YT YouTube2Text

YouTube2Text → Transcripts → AI AGENTS DEBATE: These Jobs Won't Exist In 24 Months!

AI AGENTS DEBATE: These Jobs Won't Exist In 24 Months! — Transcript

by The Diary Of A CEO · 27,465 words · 3,903 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00I think a lot of people don't realize
0:01how massive the positive impact AI is
0:04going to have on their life. Well, I
0:06would argue that the idea that this AI
0:07disruption doesn't lead us to human
0:09catastrophe is optimistic. For example,
0:12people are going to be unemployed in
0:14huge numbers. You agree with that, don't
0:16you? Yes. If your job is as routine as
0:19it comes, it's gone in the next couple
0:21years. But it's going to create new
0:23opportunities for wealth creation. Let
0:26me put it to you this way. We have
0:27created a new species and nobody on
0:29earth can predict what's going to
0:31happen. We are joined by three leading
0:33voices to debate the most disruptive
0:35shift in human history, the rise of AI.
0:38And they're answering the questions
0:39you're most scared about. This
0:41technology is going to get so much more
0:43powerful. And yes, we're going to go
0:44through a period of disruption. But at
0:46the other end, we're going to create a
0:47fair world. It's enabling people to run
0:49their businesses, make a lot of money,
0:51and you can solve meaningful problems
0:53such as the breakthroughs in global
0:54healthcare and education will be
0:56phenomenal. and you can live an
0:57incredibly fulfilling existence. Well, I
0:59would just say on that front, this has
1:01always been the fantasy of technologists
1:02to do marvelous things with our spare
1:04time, but we end up doom scrolling,
1:06loneliness epidemic, right? Falling
1:07birth rates. So, the potential for good
1:09here is infinite and the potential for
1:11bad is 10 times. For example, there's
1:14war, undetectable beat fakes and scams.
1:17So, people don't understand how many
1:18different ways they are going to be
1:20robbed. Look, I don't think blaming
1:22technology for all of it is the right
1:23thing. all these issues, they're already
1:25here. You're all fathers here. So, what
1:27are you saying to your children? Well,
1:29first of all, this has always blown my
1:32mind a little bit. 53% of you that
1:34listen to this show regularly haven't
1:36yet subscribed to the show. So, could I
1:38ask you for a favor before we start? If
1:40you like the show and you like what we
1:41do here and you want to support us, the
1:42free simple way that you can do just
1:44that is by hitting the subscribe button.
1:46And my commitment to you is if you do
1:47that, then I'll do everything in my
1:49power, me and my team, to make sure that
1:51this show is better for you every single
1:52week. We'll listen to your feedback.
1:54We'll find the guest that you want me to
1:56speak to and we'll continue to do what
1:58we do. Thank you so
1:59[Music]
2:05much. The reason why I wanted to have
2:08this conversation with all of you is
2:10because the subject matter of AI, but
2:12more specifically AI agents, has
2:15occupied my free time for several weeks
2:18in a row. And actually amad when I
2:20started using replet for me it was a
2:24paradigm shift. There was two paradigm
2:26shifts in a row that happened about a
2:28week apart. Chat GPT released their
2:30image generation model where you could
2:31create any image. It was incredibly
2:33detailed with text and all those things.
2:35That was a huge paradigm shift. And then
2:37in the same week I finally gave in to
2:39try and figure out what this term AI
2:41agent was that I was hearing all over
2:42the internet. I heard vibe coding. I
2:44heard AI agent. I was like I will give
2:45it a shot. Mhm. And when I used Replet,
2:4820 minutes in to using Replet, my mind
2:51was blown. And I think that night I
2:53stayed up till 3 or 4 a.m. in the
2:55morning. For anyone that doesn't know,
2:58Replet is a piece of software that
3:00allows you to create software. Mhm. And
3:04pretty much any software you you want.
3:06So someone like me with absolutely no
3:07coding skills was able to build a
3:10website, build in Stripe, take payment,
3:13integrate AI into my website, add Google
3:16login to the front of my website and do
3:18it within minutes. I then got the piece
3:21of software that I had built with no
3:22coding skills, sent it to my friends,
3:23and one of my friends put his credit
3:25card in and paid. Amazing. So I just
3:27launched a SAS company with no coding
3:28skills.
3:30to demonstrate an AI agent in a very
3:33simple way. I used an online AI agent
3:36called operator to order us all some
3:38water from a CVS around the corner. The
3:40AI agent did everything end to end and
3:42people will be watching on the screen.
3:43It put my credit card details and it
3:45picked the water for me. It gave the
3:46person a tip. It put some delivery notes
3:48in at some point a guy is going to walk
3:50in. He has not interacted with a human.
3:53He's interacted with my AI agent. And I
3:55just the reason I use this as an example
3:57is again it was a paradigm shift moment
3:58for me when I heard about agents. Mhm.
4:00about a month ago and I went on and I
4:02ordered a bottle of electrolytes and
4:04when it when my doorbell rang I freaked
4:07out. I freaked out. But am who are you
4:11and what are you doing? So uh I started
4:15programming at a very young age. You
4:17know I I built my first business when I
4:19was a teenager. I used to go to uh
4:22internet cafes and program there. And I
4:24realized that they don't have software
4:26to manage the business. I was like oh
4:27why don't you create accounts? I don't
4:29have a server. It took me two years to
4:31build that piece of software. And that's
4:33sort of embedded in my mind this idea
4:35that hey like you know there there's a
4:38lot of people in the world with really
4:40amazing ideas especially in the context
4:42where they live in that allows them to
4:44build uh businesses. However, the main
4:49source of uh friction between an idea
4:52and software or call it an idea and
4:55wealth creation is u infrastructure is
4:59physical infrastructure is a meaning a
5:01computer in front of you. It is um an
5:04internet connection. It is the set of
5:07tools and skills that you need to build
5:09that. If we make it so that anyone who
5:12has ideas who who wants to solve
5:15problems will be able to do it. I mean
5:17imagine the kind of world that we we
5:19could live in where no one can be anyone
5:22who has merit anyone who can think
5:25clearly anyone who can generate a lot of
5:28ideas can generate wealth. I mean that's
5:31an amazing world to live in right
5:33anywhere in the world. So with with
5:34Replet the company that I started in
5:362016 the the idea was like okay coding
5:40is difficult how do we solve coding and
5:43um we built every part of the process
5:45the hosting the code editor the only
5:48missing thing was the you know the AI
5:50agent and so over the past two years
5:53we've been working on this AI agent that
5:57you can just you know similar since chat
5:58GPT you know this revolution with Genai
6:01and you can just uh speak your ideas
6:03into existence. I mean this starts you
6:05know sounding religious like this is
6:07like the you know the gods you know that
6:09the myths that that um that humans have
6:12created they used to imagine a world
6:14where you can you can be everywhere and
6:17anywhere at once that's sort of the
6:18internet and you can also speak your
6:20ideas into into existence and um you
6:23know it's it's still early I think uh
6:25replet agent is is a fantastic tool and
6:27I think this technology is going to get
6:29so much more powerful
6:31specifically what is an AI agent. I've
6:34got this um graph actually here which I
6:36don't need to pass to any of you for you
6:38to be able to see the growth of AI
6:41agents. But this graph is Google search
6:43trend data. This also resembles our
6:45revenue too.
6:47Oh, okay. Right. The the water has
6:49arrived. Hello. Thank you. You can come
6:51on in. Can I have a go, please? Yes.
6:54It's
6:553951. Great. Thank you so much. Thank
6:58you. Thank you. I mean this is this is
7:00like a supernatural kind of power. You
7:03conjured water. I conjured water from my
7:05mind. Yeah. And it's shown up here with
7:08us and it clearly thinks we need a lot.
7:11But but just to define the term AI agent
7:14for someone that's never heard the term
7:15before. Yeah. Yeah. So uh I assume most
7:18of the audience now are familiar with
7:20chat, right? You can go in and you can
7:22talk to an AI. It can search the web for
7:25you. It has a limited amount of tools.
7:27Uh maybe it can call a calculator to do
7:30some addition subtraction for you, but
7:32that's about it. It's a request response
7:35style. Agents are when you give it a
7:38request and they can work indefinitely
7:42until they achieve a goal or they run
7:45into an error and they need your help.
7:47It's an AI bot that has access to tools.
