, for instance, with, you
2:15:42know, a
2:15:43a a
2:15:44a 10
2:15:45gigabyte per second transmission system,
2:15:47you are fine with a normal data center.
2:15:49With an AI system that is training all
2:15:51the time,
2:15:52so you have to have ensure 99.5% uptime,
2:15:55etc., etc., and it's transferring data
2:15:56and the rest of it, your connectivity
2:15:58needs go through the roof.
2:16:00You have You have one of Ghana's leading
2:16:02technology companies sitting here
2:16:04telling you that when they went in, they
2:16:05recognized the scale of it and decided
2:16:07that they're going to sequence. Like
2:16:08first you see an established model and
2:16:10then think about how they now build,
2:16:11localize, etc. That tells you that that
2:16:13kind of wisdom must also be applied when
2:16:16you're thinking about the national AI
2:16:17compute. So that is a critical point.
2:16:19We're not saying that don't attempt it
2:16:20at all. We're saying we should analyze
2:16:22the processes, the failures, etc. etc.
2:16:24You talk about water.
2:16:26And you say that we have existing data
2:16:28centers and they've not had trouble
2:16:29because they are using water. No, they
2:16:31don't because they are CPUs, they are
2:16:32air cooling.
2:16:33That is why my computer and the computer
2:16:35in uptown, they're able to use air
2:16:37conditioners and they're fine. They
2:16:38don't need liquid cooling and stuff like
2:16:39that.
2:16:40If you build an AI data center and that
2:16:42compute, that hyperscaler comes on
2:16:44stream or come online, you need liquid
2:16:46cooling. That's a complete different
2:16:47dynamic from air cooling. That detail,
2:16:50that level of detail
2:16:51>> Sorry, sorry to interject you, but
2:16:5211 years ago, when President Muhammad
2:16:55built our national data center, it's
2:16:56liquid cooling. It's not air cooling. So
2:16:58we've been doing liquid cooling. When
2:17:00you go to the Onyx Tier 4, it's liquid
2:17:02cooling.
2:17:03>> it's
2:17:05It's nowhere near.
2:17:06If you're even a GPU
2:17:07>> Irrespective of whether it's a GPU or a
2:17:09CPU, before you build that system, you
2:17:12have what the power capacity needs are
2:17:15and what the water capacity needs are.
2:17:17And what you should realize is today,
2:17:19the water capacity need of a Tier 3 data
2:17:22center that's built today is even less
2:17:24than a Tier 3 data center that we built
2:17:2611 years ago because technology is
2:17:28evolving.
2:17:30>> point
2:17:30>> The technology is evolving, so it's
2:17:31different.
2:17:32>> No, but and that is why
2:17:33>> The The scale doesn't change.
2:17:34>> Let me end on that point. That's why
2:17:36Kenya attempted this G42
2:17:39um project, etc. etc. They're going to
2:17:40build this 1,000 MW data center and the
2:17:42rest and they backed out. The reason is
2:17:44they didn't do that critical analysis
2:17:46that you do in a place like this where
2:17:48you say it's not just because you want
2:17:49the vision, but you plan every single
2:17:52sequence of it to make sure that you're
2:17:53making the right decisions. So the point
2:17:55was not that oh until you solve all your
2:17:57problems, don't attempt to build a data
2:17:59center and the rest of it. The point is
2:18:00that the reason why America, China, all
2:18:02of these countries are seeing explosive
2:18:04demand in energy
2:18:06even though they used to have data
2:18:07centers before.
2:18:08And on the onset of AI hyperscalers,
2:18:10those energy demands have gone through
2:18:11the roof. It's because there's a scaling
2:18:13law.
2:18:13>> Okay.
2:18:14>> These are issues that are
2:18:15detail-oriented and I'm not saying
2:18:17therefore that the means you shouldn't
2:18:19do it or you should do it. I'm just
2:18:19saying that the analysis is different.
2:18:21>> I'm not saying
2:18:22just 30 seconds.
2:18:24>> Yeah.
2:18:25I hear you on the scale and like I said,
2:18:26this is not an off-the-cuff decision.
2:18:29The assumption would be to be that
2:18:30government has not thought through.
2:18:31>> But you talk to the sugar.
2:18:34>> You see, you you give the similar
2:18:36analogy in your earlier submission of
2:18:38why Chinwoke
2:18:39our good friend Herman Chinwoke
2:18:42did not succeed and why some George
2:18:43succeeded.
2:18:44>> software, I think it's that one of the
2:18:46payroll so
2:18:46>> It's because you underestimate the power
2:18:48of political will.
2:18:50Herman Chinwoke was
2:18:51>> didn't have political will?
2:18:52>> Herman Chinwoke was a private
2:18:54businessman trying to break through
2:18:56politics and the red tape of political
2:18:58bureaucracy with a solution.