7:50Those tools are access to the to a web
7:53browser like operator, access to a
7:55programming environment say like replet,
7:58access to um you know credit cards. The
8:02more tools you give the agent, the more
8:03powerful it is. Of course there's all
8:05these consideration around security and
8:07safety and all of that stuff. But uh the
8:10the most important thing is the AI agent
8:12will determine when it finished
8:15executing. Uh today AI agents can run
8:18for anywhere between you know 30 seconds
8:22to 30 minutes. Uh there's a recent paper
8:25that came out that's showing that every
8:287 months the number of minutes that the
8:31agent can run for is doubling. So we're
8:34at like 30 minutes now. In seven months
8:36we're going to be at an hour then you
8:38know 2 hours. Pretty soon we're going to
8:40be at days. And at that point, you know,
8:43AI agent is doing labor is doing kind of
8:45humanlike labor and actually uh OpenAI's
8:49new model 03 beat the expectation. So it
8:52it sort of doubled coherence over long
8:55horizon tasks in just three or four
8:58months. So we're in this massive and I
9:01mean this looks this exponential graph,
9:04you know, that shows you the massive
9:06trend we're on.
9:09Brett, give us a little bit of of your
9:11background, but also I saw you writing
9:13some notes there. There was a couple of
9:15words used there that I thought were
9:16quite interesting, especially
9:18considering what I know about you. The
9:19word God was used a few times.
9:22Well, uh, let me just say I'm an
9:24evolutionary biologist, and probably for
9:27the purposes of this conversation, it
9:30would be best to think of me as a
9:32complex systems theorist. One of the
9:34things that I believe is true about AI
9:36is that this is the first time that we
9:40have built machines that have crossed
9:42the threshold from the highly
9:44complicated into the truly complex.
9:48And I will say I'm listening to this
9:51conversation with a um a a mixture of
9:55profound hope and dread
9:59because seems to me that it is obvious
10:03that the potential good that can come
10:06from this technology is effectively
10:09infinite. But I would say that the harm
10:11is probably 10 times. It's a bigger
10:13infinity. And the question of how we are
10:16going to get to a place where we can
10:19leverage the obvious power that is here
10:22to do good and dodge the worst harms. I
10:25have no idea. I I know we're not
10:27prepared. So I hear you talking about
10:31agents and I think um that's marvelous.
10:34We can all use such a thing right away
10:37and the more powerful it is, the better.
10:38The idea of something that can solve
10:40problems on your behalf while you're
10:43doing something else is marvelous. But
10:45of course, that is the precondition for
10:50absolute devastation to arise out of a
10:53miscommunication, right? to have
10:55something acting autonomously to
10:58accomplish a goal, you damn well better
11:01understand what the goal really is and
11:02how to how to pull back the reins that
11:06it starts accomplishing something that
11:08wasn't the goal. The potential for abuse
11:11is also utterly profound. You know, you
11:14can imagine, just pick your your dark
11:18mirror
11:19uh fantasy dystopia where something has
11:24been told to hunt you down until you're
11:25dead and it sees that as a, you know, a
11:29technical challenge.
11:31So, I don't know quite how
11:34to balance a discussion about all of the
11:38things that can clearly come from this
11:39that are utterly transcendent. I mean, I
11:42do think it is not inappropriate to be
11:45invoking God or biblical metaphors here.
11:49You know, you're uh producing water
11:52seemingly from thin air. I believe that
11:54does have an exact biblical parallel.
11:56Uh so, uh any case, the the power is
12:00here, but so so too is the need for
12:04cautionary tales, which we don't have.
12:06That's the problem is that there's no
12:07body of myth that will warn us properly
12:10of this tool because we've just crossed
12:11a threshold that is similar in its
12:16capacity to alter the world as the
12:19invention of writing. I really think
12:21that's that's where we are. We're
12:23talking about something that is going to
12:24fundamentally alter what humans are with
12:28no plan. You know, writing alters the
12:31world slowly because the number of
12:33people who can do it is tiny at first
12:35and remain so for thousands of years.
12:38This is changing things weakly and
12:41that's an awful lot of power to just
12:43simply have dumped on a system that
12:45wasn't well regulated to begin with.
12:48Dan, yeah. So, I'm an I'm an
12:50entrepreneur. Um, I've been building
12:51businesses for the last 20 plus years.
12:53I'm completely well positioned between
12:56the two of you here. the excitement of
12:58the opportunity and the terror uh of
13:01what could go on. There's this image
13:02that I saw of New York City in 1900 and
13:08every single vehicle on the street is a
13:10horse and cart and then 13 years later
13:12the same photo from the same vantage
13:15point and every single vehicle on the
13:17street is a car. And in 13 years all the
13:20horses had been removed and cars had
13:22been put in place. And um if you had
13:25have interviewed the horses in 1900 and
13:28said uh how do you feel about your level
13:30of confidence uh in in the world? The
13:33horses would have said well we've been
13:34part of uh humanity for you know horse
13:36and hoof hand and hoof for for many many
13:39years for for thousands of years.
13:41There's one horse for every three
13:42humans. Like how bad could it be? You
13:45know we'll always have a special place.
13:46will always be part of society. Um, and
13:51little did the horses realize that that
13:53was not the case. That the horses were
13:55going to be put out of of uh business
13:57very very rapidly. And to reason through
14:01analogy, you know, there's a lot of us
14:02who are now sitting there going, "Hey,
14:04wait a second. Does this make me a horse
14:05in 1900?" I think a lot of people don't
14:08realize how massive these kind of
14:10technologies are going to have as an
14:11impact.
14:12You know, one minute we're ordering a
14:14water and that's cute and the next
14:16minute it can run for days and in your
14:19words uh it doesn't stop until it
14:21achieves its goal and it comes up with
14:23as many different ways as it could
14:24possibly come up with to achieve its
14:26goal and in your words it better know
14:29what that goal is. I'm thinking a lot as
14:32Daniel's speaking about the vast
14:35application of AI agents and where are
14:37the bounds because if if this thing is
14:40going to get incrementally smarter well
14:43incrementally might be an understatement
14:44it's going to get incredibly smart
14:46incredibly quick and we're seeing this
14:48AI race where all of these large
14:50language models are competing for
14:51intelligence with one another and if
14:53it's able to traverse the internet and
14:56click things and order things and write
14:58things and create things and all of our
15:00lives run off the internet today. What
15:02can't it do? It's going to be smarter
15:04than me.
15:06No doubt it already is. And it's going
15:08to be able to take actions across the
15:10internet, which is pretty much where
15:11most of my professional life operates.
15:13It's like how I build my businesses.
15:15Even this podcast is an internet product
15:17at the end of the day because you can
15:19create we've done experiments now and I
15:21can show the graphs on my phone to make
15:22AI podcasts and they have we've just
15:25managed to get it to have the same
15:26retention as the dire CEO. Now with the
15:29image generation model retention as in
15:30viewer retention the percentage of
15:32people that get to one hour wow is the
15:34same now. So we can make the video, we
15:37can publish it, we can script it, you
15:39can synthesize my voice sounds like me.
15:42So what what is it going to be able to
15:44do? Mhm. And can you give me the variety
15:47of use cases that the average person
15:49might not have intuitively conceived?
15:51Yeah. So I I tend to be an an optimist
15:54and and part of the reason is because I
15:56try to understand the limits of the of
15:58the technology. What can it do is
16:00anything that we can any sort of set of
16:03human data that we can train it on. What
16:06can it not do is anything that uh humans
16:10don't know what to do because we don't
16:12have the training data. Of course, it's
16:14super smart because it integrates
16:17massive amount of knowledge that you
16:18wouldn't be able to read, right? It also
16:21much faster. It can run through massive
16:24amount of computation that you know your
16:26brain you can't even comprehend because
16:28all of that they're smart. They can take
16:31actions but we know the limits of what
16:34they can do because we trained them.
16:37They're able to simulate what a human
16:39can do. So the reason you were able to
16:41order the water there is because it was
16:44trained NSF data that includes clicking
16:48on Door Dash and ordering water. I
16:51applaud your optimism and I like the way
16:52you think about these puzzles, but I
16:54think I see you making a mistake that we
16:57are about to discover is very common
17:00place. So we have several different
17:02categories of systems. We have simple
17:05systems, we have complicated systems, we
17:09have complex systems and then we have
17:11complex adaptive systems.
17:14And to most of us, a highly complicated
17:18system appears like a complex system. We
17:20don't understand the distinction.
17:22Technologists often master highly
17:25complicated systems and they know, you
17:28know, for example, a computer is a
17:31perfectly predictable system inside.
17:32There's it's deterministic. Mhm. But to
17:35most of us, it functions, it's it is
17:38mysterious enough that it feels like a
17:40complex
17:41system. And if you're in the position of
17:44having mastered highly complicated
17:46systems and you look at complex systems
17:48and you think it's a natural extension,
17:50you fail to
17:51anticipate just how unpredictable they
17:54are. So even if it is true that today
17:57there are limits to what these machines
18:00can do based on their training data, I
18:04think the problem
18:05is to see what's going to happen. You
18:09really want to start thinking of this as
18:10the evolution of a new species that will
18:13continue to evolve. It will partially be
18:16shaped by what we ask it to do, the
18:18direction we lead it, and it will
18:19partially be shaped by things we don't
18:21understand. So, how does this computer
18:25that that we have work? Well, one of the
18:27things that it does is we plug them into
18:30each other using language. It's almost
18:33as if you've plugged an Ethernet cable
18:35in between human minds. And that means
18:38that the cognitive potential exceeds the
18:41sum of the individual minds in question.
18:45Your AIS are going to do that. And that
18:47means that our ability to say what they
18:48are capable of does not come down to
18:51well we didn't train it on that data. As
18:54they begin to interact that feedback is
18:56going to take them to capabilities we
18:59don't anticipate and may not even
19:00recognize once they become present.
19:02That's one of my fears. This is an
19:04evolving creature and it's not even an
19:08animal. If it were an animal, you could
19:09say something about what the limits of
19:11that capability are. But this is a new
19:14type of biological creature and it will
19:18become capable of things that we don't
19:20even have names for. Even if it didn't
19:23do that, even if it just stayed within
19:25the boundaries that you're talking
19:26about, you mentioned about it having
19:28median level intelligence. Well, that by
19:30definition means 50% of the people on
19:32the planet are less intelligent than uh
19:35AI. uh you know to a degree it's almost
19:38as if we've just invented a new
19:40continent of remote workers. Um there's
19:43billions of them. They've all got a
19:45masters or a PhD. They all speak all the
19:47languages. Anything that you could call
19:49someone or ask someone over the internet
19:51to do, they're there 24/7 and they're 25
19:54cents an hour. Uh if that um so like if
19:59that really happened like if we really
20:01did just discover that there were a
20:03billion extra people on the planet who
20:04all had PhDs and were happy to work
20:06almost for free that would have a
20:08massive disruptive impact on society.
20:10Like society would have to rethink how
20:12everyone lives and works and gets
20:15meaning. Um so like and that's that's if
20:18it just stays at a median level of
20:19intelligence. Like it's it's pretty
20:21profound. I still think it's it's a
20:23tool. this is power that is there to be
20:26harnessed by entrepreneurs. You know, I
20:29I think that the world is gonna get
20:31disrupted, right? Um and the the you
20:34know, the this you know post-war world
20:36that we created where you go through
20:39life, you go through 12 years of
20:40education, you get to college and you
20:43just check the boxes, you get a job. We
20:46can already see the fractures of that
20:48that it is, you know, this American
20:50dream is perhaps no longer there. And so
20:52I think the world has already changed.
20:55So but but like what are the
20:56opportunities? Obviously there are
20:58downsides. The opportunities is for the
21:00first time access to opportunity is is
21:03equal. And I I do think there's going to
21:05be more inequality. And the reason for
21:08this inequality is because actually
21:11Steve Jobs uh you know made this
21:13analogy. It's like the the best taxi
21:15driver in New York is like 20% better
21:18than the you know the you know average
21:21taxi driver. the best programmer uh can
21:24be 10x better. You know, we call we say
21:26the 10x engineer. Now, the variance will
21:30be in the thousandx, right? Like the
21:33best the best entrepreneur that can
21:36leverage those agents could be better uh
21:39could be a thousand times better than
21:41than someone who doesn't have the grit,
21:44doesn't have the skill, doesn't have the
21:46ambition. Right? So, so that that will
21:48create a world. Yes, there's massive
21:50access to opportunity, but there are
21:52people who will take seize it and then
21:55they'll seize it and then there'll be
21:57people who don't. I imagine it almost
21:59like a um a marathon race and AI has two
22:02superpowers. One superpower is to
22:04distract people um such as Tik Tok
22:06algorithm. That's right. And the other
22:08superpower is to make you hyper
22:10creative. So you become a hyper consumer
22:11or a hyper creator. And in this marathon
22:14race, the vast majority of people have
22:16got their shoes tied together cuz AI is
22:18distracting them. Some people are
22:20running traditional race. Some people
22:23have got a bicycle and some people have
22:25got a Formula 1 vehicle. And it's going
22:28to be very confronting when the results
22:31go on the scoreboard and you see, oh,
22:33wait a second. There's a few people who
22:35finished this marathon in about 30
22:37minutes. And there's a lot of us who
22:38finished in like 18 hours because we had
22:42our shoes tied together. And I can't
22:44understand if we've got equal
22:46opportunity why there's so much
22:47disparity between how fast it you know
22:50and do, you know, I'm using an analogy,
22:51but this idea that, you know, someone
22:54like a lot of people are going to start
22:55earning a million dollars a month and a
22:57lot of people are going to say, "Hey, I
22:58can't even get a job for $15 an hour."
23:00there's going to be this kind of
23:02interesting wedge. Well, but I I hear in
23:05what both of you are saying
23:07a kind of assumption that this will all
23:12be done on the up and up. And I do want
23:17to just I am not a doomer. I agree that
23:21the doomers are likely incorrect that
23:24their fears are misplaced. But I do
23:27think we have a question of a related
23:29rates problem.
23:30You know I said the potential for good
23:32here is infinite and the potential for
23:35bad is 10 times.
23:37Right? What I mean is there are lots of
23:41ways in which this obviously empowers
23:43people to do things that they were going
23:45to be otherwise stuck in the mundane
23:48process of learning to code and then
23:51figuring out how to make the code work
23:53and bring it to market and all of that.
23:55And this solves a lot of those problems
23:56and that's obviously a good thing.
23:58really the what we should want is the
24:00wealth creation object as quickly as we
24:03can get there. But the problem is you
24:06know as much as it that hyper creative
24:10individual is empowered to make wealth
24:13the person who is interested in stealing
24:16may be even more empowered. And I'm
24:18concerned about that at at a pretty high
24:21level. the the abuse cases may outnumber
24:24the use cases and we don't have a plan
24:27for what to do about that. Um to can I
24:31can I give you a quick uh like
24:32introduction here like the optimistic
24:34view? OpenAI invented uh GP the first
24:37version GPT came out in 2019. 2020 was
24:41GPT2 and so OpenAI you know now they get
24:44a lot of criticism and lawsuit from IL
24:47Musk that they're no longer open source
24:50right they used to be. The reason is in
24:52GPT2 they said we are we are no longer
24:56gonna uh open source this technology
24:58because it's going to create um
25:00opportunities for abuse such as you know
25:03influencing elections um you know
25:06stealing you know grandma's credit card
25:09and so on so forth. Wouldn't you say
25:11Brett that it is kind of surprising how
25:14little abuse we've seen so far?
25:17I don't know how much abuse we've seen
25:19so far. I don't know how any of us do.
25:21And I also even the example that you
25:25suggest where chat GPT is no longer open
25:29source to prevent abuse. I'm taking
25:31their word for it that that's the
25:33motivation. Where as a systems theorist,
25:36I would say, well, if you had a
25:38technology that was excellent at
25:42enhancing your capacity to wield power,
25:46then open sourcing it is a failure to
25:49capitalize on that. and that the most
25:52remunerative use is to keep it private
25:55and then either sell the ability to
25:58manipulate elections to people who want
26:00to do so or sell the ability to have it
26:02kept off the table for people who don't.
26:05And I would expect that that's probably
26:07what's going on. There's no if you have
26:09a technology as transformative as this,
26:12giving it away for free is
26:13counterintuitive, which leaves those of
26:16us in the public more or less at the
26:18mercy of the people who have it.
26:21So I I don't see the reason for comfort
26:25there. We are at the dawn of this
26:29radical transformation of humans that by
26:33its very nature as a truly complex and
26:38emergent innovation. Nobody on earth can
26:40predict what's going to happen. We can
26:43we're on the event horizon of something.