2:19:01Whoever he was selling that solution to,
2:19:03if they did not appreciate what he was
2:19:04doing, will not get it. That's different
2:19:07from when you have a cabinet-backed
2:19:09decision
2:19:10that says this is the direction
2:19:11government is going today.
2:19:12>> had a new data center, right? That
2:19:19>> That again is incorrect.
2:19:20>> What is incorrect?
2:19:21>> That's incorrect.
2:19:21>> Government agencies were not using it.
2:19:23>> No, Bright, that's incorrect because
2:19:25that government data center with all of
2:19:27the challenges we came to meet in 2025
2:19:30was hosting our entire GIFMIS payment
2:19:32roll. GIFMIS is the is the platform on
2:19:34which all government payments are
2:19:36processed. So to say that government
2:19:37agencies were not using it is not
2:19:38factual.
2:19:43the National Identification Authority,
2:19:45which is our single source of truth for
2:19:48verification
2:19:49>> That was
2:19:49>> is running on the national data center.
2:19:51But you just said it was not being used
2:19:53by government agencies.
2:19:54>> I've shown you. I've shown you. I've
2:19:54shown you.
2:19:55>> That's incorrect.
2:19:56>> No, we have 98% utility of the
2:20:04>> Let me let me let me let me go to Ghana
2:20:05from the from the industry point of
2:20:06view. Um, uh,
2:20:09how did you go about the challenge? Um,
2:20:11you also need electricity, right? Um,
2:20:14you you you you you do. You try to run
2:20:17an AI lab.
2:20:18Are there some little lessons there that
2:20:20we can draw as a country as we begin our
2:20:23journey that you can share?
2:20:25>> Yes, uh, even so
2:20:27uh, in the heat of the extreme
2:20:29>> [laughter]
2:20:34>> Let's Let's bring it Let's bring it back
2:20:36down to our level so we can all connect
2:20:38to I mean, are there any lessons? I
2:20:40mean, one of the key lessons from the
2:20:40documentary obviously is Alex talking
2:20:42about, you know, even finding the people
2:20:44to run it um, with the skill and the
2:20:46know-how. So, training, we haven't even
2:20:47talked about that, but it was in the
2:20:48documentary, so it's fine. So, that was
2:20:49one lesson that you need to do. Then, he
2:20:51says data.
2:20:52>> Mhm.
2:20:52>> You you couldn't find the data. So, if
2:20:54you want to do this, fix the data
2:20:56fragmentation problem. Another
2:20:58recommendation. Are there any other
2:20:59things because other things have come up
2:21:01that maybe you you try we haven't
2:21:03explored yet, but that is a way to learn
2:21:05as a country before we we we set off on
2:21:07this journey. Any anything else you can
2:21:09>> Right. So, um, um, right. And the and
2:21:12the honorable minister there, I think
2:21:14they're both right, but maybe in a
2:21:15different time zone.
2:21:17So,
2:21:17>> [laughter]
2:21:18>> so,
2:21:19um, for Hubtel
2:21:21the challenges were glaring. At first,
2:21:24we didn't even know we were on a
2:21:25parallel tangents to these big frontier
2:21:29companies.
2:21:30Cuz the ideas and the
2:21:33the critical thinking that went into
2:21:35some of the projects we were working on,
2:21:37I mean, helping ECG, our own
2:21:40uh, call center application, and then
2:21:42fraud.
2:21:44Right?
2:21:45Very big issues.
2:21:46So, in our attempt to do that, we
2:21:49thought, "Okay, we have brilliant people
2:21:51in Ghana.
2:21:52Some are into machine learning and other
2:21:54things. At that time, the the AI thing
2:21:57hadn't
2:21:59been experienced as an app just yet.
2:22:03So, we got very very
2:22:04>> before they they before we had the first
2:22:08chatbots, you know, ChatGPT and co.
2:22:10became a a thing.
2:22:11>> Before it became mainstream.
2:22:12>> Mainstream, okay.
2:22:13>> Right? We had machine learning engineers
2:22:16and that's different from AI, right? So,
2:22:18they were to come in and use the same
2:22:21skills and knowledge that these
2:22:25very good guys in the in the other
2:22:28countries are doing and then
2:22:30get us a model, something that would say
2:22:32that, "No, this meter
2:22:34appears that you don't pay bills. Why?
2:22:37Let's look at
2:22:38your your Let's backtrack your record."
2:22:42>> Yes, it would have been done. We do risk
2:22:44profiling, we'll check and then we'll
2:22:46see that, "Okay, it's because of A, B,
2:22:48C." Then the dots will be connected.
2:22:50>> Mhm.
2:22:50>> Right? Then we'll scale it across. So,
2:22:53it is when the scale question came in,
2:22:56then we realized it's it's quite beyond
2:22:59what we intend to do. So, now the
2:23:02thinking is to scale back, own what you
2:23:05can control, your own information, your
2:23:07own data, your own processes. And then
2:23:10you can rent responsibly rent because
2:23:14these frontier companies or these AI
2:23:15companies, they will sell you their
2:23:18magic wands.
2:23:19>> Mhm.
2:23:19>> But behind the scenes, they are they are
2:23:22intelligently also taking I'm happy the
2:23:25honorable minister and others are
2:23:26mentioning. There's a different tier
2:23:28that when you are on subscribed to them,
2:23:31they promise they promise that they will
2:23:34not use your data for training their
2:23:35models.
2:23:36>> Mhm.
2:23:36>> So, Hubtel responsibly, according to our
2:23:39internal AI policies too,
2:23:42we had word with them, we had meetings
2:23:44with them. We still have ongoing
2:23:46meetings with them. I mean, they
2:23:47couldn't believe there's a company in
2:23:49Ghana that is even can't pay that much
2:23:52money, right? But we did. Right? And
2:23:55then they put us on a tier that we can
2:23:58run their engines in our vehicles that
2:24:02we own. The data and the context layer
2:24:04that we own. Because, for example, see,
2:24:07there was um
2:24:09um um a model and open with they call it
2:24:11open with let's say open source, all
2:24:13right? That we got in we hosted
2:24:16ourselves. We asked you the question
2:24:18about how they were telling us it's
2:24:20headquartered in Ki- Kigali.
2:24:21>> Mhm.
2:24:22>> Right? It doesn't make any sense. So, we
2:24:24need to go we need to take frantic steps
2:24:28to teach it and train it our ways.
2:24:31And then it starts producing the results
2:24:33that we want.
2:24:34>> Yeah.
2:24:35>> So,
2:24:36that those are some of the lessons hard
2:24:38lessons that we had to learn that it's
2:24:39not the hype of hey, these guys have
2:24:42brought this new feature and you can
2:24:44talk to it and you can just upload some
2:24:46document and it will give you some
2:24:49flashy output. That's That's That's not
2:24:51it. There's something more to it. We
2:24:53need to do that work of organizing
2:24:55ourselves properly internally,
2:24:57>> Yeah.
2:24:58>> making sure that we expose the right
2:25:00data responsibly, we own what we can
2:25:03control.
2:25:04>> before you begin to scale back?
2:25:05>> Yes, we own what
2:25:06>> What you're trying to scale before you
2:25:07realize you have to pull back?
2:25:08>> Because the temptation with the with the
2:25:09sweetness of AI is that see, oh, my
2:25:12phone now I have a medical record of my
2:25:14mom 74 years.
2:25:17If you don't take care, the temptation
2:25:18is just running that PDF through some
2:25:21chat GPT and it will give you some
2:25:24results, right? But they will take your
2:25:26information cuz you are you you just
2:25:28signed up as a individual person. You
2:25:31are not on any band that would restrict
2:25:33them from using your data.
2:25:34>> Mhm. Interesting.
2:25:35>> Right? So, there is some sweet spot
2:25:37there. That's what is creating all this
2:25:39excitement about AI AI AI because it can
2:25:42things that previously you couldn't do.
2:25:44>> Yeah.
2:25:44>> Right? But we expect it to be very
2:25:47responsive.
2:25:49>> Um Osumafu has has a question or
2:25:50contribution.
2:25:52Um please give the microphone to Osu. Um
2:25:55then Osu, please stand up for me and
2:25:56then let's and then I I also see
2:25:58um my final round on this Anthony
2:26:01Akonnor Edu.
2:26:02Also has has a question I'll read out. I
2:26:04think that came on our social media
2:26:05handles. Yes, Osu.
2:26:06>> You know, um Evans, you had to put me on
2:26:08the spot like that.
2:26:09>> Ah, yes. I know you had a question, so.
2:26:11>> Um yes, so the the the debate has been
2:26:13very um
2:26:16very hot from right
2:26:18uh taking us to theoretical stances all
2:26:20the way to being empirical data driven.
2:26:22Yeah, it's good. It's good the questions
2:26:23you're asking. I think the honorable
2:26:24minister has given a lot of answers. In
2:26:27fact, if you look at the AI policy,
2:26:28which I've read partly, not fully,
2:26:31um some of the answers are there. So,
2:26:33you ask a question, I'm not trying to
2:26:35get the trade, but you ask the question,
2:26:37can a country that fails to learn teach?
2:26:40And I see in the policy that yeah, from
2:26:43all the way from governance, we we've
2:26:45seen it's there that what we
2:26:47experienced in the past and how we're
2:26:49going to make it better all the way to
2:26:50localization of what he was talking
2:26:51about. Um
2:26:54um some hallucinations here and there,
2:26:56right? So, localization, how we instead
2:26:59of it speaking direct I mean strictly
2:27:01English, how can we localize all of it?