26:45And the problem is you know we can talk
26:48about the obvious disruptions the job
26:49disruption and that's going to be
26:51massive. And does that lead some group
26:54of elites to decide, oh well, suddenly
26:56we have a lot of useless eaters and what
26:58are we going to do about that? Because
27:00that conversation tends to lead
27:01somewhere very dark very quickly. Um,
27:04but I think that's just the beginning of
27:06the the various ways in which this could
27:08go wrong without the doomer scenarios
27:10coming into play. This is an
27:12uncontrolled experiment in which all of
27:15humanity is downstream. Yeah. So I I was
27:18trying to make the point that OpenAI has
27:21been sort of wrong about the uh the sort
27:25of uh how big of a potential for harm it
27:27is like you know I think we would have
27:30heard about it in the news like uh the
27:33sort of how much harm it's done and
27:35maybe you know some of it is working in
27:36the shadows but like the few incidents
27:39that we've heard about where you know
27:41the cause of LLM's large language models
27:43the technology that's powering chat has
27:46been a huge headliners uh like New York
27:49Times talked about this kid that was,
27:52you know, perhaps goated by some kind of
27:54chat software that, you know, helps
27:56teenagers to be less lonely into into
27:58suicide, which is which is tragic. And
28:00obviously, these are the kind of safety
28:02and and abuse uh issues that we want to
28:04we want to worry about. But these are
28:06kind of these isolated uh incidents and
28:09we do have open- source large language
28:12models. Obviously, the thing that
28:13everyone talks about is DeepSeek.
28:15Deepseek is uh coming from China. So
28:18what is Deepseek's incentive? You know
28:21perhaps the incentive is to destroy the
28:24AI industry in the US. Uh you know when
28:26they released Deepseek, you know the
28:28market tanked the market for Nvidia, the
28:30market for AI and all of that. But there
28:32is an incentive to open source. Meta is
28:34open sourcing Llama. Llama is another AI
28:37similar to Chat GPT. The reason they're
28:39open sourcing Llama and Zuckerberg just
28:41says that out loud is basically they
28:44don't want to be beholden to open AI.
28:47They don't sell AI as a service. They
28:49use it to build products. And there's
28:52this concept in business called
28:54commoditize your compliment because you
28:56need AI as technology to run your
28:58service. The best strategy to do is to
29:01open source it. So these market forces
29:04are going to create conditions that I
29:07think are actually beneficial. So I'll
29:10give you a few few examples. One is
29:13first of all the AI companies are
29:15motivated to create AI that is safe so
29:17that they can sell it. Second there are
29:20security companies investing in AIs that
29:23allows them to protect from the sort of
29:25malicious uh acting of of AI. And so you
29:28have the free market and we've always
29:30had that you know but generally as
29:32humanity we've been able to leverage uh
29:36the same technology to protect against
29:38the abuse. So I I don't really
29:41understand this and maybe this is
29:43actually this is the exact discussion
29:45that you would expect between somebody
29:47at the frontier of the highly
29:49complicated staring at a complex system
29:51and a biologist who comes from the land
29:53of the complex and is looking back at
29:56highly complicated
29:58systems. In game theory we have
30:00something called a colle collective
30:01action problem. And in the market that
30:05you're
30:06describing, an individual company has no
30:10capacity to hold back the abuses of AI.
30:15The most you can do is not participate
30:17in them. You can't stop other people
30:19from programming LLMs in some dangerous
30:23way. And you can limit your own ability
30:26to earn based on your own limitations of
30:28what you're willing to do. And then
30:30effectively what happens is the
30:31technology gets invented anyway. It just
30:34that the dollars end up in somebody
30:35else's pocket. So the incentive is not
30:38to restrain yourself so that you can at
30:40least compete and participate in the
30:42market that's going to be opened. And
30:46so the number of ways in which you can
30:49abuse this technology. Let's take a
30:51couple.
30:53What is to stop somebody from training
30:56LLMs
30:58on an
31:00individual's creative output and then
31:03creating an LLM that can out compete
31:07that individual can effectively not only
31:09produce what they would naturally
31:11produce over the course of a lifetime
31:12but can extrapolate from it and can even
31:15hybridize it with the insights of other
31:17people so that effectively those who
31:20have the LLM can train it on the
31:22creativity ity of others, not cut them
31:24in on the use of that insight. You can
31:27effectively end up putting yourself out
31:29of business by putting your creative
31:31ideas in the world where they get sucked
31:32up as training data for future LLMs.
31:36That is unscrupulous, but it's
31:38effectively guaranteed. In fact, it's
31:40already happening. So, that's a problem.
31:43And likewise, what what would stop
31:45somebody from interacting with an
31:48individual and training an LLM to become
31:51like a personalized con artist?
31:53Something that would play exactly to
31:55your blind spot. That does happen. That
31:56that is starting to happen. Um people
31:58get phone calls and it sounds like their
32:00daughter and I've I've lost my phone and
32:02I'm borrowing a friend's phone and all
32:04of that sort of stuff. What's what's
32:06interesting is that I I think you make a
32:08a really good point.
32:10I worry about the impact on society. And
32:12yet when I look at every single
32:14individual who uses AI regularly, it
32:18almost has nothing but profoundly
32:20positive impact on their life. I look at
32:23people like um I was just spending some
32:25time with my parents-in-law um who are
32:28in their 70s and early 80s and they use
32:31AI regularly for all sorts of things
32:33that they find incredibly valuable and
32:36that improves the quality of their life.
32:38I personally did a an M&A a mergers and
32:40acquisitions deal where I bought a
32:42company last year and the AI was so
32:46powerful at helping that process. The
32:48conversations were transcribed and they
32:50were turned into letters of intent and
32:52then press releases and uh legal
32:55documents and we probably shaved
32:58$100,000 worth of uh costs and and we
33:01sped up the whole process and it was
33:04pretty magical to see how how it could
33:06happen. With that said, you know,
33:09there's there's all of these like, well,
33:10$100,000 worth of lawyers didn't get
33:12paid, right? So, well, what I want to
33:14know, yeah,
33:17people upset about, but if we look back
33:20at the invention of the cell phone or
33:24the invention of the social media
33:26platforms, there would be every reason
33:28to have the exactly the same
33:30perspective, right? I remember the
33:31beginning of Facebook and I remember the
33:33idea that suddenly the process that used
33:36to afflict people where you would just
33:38lose touch with most of the people who
33:40had been important to you, that was not
33:43something that needed to happen anymore.
33:44You could just retain them permanently
33:46as a part of a diffuse
33:49uh social grouping that just simply grew
33:52and value was added. There's no end to
33:54how much good that did, but what it did
33:58to us was profound and not evident in
34:01the first chapter. Say the same thing
34:04about the cell phones and the dopamine
34:05traps and the way this has disconnected
34:08us from each other, the way it has
34:10disconnected us from nature, the way it
34:13has altered the very patterns with which
34:16we think. It has altered every
34:19classroom. Mhm. So, and those things I
34:22think are going to turn out to have been
34:24sort of minor foreshadowings of the
34:28disruption that AI will produce. So, I
34:31agree with you today. The amount you can
34:33do with AI, there's a tremendous amount
34:34of good. There's a little bit of harm.
34:36Maybe that's something we need to worry
34:37about. But as this develops, as we get
34:41to, you know, to peer over the edge of
34:44this cliff that we're headed to, I think
34:45we're going to discover that we can't
34:48yet detect the nature of the alteration
34:50that's coming. So, I just wanted to add
34:52some context to that cuz Amjad, I I saw
34:53the interview you did in a newsletter in
34:552023 where you said, "I wouldn't prepare
34:57for AGI in the same way that I wouldn't
34:59prepare for the end of days." It's
35:00effectively the end of days if the
35:02vision of AGI that some of these
35:04companies have comes to bear because
35:05it's called the singularity moment
35:07because you can't really predict what
35:08happens after that. And so like how
35:11would you even prepare for that and you
35:13want to prepare for the more likely
35:15world and that world that you can
35:16actually predict is a world where yes
35:19there's like a massive improvements of
35:21technology and there's like insane
35:23compounding effects of technology and
35:24it's pretty hard to keep up. From that
35:26it appeared that in 2023 you were saying
35:29a similar thing to Brett in terms of we
35:31can't see around the corner here because
35:32it is a singularity.