2:27:03Yes, but so, at the high level, you are
2:27:06being theoretical. Minister is being
2:27:08very high-level data things. I want to
2:27:10bring it down to our level.
2:27:11>> Mhm.
2:27:12>> Um what does it mean for the ordinary
2:27:14Ghanaian who fears that
2:27:16will I still keep my job or am I losing
2:27:20it? So, is it an opportunity or a
2:27:22distraction? I want us to bring it home
2:27:23because
2:27:25you are taking us to
2:27:26>> So, so you see why I called you.
2:27:27>> [laughter]
2:27:27>> Because that's where we need to end it.
2:27:29In the in the in the in the eventual
2:27:30analysis,
2:27:31um
2:27:32really that the person listening to us
2:27:35right now, Luke for example, whose job
2:27:37he told us is under threat. Wants to
2:27:39know when everything is said and done,
2:27:42by the time we finish implementing this
2:27:43policy, what would it mean for me?
2:27:45>> Correct.
2:27:46>> Right?
2:27:46>> Yeah.
2:27:46>> Um
2:27:47>> 1 million coders answers that program.
2:27:49The 1 million coders
2:27:50>> I'll come to you, Adam. Yeah.
2:27:51>> So, 1 million coders project, and we've
2:27:53had real-time testimony of people's um
2:27:57experience with them. That's why I think
2:27:58that people need to take and and and let
2:28:00me use this platform to ask our people
2:28:03to enroll on that program. The the
2:28:05www.1millioncoders.gov.gh.
2:28:08We're doing some work with the money
2:28:09Ghana on how the utility of the program
2:28:13translates from training now into jobs.
2:28:15And even that comes back to the question
2:28:17from the the translator from UG. For
2:28:19example, in Ghana today,
2:28:21there is a there's a a company that has
2:28:24employed 1,500 people doing data
2:28:26annotation
2:28:28here in Ghana. 1,500 people. Ghanaians,
2:28:31not foreigners.
2:28:33So, there is Sorry?
2:28:35Teleperformance. Oh, Teleperformance.
2:28:36Okay. Concentric is doing some other
2:28:37kinds of BPO as well.
2:28:39But Teleperformance is doing that as
2:28:41well. So, and and we're having
2:28:42conversations with Teleperformance and
2:28:44Concentric between both of them over the
2:28:45next 3 years to employ 10,000 people.
2:28:47And how do we position
2:28:50the 1 million coders to be a pipeline
2:28:52for them? How do we do the And that's
2:28:55what we're doing with Imani uh
2:28:57um
2:28:57Selorm and Franklin. We're working with
2:29:00them to now
2:29:01do a mapping of skills gap. What does
2:29:04industry need today? So, that we are
2:29:06training our university graduates. If
2:29:08you listen to me at KNUST, I told them
2:29:10the original partnership was to begin
2:29:12training from level 100. Now, we want to
2:29:14train level 400s. People who are ready
2:29:17to leave school. Train them, give them
2:29:19hands-on technical training in AI
2:29:22annotation, AI analytics.
2:29:24>> And Sam, because of Sam, let me ask you
2:29:25that
2:29:26Let me ask you the bread and butter
2:29:27question.
2:29:29I know Know mom is in global commerce.
2:29:31What does that mean for her?
2:29:32>> That's what I'm saying. If she takes if
2:29:34if for example and I went to a black
2:29:36market
2:29:36>> She may not have data. She may not have
2:29:37internet.
2:29:38>> no. So, go to the learning center. I
2:29:39went to the learning center in a black
2:29:40market. That's why we built learning
2:29:41centers as well with with devices. And
2:29:43in that classroom was a mother and her
2:29:46daughter.
2:29:47The mother was learning that the
2:29:48daughter was learning PHP full stack
2:29:50development. The mother was learning
2:29:52digital marketing and and and and
2:29:54digital digital marketing
2:29:56and e-commerce.
2:29:58And I asked the mother, "Why are you
2:29:59doing this with your daughter?" She
2:30:00said, "Oh,
2:30:01I've asked my daughter to learn the
2:30:03program. But me, I have a small shop and
2:30:05I want to be able to do a small online
2:30:06portal. And I was told that if
2:30:08[clears throat] I learn here, I can
2:30:09design my own portal." So, for that
2:30:12ordinary small-scale woman who runs you
2:30:15even call it an MSME, a small corner
2:30:16shop, what that skill she would get
2:30:19gives her is the ability to be able to
2:30:21now go online and open the frontiers of
2:30:23her sale. So, for the ordinary young
2:30:25Ghanaian who jumps on board the 1
2:30:28million coders program for example and
2:30:30gets a new skill. And that's why I keep
2:30:31saying the 1 million coders is a program
2:30:33that skills, reskills, and upskills.
2:30:36Because you may have a certain skill
2:30:37sets today like an interpreter who used
2:30:40to do regular interpretation. But today,
2:30:42we'll go and learn data annotation and
2:30:44can then begin to train the models that
2:30:46are doing the translation. So, you are
2:30:48reskilling or upskilling that
2:30:50individual. And that's what we're trying
2:30:51to do with our civil servants as well.
2:30:52Because a civil service that does not
2:30:55understand how to use AI for public
2:30:57service is useless in the 21st century.
2:31:00In this day and age. So, for the
2:31:01ordinary Ghanaian, there are
2:31:03opportunities. There are huge There are
2:31:05going to be job losses. No, let nobody
2:31:07kid you. Jobs will be lost. But a lot of
2:31:10new jobs will open up. How do we begin
2:31:13to position ourselves
2:31:14>> net impact on jobs will be positive.
2:31:16>> Positive, absolutely.
2:31:17>> Adam.
2:31:17>> Thank you so much. So, first to the
2:31:19point of you know, do we check these
2:31:21boxes first? I I don't think it's either
2:31:23or. I think it's and both. I'm I'm
2:31:26in agreement with the honorable minister
2:31:27there and I also believe in action. And
2:31:29the reason why I've been chomping at the
2:31:30bit to speak here within the final
2:31:32minutes is because I know my
2:31:34constituents that I speak with on a
2:31:35daily basis who are those average
2:31:37Ghanaians, who are the momo shop
2:31:39vendors, the barbers, the beauticians,
2:31:41you name it, right? They're listening to
2:31:43this panel right now and honestly even
2:31:46me who who who uses AI 12 hours a day, 7
2:31:48days a week, I'm a little bit lost and
2:31:50I'm learning myself amongst these these
2:31:52esteemed colleagues. But the people that
2:31:54I deal with on a day-to-day basis have
2:31:56no idea for a large part how this is
2:31:59going to affect them upstream,
2:32:00downstream. They want to know how they
2:32:01can benefit from this today. So, if I
2:32:03get 10 seconds, I would like to say
2:32:05>> Particularly that.
2:32:06>> to the people that are listening right
2:32:07now that are working hard and want more
2:32:10but don't know how to get there, who
2:32:12have dreams of just you know,
2:32:15the people I'm dealing with aren't
2:32:16trying to drive a Bentley down the road.
2:32:18They're trying to pay their school fees.
2:32:20They're trying to make sure that they
2:32:22can raise a family, right? They're
2:32:23trying to get a bicycle so they don't
2:32:24have to take the trotro, right? And the
2:32:27amount of difference you can make in
2:32:29these people's life is transformative.
2:32:31It can augment them in a way where
2:32:33>> How do you deploy it to benefit them?
2:32:34>> So, you you don't need to have money,
2:32:37you don't need to be a tech expert, you
2:32:38need to be able to articulate what the
2:32:40problem is you have and have a phone,
2:32:43right? And then have somebody point you
2:32:44in the direction and that's where AI for
2:32:45Ghana lives right there.
2:32:47>> Okay, just explain that cuz that for me
2:32:49is where I want to end and that's why
2:32:51when I know if I call him he would he
2:32:52would land me there.
2:32:54You know what the policy is.
2:32:56You've seen you have the critique but
2:32:59you say it must be about the ordinary
2:33:01Ghanaian and I agree with you. If you
2:33:03can lay for me in bullet points, what
2:33:05should we do
2:33:07so that it benefits that person? What
2:33:09would you say
2:33:10that is? You work with them.
2:33:12>> I I
2:33:13I think there's a lot of macro issues
2:33:15that are out of my lane so I'm not going
2:33:17to speak to them. What I can say
2:33:19>> What are the basic things we need to do
2:33:20with this AI policy you're trying to
2:33:22roll out that will benefit them.
2:33:24>> Here are the gaps. The gaps that I see
2:33:25are people don't know where to start.
2:33:28It's people don't know how to articulate
2:33:30what they need from the tool and people
2:33:32hit paywalls. Whether that's a lack of
2:33:34funding for the data or on the free
2:33:36tools, they hit a paywall which for
2:33:39those of you who don't know, they've
2:33:40used it too much, they have to wait till
2:33:41the next day, etc. For me, that's a good
2:33:43problem because it means that somebody
2:33:44I'm working with is using it enough that
2:33:46they hit a paywall. I love that. Well,
2:33:48what does somebody need to do? They need
2:33:50to be able to articulate what it is that
2:33:52their problem is and have a phone on on
2:33:55them, a smartphone. That's it. And
2:33:57somebody to point them in the right
2:33:59direction. And again, that's what AI for
2:34:00Ghana does is we cuz that's a universal
2:34:02problem as far as I can tell. The people
2:34:04I lead in the West, they they hear all
2:34:06this hype about AI and they don't know
2:34:07where it to apply it in their own
2:34:09>> So, what I learned China has done with
2:34:10what you're saying, they set up AI
2:34:11training centers across the country
2:34:14where every district can just walk in
2:34:16free and you're trained.