35:34Sorry you also used AGI artificial
35:37general intelligence. It'd be
35:38interesting to know what your definition
35:40of AGI is. say that. Yeah. So, so what
35:42what I was saying there is even if I'm
35:44wrong that you can actually create a
35:49unbounded seemingly conscious artificial
35:53intelligence that can entirely replace
35:57humans and can act autonomously in a way
36:00that even humans can't act and can
36:03coordinate across different AIs,
36:05different data centers to take over the
36:07world. Even if if that's so so the
36:09definition of AGI is artificial general
36:12intelligence meaning that AI can acquire
36:15new skills in efficiently in the same
36:18way that a humans can acquire skills.
36:20Right now AIs don't acquire skills
36:22efficiently you know they they require
36:24massive amount of energy and compute
36:26entire data set of compute to acquire
36:27these these skills and I think there's a
36:31again a limit on how general
36:34intelligence can get I think for most of
36:37the time we're lagging in terms of what
36:41humans are are capable of doing the
36:44singularity is based on this concept of
36:47intellig intelligence explosion so once
36:49Once you create an AGI, once you create
36:52an artificial general
36:53intelligence, that intelligence will be
36:56able to modify its own source code and
36:59create the next version that is much
37:01more intelligent. And the next version
37:03creates the next version and the next,
37:06you know, for infinity, right? Within a
37:09week, within a week, perhaps within
37:11milliseconds at some point. Yeah. Right.
37:13Uh because it might invent new computing
37:15substrate and and and all of that.
37:18perhaps they'll use quantum computing
37:19and and so then you have this
37:22intelligence
37:24explosion in a way that it is impossible
37:27to predict how the world is going to be
37:29and what I'm saying is this is sort of
37:31like an end of time story like how would
37:35you even prepare for that so if if
37:38that's coming like why would I spend my
37:41time preparing for I think it's unlikely
37:43to happen can't see around the corner
37:45yeah but I'd rather prepare
37:48what I was saying there. I'd rather
37:49prepare for the more likely world in
37:53which we have access to tremendous
37:55power, but the world's not ending and
37:58humans are still uh important.
38:02I don't I don't know why you say more
38:05likely. I mean, I I think the structure
38:07of your argument is is sound. You would
38:10prepare for the world that might happen
38:12for which you can prepare. There's
38:14literally no point in trying to prepare
38:16for a world You can't predict it all.
38:18The only thing you can do is just sort
38:20of upgrade your own skills and pay
38:23attention. But if I have one message for
38:25the technologists, it's that your
38:28confidence about what this can and
38:32cannot do is misplaced because you have
38:35without noticing stepped into the realm
38:38of the truly complex. In the truly
38:42complex, your confidence should drop to
38:45near zero. that you know what's going
38:47on. Are these things conscious? I don't
38:50know. But will they be highly likely
38:52they will become conscious and that we
38:53will not have a test to tell us whether
38:56that has happened. Elon Musk predicts
38:58that by
39:002029 we will have AI with us AGI that
39:04surpasses the combined intelligence of
39:06all
39:07humans. And Sam Alman actually wrote a
39:10blog three months ago that I read where
39:11he said we are confident now Sam Alman
39:14being the founder of open arrow which um
39:15created chatbt we are confident now that
39:17we know how to build AGI as we have
39:20traditionally understood it. When I put
39:22these things together, I go back to the
39:24central question of what role do humans
39:26have in in this in the sort of
39:29professional output in GDP creation. If
39:32it's smarter than all humans combined,
39:36if Elon Musk is correct there, and it's
39:38able to take actions across the internet
39:41and continue to learn. This is like a
39:43central question that I'm hoping I can
39:44answer today, which is like where do we
39:46go? Yeah. I I mean in my vision of the
39:49world we are in the creative seat. We're
39:52sitting there where um we we are
39:56controlling swarms of intelligent being
39:59to do our job. You know the way you run
40:01your business for example you're sitting
40:03at a computer you have an hour to work.
40:05Yeah. and you're going to launch like a
40:07thousand SDR you know sales uh you know
40:11representative to go like grab as as
40:13many leads as possible and you're
40:15generating new update on replet for for
40:18your website here and then uh on this
40:20side you you you're actually you you
40:23have an AI that's crunching data about
40:26your uh your existing business to to
40:28figure out how to improve it and these
40:30AIs are kind of somehow all coordinating
40:33together and I am trying to privilege
40:36the human like this is my my mission is
40:39to uh build tools for people. I'm not
40:42building tools for agents and agents are
40:45a tool and so ultimately not only do I
40:48think that humans have a privileged
40:51position in the in the world and in the
40:53universe. We don't know where conscious
40:56consciousness is coming from. We don't
40:58really have the science to explain it.
41:01Um I think humans are special. That's
41:04one side is is my belief that humans are
41:07are special in the world and another
41:09side which I understand that the
41:11technology today and I think for the
41:14foreseeable future y is going to be a
41:17function of us training data. So there
41:19was this whole idea like what if chat
41:21GPT generates uh pathogens. Well have
41:24you trained it on pathogens? They were
41:26doing that kind of stuff in Wuhan. know
41:28I mean I mean a lot of the biotech
41:30companies are essentially using
41:32artificial intelligence like I can think
41:34of Abselera I think it's Abselera in
41:37Canada their whole business is using AI
41:39to create new vaccines using artificial
41:42intelligence and bigger data sets than
41:44we've never have had before and I know
41:46cuz I was I was very close to one of the
41:47founders of people involved in Absela so
41:50that that work is going on anyway and if
41:51we think about Wuhan that's it's quite
41:53probably well known now that it came out
41:55of a lab and people working in a lab and
41:57in that scenario that had a huge impact
41:59and shut down the world. What I'm the
42:01central question I'd love to answer
42:02before I throw it back open to the the
42:04room is what jobs because I know that
42:07you have this perspective. What jobs are
42:10going to be made redundant in a world
42:11where I am sat here as a CEO with a
42:13thousand AI agents, right? I was
42:15thinking of all the names of my of the
42:17people in my company who are currently
42:18doing those jobs. I was thinking about
42:19my CFO when you talked about processing
42:21business data, my graphic designers, my
42:23video editors, etc.
42:25So what what jobs are going to be
42:26impacted? Yeah, all of those. Uh so I I
42:29think and what do they do? You maybe
42:31this is useful for for the audience. I
42:34think if your job is as routine as it
42:37comes, your job is gone in the next
42:40couple years. So meaning if you if in in
42:43those jobs for example uh quality
42:46assurance jobs data entry jobs you're
42:49sitting in front of a computer and
42:50you're supposed to click uh and and type
42:53things in a certain order operator and
42:55those technologies are coming on the
42:57market really quickly and those are
42:58going to displace a lot of a lot of uh
43:01labor accountants accountants noise yes
43:05I mean I've just pulled a ligament in my
43:07in my foot and they did an MRI scan and
43:09I had to wait a couple of days for
43:10someone to look at the MRI scan and tell
43:12me what it meant. Yeah, I'm guessing
43:14that that's gone. Yeah, I think I think
43:16the healthcare ecosystem is hard to
43:18predict because of regulation and and
43:21again there there's so many limiting
43:22factors on how this technology can
43:24permeates the economy because of
43:26regulations and and people's willingness
43:28to to take it. But you know things
43:30unregulated jobs uh that are
43:34purely text in text out. if your job,
43:36you know, you get a you get a message
43:38and you produce some kind of artifact
43:40that's like probably text or images that
43:43that job is is at risk. So, just to give
43:46you some stats here as well, about 50%
43:48of Americans who have a college degree
43:50currently use AI. The stats are
43:52significantly lower for Americans
43:54without a college degree. So, you can
43:55see how a splinter might emerge there
43:57and that crack will write widen because
44:00people like us at this table are all
44:01messing around with it. But my mom and
44:03dad in Plymouth in the southwest, rural
44:05England, haven't got like they just
44:07figured out iPhones. So like I got them
44:08an iPhone and now they're like texting
44:09me back. AI is a million miles away. And
44:12if I start running off with my AGI, my
44:13agents, that gap is going to widen.