2:34:18Is that Are you recommending something
2:34:19like that for Ghana?
2:34:20>> No, what I'm what I'm saying is, you
2:34:21know, I think that's great. You know,
2:34:22it's an ecosystem, right? We're all
2:34:23working together. So, I I support any of
2:34:25the AI labs that are around, right? What
2:34:27we do at AI for Ghana is make it
2:34:29available to anyone for free,
2:34:31individuals, right? and we train them on
2:34:33WhatsApp, going back and forth. And the
2:34:34way we train them, we the way they learn
2:34:36AI is by applying it in a practical way
2:34:39to their life. So, they don't go to a
2:34:40session and learn about AI. You know,
2:34:42you see these courses, you learn deep
2:34:43seek and chat GPT and cloud and then
2:34:45you're like, "Okay, now what do I do?"
2:34:46Right? What we do is we ask some
2:34:48questions to ascertain exactly where
2:34:50they're at, what their skill level is,
2:34:52what they communicate. They can
2:34:53provide voice notes or they can type it,
2:34:55whatever they like. And then we put
2:34:57together a step-by-step game plan that
2:34:59starts them on the road with prompts,
2:35:01with which tools to download that apply
2:35:02to their system.
2:35:03>> Solve real problems in your life.
2:35:04>> Solve real problems and they're really
2:35:05the AI wall, they're solving the
2:35:07problems. So, look, I can't solve all
2:35:09these big problems and I love I love the
2:35:11you know, the argument between the two,
2:35:12right? But what I can say is that I
2:35:15believe in action, too, right? So, I'm
2:35:17my constituents are the ones that want
2:35:19to know how do I use this on Monday
2:35:20morning? How how do I do this to further
2:35:22my life? And the people that I'm working
2:35:23with are creating things that weren't
2:35:25there before. Whether that's money,
2:35:27whether that's time, whether that's an
2:35:28opportunity, whether that's a SWOT
2:35:30analysis of themselves because they
2:35:32don't they're like I don't know what to
2:35:33where to start. I have no money and and
2:35:34I don't know what to do but I want more.
2:35:35And there's no lack of ambition in
2:35:37Ghana.
2:35:37>> Mhm.
2:35:37>> And there's no lack of entrepreneurship.
2:35:40What did I say? What did I say? Right?
2:35:41It's like it's out there. So, you just
2:35:44need to be able to engage the people and
2:35:46show them where to start and let them go
2:35:48because there are people doing wonderful
2:35:50things on the ground level and that's
2:35:51what I'm excited about.
2:35:52>> Daniel, the same question to you based
2:35:54on your experience. How can we deploy
2:35:55this so that it answers an opportunity
2:35:57for for the majority of Ghanaians
2:36:00um in informal sector mainly because
2:36:01that's where the bulk of us sit.
2:36:04>> I mean, I think I think there are many
2:36:06tiers. I mean, you are pointing at
2:36:08informal sector but I think the first
2:36:09thing is that yes, there needs to be I
2:36:11mean, without access you are at a
2:36:13disadvantage and I think a lot of what
2:36:14is being discussed is providing access.
2:36:18Once access is provided, I think there's
2:36:21so many opportunities but ultimately
2:36:23people there's a layer of responsibility
2:36:26on the individuals to seize
2:36:28opportunities that are being provided.
2:36:30Whether that is the 1 million coder
2:36:32there just is.
2:36:33In my personal example,
2:36:35when I started to initially work with
2:36:38AI, it took me a very short time to let
2:36:42people go. I mean, I had a podcast
2:36:44somebody was writing my my show notes
2:36:48and eventually I probably most of the
2:36:50time still kind of wrote them. I
2:36:52employed two or three graduates in
2:36:54Ghana. I had people hired from Nigeria.
2:36:57When I put in the AI first time, I
2:37:00looked at it and I was like, okay.
2:37:01I'm sorry. It's just better than
2:37:04everything that they have done. But when
2:37:06you look at it, if somebody really had
2:37:08the skill and would go deep, that person
2:37:10could have
2:37:11actually been doing 10 jobs and they
2:37:13could have done the same show no jobs
2:37:15for somebody in the US and so on and I
2:37:17work with a lot of people. We also do
2:37:19trainings not necessarily for the
2:37:20informal sector, but a lot of Gani young
2:37:23Gani ins are able to have a full day
2:37:26time job that we work with. They are
2:37:28doing jobs for people outside of the
2:37:30country and it's all based on skill.
2:37:33But of course before the skill there is
2:37:35the access part that they have been
2:37:37discussing, but I think there is a lot
2:37:38of opportunity
2:37:40and it's a question of how do we seize
2:37:43it and I think the only way that works
2:37:45is to be entrepreneurial and I think
2:37:48even at
2:37:50politic as a politician as a minister
2:37:53being entrepreneurial and seizing the
2:37:54opportunity we can't stay away and
2:37:57opportunities will open up and then it
2:37:58goes down private sector the same, but
2:38:01ultimately responsibility on the
2:38:03individual as well.
2:38:05>> And I want to wrap up minister. I think
2:38:08he said something about the approach for
2:38:10the ordinary person
2:38:12not just train them, but train them in
2:38:14the way that they useful in your daily
2:38:16lives and your businesses to be
2:38:18productive. Um,
2:38:20is there an approach that can be
2:38:20skilled?
2:38:21>> I mean that's the whole essence of the 1
2:38:22million coders program.
2:38:24>> Okay.
2:38:24>> It's it's putting
2:38:26giving people real giving people real
2:38:29world skill sets that are applied in
2:38:33real world situations. Okay and so once
2:38:36again I can't say it more more than
2:38:38enough. Please go to
2:38:40www.1millioncoders.gov.gh
2:38:44and take one of the course. Currently
2:38:46there are about 37 different programs
2:38:48that are running on that portal that
2:38:50will help us and and and and give skill
2:38:52sets. It's extremely important. The last
2:38:54point I just want to make is about the
2:38:57use of foreign resource
2:39:01and even the
2:39:03the perception of a tailored arrangement
2:39:06that gives you sovereignty over your
2:39:08data.
2:39:09That is a mirage that is sold to many
2:39:12African companies, not just Ghanaian
2:39:13companies. When you read the US Cloud
2:39:16Act,
2:39:17cuz the US Cloud Act
2:39:19gives access
2:39:21from all top 10 US tech companies
2:39:26to the US government without need of
2:39:28informing the data subject
2:39:31and without a court warrant if the
2:39:33subject is not a US citizen to access
2:39:35your data without your knowledge. So,
2:39:38I've had this back and forth with Google
2:39:40and Amazon Web Services where they give
2:39:42you all kinds of agreements that say you
2:39:45are protected and data sovereign. But
2:39:47the US Cloud Act is an act of Congress
2:39:50that supersedes any such arrangement
2:39:52that they may have with you as a private
2:39:54company. And so, data residency and
2:39:56sovereignty, and especially when you
2:39:57look at what they tried to do with our
2:39:58health records, it's extremely important
2:40:01that we begin to think about how we sign
2:40:04those agreements and how we utilize
2:40:07those platforms. And that's why we need
2:40:09something that's
2:40:09>> Thank you, Sam. And Bright, I know you
2:40:11have final words, and do it for me a
2:40:12minute, but in doing so, think about the
2:40:14ordinary person as you've just I want to
2:40:15end there.
2:40:16How How do cuz we're going to do this.
2:40:19How do we do this in the way that
2:40:20benefits them?
2:40:21>> detail is is boring, but you can avoid
2:40:23it. So, for instance, the minister is
2:40:24talking about the Cloud Act. The fact is
2:40:26that every Ghanaian I don't I don't want
2:40:27to do the Cloud Act. I'm just I'm just
2:40:28saying that almost every Ghanaian uses
2:40:30foreign email service.
2:40:31>> Yeah.
2:40:32>> So, if we worried about our data being
2:40:33exposed to the US government, then we
2:40:35have to literally build our own email
2:40:37services because already the US
2:40:39government sees everything that you put
2:40:40on Gmail or any of those things on
2:40:41Google.
2:40:41>> Yeah.
2:40:42>> The second point is that
2:40:44um when you look at Uber and all the uh
2:40:47ride-sharing apps and all the e-commerce
2:40:48apps, a lot of the people that have
2:40:50benefited are ordinary people. Nobody
2:40:52taught them how to use mapping services.
2:40:54There were very little attempts to do
2:40:56that. Somehow, we figured out how to use
2:40:58mapping services, how to do geolocation,
2:41:00and deliver food and the rest of it. I'm
2:41:01a big believer in the fact that when it
2:41:03comes to those kind of things, typically
2:41:04the government is an impediment rather
2:41:06than an enabler.