44:15Women are disproportionately affected by
44:18automation, which is what you were
44:19talking about there. with about 80% of
44:21working women in an at risk job compared
44:23to just over 50% of men according to the
44:25Harvard Business Review and jobs
44:27requiring only a high school diploma
44:29have an automation risk of 80% while
44:32those requiring a bachelor's degree have
44:34an automation risk of just 20%. So we
44:37can see again how how this will cause a
44:39sort of a it's also a huge risk um with
44:42business processor out outsourcing which
44:44is essentially western countries sending
44:47jobs to India to the Philippines like at
44:50the moment millions of people have been
44:51lifted out of poverty through the
44:53ability to do those kind of business
44:55process auto outsourcing jobs and those
44:58are all going to go but these they're
45:00going to have a thousand employees but
45:01but but uh also uh these people are
45:04actually already transitioning to
45:05training AI eyes. Mhm. You know, so so
45:08there's going to be a massive industry
45:09around training AI until they're
45:11trained. Well, no, you you have to
45:14continuously acquire new skills and this
45:15is what I'm talking about. I mean, this
45:17is again if AI is a function of its
45:18data, then you need increasingly more
45:20data. And by the way, we ran out of
45:22internet data. I was actually thinking
45:24interestingly that this might not be
45:25great for the United States or the UK,
45:26the Western world because it is going to
45:28be a leveler where now a kid in India
45:30doesn't need a Silicon Valley office and
45:33$7 million in investment to throw up a a
45:35software company basically. My yeah my
45:38belief is that so I have a I have a more
45:40broad definition of AGI and the
45:42singularity and for me AGI is do we have
45:45artificial general intelligence in terms
45:47of generally speaking can AI just do
45:50stuff that humans used to be able to do
45:52and we've already crossed that point we
45:54have this general intelligence that we
45:56can now all access and 800 million
45:58people a week are now using uh chat GBT
46:00it's it's exploded in the last 3 months
46:03and then to me a
46:05singularity Uh when the first tractor
46:08went out onto a farm, for me that was a
46:10singularity moment. Uh because everyone
46:12who worked in farming, it used to take a
46:14100 people to plow a field and now a
46:16tractor comes along and two guys with
46:18the tractor can now plow the field in
46:20just as much time and now 98 people out
46:22of 100 are completely out of a job. We
46:25also always underestimate a technology
46:28if it does go on to change history. When
46:30you look back through cars, horses,
46:32planes, the Wright brothers just thought
46:34of a plane as being something that that
46:35army could use, we had no idea of the
46:38application. So someone said to me
46:40recently, they said, "When it does
46:41change the world, we underestimate the
46:43impact that it will change the world."
46:45And I see people now with their
46:47estimations of AI and AI agents already
46:49incredibly optimistic. And so if history
46:52holds here, we're undershooting the
46:54impact it's going to have. And I think
46:56this is the first time in my life where
46:57the industrial revolution analogies seem
47:00to fall a little bit short. Yeah.
47:02Because we've never seen intelligent
47:04that's like you could I could think of
47:06this as a I'm not an intelligent person
47:07on this, but I could see that as like
47:08the disruption of muscles, whereas this
47:11is the disruption of intelligence. That
47:14that's that's exactly the thing is that
47:16what makes human beings special is our
47:18cognitive capacity and very specifically
47:21our ability to plug our minds into each
47:23other. So that the sum is is uh or the
47:27whole is greater than the sum of the
47:29parts. That's what makes human cognition
47:31special. And what we are doing is we are
47:34creating a something that can
47:37technologically surpass it without any
47:39of the preconditions that make that a
47:44safe process. So yes, we've
47:45revolutionized the world how many
47:47different times? It's innumerable. But
47:49you know, we've we've made farming
47:51vastly more efficient. That's different
47:53than taking our core competency as a
47:56species and surpassing ourselves with
47:58the product of our of our labor. I think
48:01your question is a good one. Then then
48:03what does become? We only have one thing
48:05left. Um we have our muscles which we
48:08got rid of in the industrial revolution
48:10and then we have our intellect which is
48:11this digital revolution. Now we're left
48:13with emotions and agency. So we
48:15essentially the the the agency idea I
48:19think we used to judge people on IQ and
48:21now IQ is the big leveler and now going
48:23forward for the next 10 years we're
48:25going to look at are you a high agency
48:27person or a low agency person. Do you
48:29have the ability to get things done and
48:32coordinate agents? Do you have the
48:33ability to start businesses or give
48:36orders to digital armies? uh you know
48:39and and essentially these high agency
48:41people are going to thrive in this new
48:44world because they have this other thing
48:46that's been bubbling under the service
48:47surface which is really interesting when
48:49you said agency is going to remain as an
48:51important thing we're sat here talking
48:54about AI agents and the crazy thing in a
48:56world of AI agents that have super
48:58intelligence is I can just tell my agent
49:00listen I'm going on holiday please build
49:02me a SAS company that spots a market
49:04opportunity throw up the website post it
49:06on my social media channel I'll be in
49:08Hawaii I and this new agentic world is
49:12stealing that too cuz now it can take
49:14action in the same way that I can browse
49:16the internet. I can call Domino's Pizza,
49:19speak to their agentic agent, organize
49:21my pizza to be there before I even wake
49:22up. And in fact, predictability, you
49:24know, OpenAI now learns and Sam Alman
49:27said that they've expanded the memory
49:29feature. So, it knows every it's knowing
49:30more and more and more and more about
49:32me. It'll almost be able to predict what
49:34I want when I want it. It'll know
49:35Steve's calendar. He's arriving at the
49:37studio. Make sure his cadence is on the
49:39side. Make sure his iPad has the brief
49:42on it. Do the brief. Do the research for
49:44me. And everything else say remember
49:46Brett's birthday so when I arrive
49:48there'll be something. In fact, it's
49:50removing my need for any agency. Yes.
49:52And you know again I don't know how to
49:54make this point so that it occurs to
49:57people what I'm really suggesting
49:59but today maybe it's not conscious but
50:04well let me put it to you this way.
50:07If you're
50:08conscious, you started out as a child
50:11that wasn't. And although this may not
50:14fully encapsulate it, you are
50:15effectively an LLM, right? You go from
50:18an unconscious infant to a highly
50:21conscious adult. And the process by
50:24which you do that has a lot to do with
50:27being trained effectively on words and
50:30other things in an environment in
50:31exactly the way that we now train these
50:33AIs. So the idea that we can take
50:36consciousness off the table, it won't be
50:38there till we figure out how to program
50:39it in and we're safe because we don't
50:41know how consciousness works. I take the
50:42opposite lesson. We've created the exact
50:45thing that will produce that phenomenon
50:47and then we can have philosophers debate
50:49whether it's real consciousness or it
50:51just behaves exactly as if it were. And
50:52the answer is those aren't different.
50:54Doesn't matter. And the same thing is
50:56true for agency. you know, especially if
50:58you've created an environment in which
51:00these AIs are de facto competitors, what
51:04you're effectively doing is creating an
51:06evolutionary environment in which they
51:08will evolve to fill whatever niches are
51:10there. And we didn't spell out the
51:11niches. So, I have the sense we have um
51:15we have invited we have we have created
51:17something that truly is going to
51:19function like a new kind of life and
51:21it's especially troubling because it
51:23speaks our language. So that leads us to
51:26believe it's more like us than it is and
51:28it's actually potentially quite
51:30different. So but by the way, he's the
51:33optimist here, right? Like he's so
51:36optimistic about LMS and how how they're
51:39going to they're going to evolve. Yes.
51:42It's amazing. It's amazing technology.
51:44Like I think it raised global IQ, right?
51:47Like 800 million people like 800 million
51:49people are that much more intelligent
51:51and emotionally intelligent as well.
51:53Like I know people who previously were
51:56very coarse and they kind of robbed
51:58people the the the wrong way. They they
52:00would say things in not so polite way
52:03and then suddenly they started putting
52:05their the you know what they're saying
52:08through chat in order to kind of make it
52:11kinder and nicer and they're more liked
52:13now. And so not only is it uh making us
52:16more intelligent but also it allows us
52:18to be the best version of ourselves. And
52:21the the scenario that you're talking
52:22about, I don't think I don't know what's
52:24wrong with that. Like, you know, you
52:26know, the I would want less agency in
52:29certain places. Like, I would want
52:32something to help me not, you know, open
52:34up a peanut butter jar at night, right?
52:38You know, there are places in my life
52:40where I need more control and I would
52:44rather seed it to some kind of entity
52:46that could help me make better choices.
52:51I mean unfortunately even if there is
52:54some small group of elites that are able
52:58to go to Hawaii while something else
53:00does the mundane details of their
53:03business
53:04building. We are rather soon going to be
53:08faced with a world that has billions of
53:11people who do not have the skills to
53:14leverage AI. Some of them will be
53:17necessary for a time. you're going to
53:19need plumbers. But this is
53:23also not a long-term solution because
53:27not only are there not enough of those
53:29jobs,
53:30um, but of course we have humanoid
53:33robots that once imbued with AI capacity
53:37will also be able to take, you know,
53:39they'll be able to crawl under your
53:40house into the crawl space and fix your
53:42plumbing.
53:43So what typically happens when you have
53:47a massive economic contraction that
53:49arises from the fact that a huge number
53:52of people are out of work is that the
53:54elites start looking at those people and
53:56thinking well we don't really need them
53:57anyway. And so the idea that this AI
54:01disruption doesn't lead us to some very
54:03human catastrophe I think is overly
54:06optimistic and that we need to start
54:08preparing right now. What are the rights
54:10of a person who has had whatever it is
54:12that they've invested in completely
54:14erased from the list of needs? Is that
54:17person responsible for not having
54:18anticipated AI coming? And is it their
54:21problem that that they are now starving
54:23and they're being eyed by others as you
54:25know a useless eater? I don't think so.
54:27How is it different than uh when the uh
54:30uh what's it called the the looming
54:31machine came and the textile workers you
54:33know that the result of the in the lad
54:35sort of revolution? uh h how is it how
54:38is it different than any time in history
54:41when uh technology uh automated a a lot
54:45of people out of out of jobs? I would
54:47say scale and speed that's how it's
54:49different and the scale and speed is
54:51going to result in a an unprecedented
54:54catastrophe because the rate at which
54:56people are going to be simultaneously
54:58sidelined not just in one industry but
54:59across every industry is just simply uh
55:03and it also did actually happen. There
55:05was a there was an uh for the first 50
55:07years of industrialization from like
55:10late 1700s to early 1800s, you actually
55:13the Charles Dickens novels are
55:15essentially people coming from the farms
55:17who are displaced arriving in cities,
55:20kids living on the streets. Uh the
55:23British decided to pick everyone up and
55:24send them over to the over to Australia,
55:27which is where I came from. um and uh
55:31you know there there were this there was
55:33this massive issue of displacement. I
55:36think we're going to go into a high
55:37velocity economy where rather than this
55:39long arc of career that lasts 45 years,
55:42we're going to have these very fast
55:45careers that last 10 months to 36
55:48months. and you invent something, you
55:51take it to market, you put together a
55:53team of five to 10 people who work
55:55together, you then get disrupted, you
55:58come Can I mention a story here? Uh
56:00there's an entrepreneur that used Replet
56:03in a similar way. Uh his name is Billy
56:05Howell. You can find him on YouTube on
56:07the internet. He would go to Upwork and
56:10he would find what people are asking for
56:13different requests for certain apps,
56:15technologies. Then he would take the
56:18what they're asking for, put it into
56:19replet, make it an application, call
56:22them, tell them, "I already have your
56:24application. Would you pay $10,000 for
56:26it?" And so, so that's sort of an
56:27arbitrage opportunity that's that's
56:29there right now. That's not arbitrage.
56:31That's theft. How is no what is it? How
56:33is that theft? You have somebody who has
56:35an idea that can be brought to market
56:38and somebody else is cryptically
56:40detecting it and then selling back their
56:42own idea to them. Well, they're paying
56:44them to do that. They're saying, "I will
56:46give you $500 if you if someone makes
56:48this for me." Right? But this is what I
56:50more or less think is going to happen
56:51across the whole economy is that yes,
56:53from this perspective, we can see that
56:56everybody is suddenly empowered to build
56:58a great business. Well, what do we think
57:00about the folks who are going to be
57:02displaced from the top? What are they
57:04going to think about all these people
57:05building all of these highly competitive
57:07businesses? And are they going to find a
57:08way to do, you know, what venture
57:11capital has done or what record
57:13producers have done? What they're going
57:15to do is they're going to take their
57:16superior position at the top and they
57:18are going to take most of the wealth
57:20that is produced by all of these people
57:22who have these ideas that in a proper
57:24market would actually create businesses
57:26for them and they're going to parasitize
57:28them. I think that we with this in
57:33introduction of AI and AI
57:35agents old value has moved and now it's
57:39not going to be the case that the idea
57:40itself is the moat and it's not going to
57:43be the case that resources are the moat.
57:44So in such a scenario you still have to
57:46figure out distribution. You still have
57:48to have for example like an audience. So
57:50if you're a podcaster now you have a
57:51million followers on Twitter. you're in
57:53a prime position because you now have
57:55something that the the great guy with a
57:58great idea with no audience has you have
58:00inbuilt distribution. So I now think
58:01actually much of the game might be
58:02moving to like yeah still about taste
58:05and idea but also the the mo is
58:08distribution. Yeah. And speaking of
58:10adaptive systems um the one of the
58:13adaptation that will happen is people
58:15will seek uh humans and will seek proof
58:19of humanity. Oh, I agree that uh
58:21authenticity is going to become the coin
58:23of the realm and anything that can be
58:27faked or cheated is going to be devalued
58:30and things you know spontaneous jazz or
58:34you know comedy that is interactive
58:36enough that it couldn't possibly have
58:38been generated with the aid of AI those
58:40things are going to become prioritized
58:42you know spontaneous oratory rather than
58:44speeches answers some of your questions
58:47no it answers it answers my question for
58:49the tiny number of people who are in a
58:51position to do those things. Stephen,
58:54you use the mo word moat. Um, which I
58:56think is a really important word for
58:57entrepreneurs. We like we like have to
58:59have a moat. We think a lot about moes
59:01and it's an industrial age. A lot of
59:02people don't even know what a moat what
59:04you mean by moat. It's just I often
59:06think about this idea of what are the
59:08moes that are left. So to define how I
59:11define a mo, you've got a castle and
59:12it's got a like a small circle of water
59:15around it. And once upon a time that
59:17circle of water defended the castle from
59:19attack and you can pull up the
59:20drawbridge so nobody could attack you
59:21very easily. It's a de defense from
59:23something. So it's your it's your
59:25shield. It's your your defense. And once
59:27upon a time as an entrepreneur, you
59:28know, I've got a software company in San
59:29Francisco called Third Webb and we
59:31raised almost $30 million. We have a
59:33team of 50 great developers. And much of
59:36our moat was you can't compete with us
59:39if you don't have the 50 developers and
59:40the $30 million in the bank. How much of
59:42that 30 million went to coding? The vast
59:45majority of it. I mean what else are we
59:46going to do? What else we do? So this is
59:48a good thing. I think modes are a bad
59:50thing. Okay, let me make the argument
59:52there. Uh so everyone is looking for
59:54modes you know for example like one of
59:56the more uh significant modes is network
59:59effects. Yeah, you know, so you can't
1:00:02compete with Facebook or Twitter
1:00:05because to move people from Facebook or
1:00:08Twitter, you need to it's the collective
1:00:09action problem. You need to move them
1:00:11all at once because if one of them
1:00:13moves, then it's the network is not
1:00:17valuable, they'll go back. So you have
1:00:19this chicken and egg problem. Let's say
1:00:21that we have a more decentralized way of
1:00:24doing social networks that will remove
1:00:27the power of Twitter to kind of censor
1:00:31and I I think you're at the other end of
1:00:33of censorship, right? And so part of my
1:00:36optimism about humanity is that um
1:00:39generally there's self-correction.
1:00:41Democracy is a self-correcting system.
1:00:43Uh free markets are largely
1:00:45self-correcting systems. There are
1:00:47obvious problems with with free markets
1:00:49that that we can discuss. But take um
1:00:52health, you know, there is obesity uh
1:00:55epidemic. This period of time when uh
1:00:59companies, you know, ran loose kind of
1:01:02making this sugary, salty, fatty kind of
1:01:06snacks and everyone gorged on them and
1:01:08everyone got very uh you know,
1:01:10unhealthy. And now you have Whole Foods
1:01:13everywhere. Today, people in Silicon
1:01:15Valley, they don't go to bars at all.