2:41:07There was a time we were thinking about
2:41:08licensing Uber in this country and
2:41:10licensing boats and the others and
2:41:12bringing all these onerous
2:41:13obligations at the local government
2:41:16level. And some of us pushed against it
2:41:17because we hadn't thought it through
2:41:18carefully and looking at the you know
2:41:20the effects. So when we think about
2:41:21ordinary people, we shouldn't always
2:41:22assume that the delivery has to be
2:41:25direct to them before we understand the
2:41:26value of things like policy. Somebody
2:41:28says theoretical, but it's not
2:41:29theoretical. If they introduce new
2:41:30licensing for developers in Ghana that
2:41:33are not well thought through, you saw
2:41:34the the uproar that that we saw in the
2:41:36country. So we have to think about
2:41:38policy as also to include the things
2:41:41that remove barriers and bottlenecks
2:41:44for ordinary people. Because when it
2:41:45comes to the basic capabilities, map
2:41:48reading, WhatsApp use, who builds
2:41:50schools to teach people how to use
2:41:51WhatsApp? Who builds schools to teach
2:41:53influencers how to do YouTube
2:41:55influencing? And now we have people
2:41:57getting diplomatic passports because
2:41:58they're amazing YouTubers. Nobody builds
2:42:00schools and things like that to develop
2:42:02it. So I think that policy is very
2:42:03critical and in serving ordinary people,
2:42:06it's really the case that we directly
2:42:08serve them. Often there are impediments
2:42:10and bottlenecks in their way and people
2:42:11like us who point them out should not be
2:42:13seen as discussing theory. It's very
2:42:15practical.
2:42:15>> Uh Bright O Gyamfi Adam, uh Sam George,
2:42:19uh Daniel, thank you very much for
2:42:20joining us. And to my audience, a clap
2:42:22for yourselves by the way. Uh thank you
2:42:24very much. Uh
2:42:25>> [applause]
2:42:26>> uh and and I'm grateful that we could
2:42:28come together have the conversation. Uh
2:42:30it's very clear to me that we need to do
2:42:32more of this because the ship has
2:42:34sailed. The question is are we even
2:42:37moving fast enough? Um but we have to
2:42:39move because that is the future.
2:42:42Um and as we've heard from those who are
2:42:44experiencing it, he lost a job but he
2:42:47gained more from the AI revolution, if
2:42:50you like. And so as has been agreed, I
2:42:52think there's a consensus that the net
2:42:54>> [music]
2:42:54>> effect of this will be positive,
2:42:56correct? Of everything we've talked if
2:42:58we do it if we get it right. And we must
2:43:00get it right. Enjoy the rest of your
2:43:02evening till we return again with the
2:43:03next edition of the Digital Economic
2:43:05Forum in December.
2:43:07We see you.
2:43:08Have a good night.
2:43:18>> [music]
2:43:37>> Hey Grace, did you see that? That is me.
2:43:41Without me that tank would just be
2:43:42decoration.
2:43:45So when the water comes, where exactly
2:43:47is he waiting?
2:43:53At least I move water.
2:43:56And I keep it safe. Without me, nothing
2:43:58flows.
2:44:02What's happening? I bring the life. I
2:44:04preserve it.
2:44:07Why is the water going on and off like
2:44:08this?
2:44:11Maybe I can do it alone. Maybe I need
2:44:14you too.
2:44:18>> [music]
2:44:18>> It's strong, reliable, built for every
2:44:22home.
2:44:25fast and tough
2:44:27water tank
2:44:29Syntex is strong and you're tough.
2:44:33>> [music]
2:44:34>> Sister Grace, why didn't you finish
2:44:36medical school?
2:44:40>> [music]
2:44:41>> My parents didn't have the money.
2:44:47>> [music]
2:44:50>> So you know I want to become a fashion
2:44:52designer.
2:44:53>> [music]
2:44:53>> I've done research.
2:44:55Three schools, one in Accra, one in
2:44:58Lagos, and one in Paris. But, can't talk
2:45:02about [music] Paris later. When the time
2:45:04comes, I don't want you to say you don't
2:45:05have the money.
2:45:06>> [music]
2:45:07>> So, I have a plan. Ecobank Junior Saver
2:45:10Account. You open this for me today.
2:45:13By the time I'm done with school,
2:45:14[music] the money will be plenty. If you
2:45:18agree, please sign here.
2:45:20>> Ecobank Junior Saver Account.
2:45:23>> [music]
2:45:26>> Just a little bit
2:45:28>> Save for who they are becoming.
2:45:31>> [music]
2:45:32>> Just a little today.
2:45:34>> Ecobank, the Pan African Bank.
2:45:40>> [screaming]
2:45:51>> Who is the good? Ghana Jollof or
2:45:53Nigerian Jollof?
2:45:54>> Ghana Jollof has no coequal.
2:45:57>> Smell alone.
2:45:58>> Oh my God.
2:46:00>> You tell them I lie.
2:46:05>> You are lying.
2:46:05>> [screaming]
2:46:17>> Put some respect on the goat.
2:46:19>> On the goat?
2:46:20>> Only goat I know lives in Tema. Every
2:46:21year we the give to you back back.
2:46:24>> We are talking about beef.

About this transcript

This page contains the full transcript of Digital Economic Forum: AI and the Ghana Economy. by JoyNews, generated from the public captions YouTube serves with the video. The transcript has 29,388 words across 4,792 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.