1:01:17They go to running clubs. That's how you
1:01:19meet. That's how you go find a date. You
1:01:21go to running clubs. And so, there was a
1:01:24shift that happened because there was a
1:01:27reaction. Obviously, cigarettes is
1:01:29another example. You know, you were
1:01:31talking about phones and our addiction
1:01:33to phones. And I see a shift right now
1:01:35like in my uh friend circle like people
1:01:39who are constantly kind of on their
1:01:40phones is already kind of frowned upon
1:01:42and they don't want to hang out with you
1:01:44because you're you're constantly staring
1:01:45on on your phone. So there's always
1:01:47these reactions and and but the problem
1:01:49is you you reference selfcorrection and
1:01:52I agree that there's actually an
1:01:54automatic feature of the universe in
1:01:55which the self-correction happens. You
1:01:57can't have a positive feedback that
1:01:58isn't reigned in by some outer negative
1:02:01feedback. But the corrections, the list
1:02:04of corrections involves things like you
1:02:06point to where people become enlightened
1:02:08and they realize that they're doing
1:02:10themselves harm with either the sugar
1:02:12that they're consuming or the dopamine
1:02:14traps on their phone and they get
1:02:16better. But also on the list of
1:02:18corrective patterns are genocide and war
1:02:22and you know parasetism. And the problem
1:02:27is these things are destructive of
1:02:30wealth. And so you allude to
1:02:35the superior fact of an open market
1:02:38without moes. Presumably the benefit of
1:02:40that is that more wealth gets created
1:02:42because people aren't kept from doing
1:02:44things that are productive. I see that.
1:02:46But then what is the product of all of
1:02:49this new wealth that is going to be
1:02:50generated by a world empowered by AI?
1:02:53Does it end up so highly concentrated
1:02:55that you have a tiny number of ultra
1:02:58elites and a huge number of people who
1:02:59are utterly dependent on them? What
1:03:02becomes of those people? The learning
1:03:04process, the self-correction pro process
1:03:07goes through harm in order to get to
1:03:09that more enlightened solution. There's
1:03:11nothing that protects us from the harm
1:03:14phase being so apocalyptically terrible
1:03:17that, you know, we get to the other side
1:03:18of it and we say, "Well, that was a hell
1:03:20of a correction." Or maybe there's
1:03:21nobody there to even say that. Those are
1:03:23also on the table. It reminds me of a
1:03:24mouse trap where you see the cheese and
1:03:26we're going, "Oh my god, my
1:03:27grandmother's going to be able to do
1:03:28some research and oh my god, my life's
1:03:30going to get easier." So you head closer
1:03:31and closer to the cheese.
1:03:34And historically, if we look at all of
1:03:37the last 10,000 years, it's a very small
1:03:39number of elites who own absolutely
1:03:41everything and a very large number of
1:03:44surfs and peasants who have a
1:03:46subsistence living. You know, if the
1:03:48elites are too greedy and they freeze
1:03:50out the peasants at too high a level and
1:03:52they try to use brutality or Yeah.
1:03:55eventually it comes back to haunt them.
1:03:56And so what you get is a recognition
1:03:59that you you need a system that does
1:04:01balance these things and you know the
1:04:03west has the best system that we've ever
1:04:05seen. It's one in which we agree on a
1:04:07level playing field. We never achieve it
1:04:09but we agree that it's a desirable thing
1:04:11and the closer we get to it the more
1:04:13wealth we create. But again,
1:04:17if AI empowers those with ill
1:04:23intent at a higher rate that it empowers
1:04:26those who are wealth creating and
1:04:28pro-social, we may be in for a massive
1:04:32regression in how fair the the market
1:04:34that the West is. Is that your top
1:04:35concern versus economic displacement?
1:04:38And I think they're the same thing. How
1:04:41are they the same thing?
1:04:43because the economic displacement is
1:04:46going to start. I don't know how
1:04:48many million people are going to be
1:04:50displaced from their jobs in the US.
1:04:52Suddenly, we're going to have a question
1:04:54about whether or not we have obligations
1:04:56to them. And you agree with that, don't
1:04:58you? Yes. But but but again, it's it's
1:05:01the no pain, no gain. I mean, we're
1:05:02going to go through a period of of
1:05:03disruption. And I think at the other
1:05:05end, the old, you know, sort of
1:05:08oppressive systems will be broken and
1:05:10we're going to create perhaps a fair
1:05:13world, but it's going to have its own
1:05:15its own problems. And what's the scale
1:05:16of that disruption in your estimation?
1:05:19It's hard to say because uh you know
1:05:22there's this concept of limiting factors
1:05:24like you know there is um regulation
1:05:28there's the appetite of people to today
1:05:30for example the health care system is
1:05:33very resistant to innovation because of
1:05:35regulation you know and that's a that's
1:05:37a bad thing on the regulation point it's
1:05:39worth saying that when Trump came into
1:05:41power he signed in a new law which is
1:05:44called removing barriers to American
1:05:46leadership in AI which revokes previous
1:05:48AI policies that were deemed to be
1:05:50restrictive. And obviously when you
1:05:52think about where the funding is going
1:05:52in AI, it's going to two places. It's
1:05:54going to America and it's basically
1:05:56going to China. That's the the vast
1:05:58majority of investment. So with those
1:06:00two in competition, any regulation that
1:06:02restricts air in any way is actually
1:06:03self-sabotage. Mh. And this is, you
1:06:06know, I live in Europe some of the time
1:06:09and it's already annoying to me that
1:06:11when Sam Alman and OpenAI released the
1:06:1303 model, this new incredible model,
1:06:15it's not in Europe because Europe has a
1:06:17regulation which prevents it from coming
1:06:19to Europe. So, we're now at a
1:06:20competitive disadvantage
1:06:22um which Sam Alman spoken about. And
1:06:24more broadly on this point of
1:06:26disruption, it was I was quite unnerved
1:06:29when I heard that Sam Wolman's other
1:06:31startup was called
1:06:33Worldcoin. And Worldcoin was conceived
1:06:36with the goal of facilitating universal
1:06:39basic income,
1:06:41i.e. helping to create a system where
1:06:43people who don't have money are given
1:06:46money by the government just for being
1:06:47alive to help them cover their basic
1:06:49food and housing needs. Which suggests
1:06:51to me that the guy that has built the
1:06:52biggest air company in the world can see
1:06:54something that a lot of us can't see
1:06:56which is there. Yeah. There's gonna need
1:06:58to be a system to just hand out money to
1:07:00people because they're not going to be
1:07:01able to survive otherwise. I
1:07:03fundamentally disagree with that. Which
1:07:04part do you disagree with? I disagree
1:07:06that first of all that humans would be
1:07:08happy with UBI. I I think that you know
1:07:12uh you know core value of humans and be
1:07:16curious about the evolutionary reasons
1:07:17is we want to be useful. It's really
1:07:19important to know that a lot of the jobs
1:07:21that are in at risk are the most high
1:07:23status, highly paid jobs in the world.
1:07:26Let's take the highest paid job in
1:07:27America um which is an
1:07:29anesthesiologist. Uh this is the highest
1:07:32paid job and highest paid salary job.
1:07:35Salari job. Yeah. And the majority of
1:07:37that job is observing a patient, knowing
1:07:40which type of medication would work best
1:07:42with their body. um giving them the
1:07:45exact right amount, monitoring the
1:07:47impact of that uh on the on the body and
1:07:50then making slight adjustments, the
1:07:52right technology and any nurse will be
1:07:55able to do that job. And you might have
1:07:57one
1:07:58anesthesiologist on on
1:08:01site supervising 10, 20, 30, 40 wards
1:08:05and the technology is, you know, doing
1:08:07the job, but that one person is there
1:08:09just to kind of supervise if something
1:08:11went wrong or if there was an ethical
1:08:12dilemma. What's wrong with that? I mean,
1:08:13if if if the precision is better where
1:08:15they are. No, there's nothing wrong with
1:08:17that except for the fact that a lot of
1:08:20people, hundreds of thousands of people
1:08:21have spent their entire life training to
1:08:23be that, they get an enormous amount of
1:08:25purpose and satisfaction about the fact
1:08:27that that's their career, that's their
1:08:28job. They have mortgages, they have
1:08:30houses, they have status, and that's
1:08:33about to go away. Well, if it's highest
1:08:35paid jobs, maybe you should start
1:08:36saving.
1:08:38Yeah. Well, I mean, but yeah, I I hear I
1:08:43hear you, but you're talking about
1:08:45people who have done vital work. Mhm.
1:08:50Highly specialized work and are
1:08:51therefore not in a great position to
1:08:53pivot pivot based on the invention of a
1:08:56technology that they didn't see coming
1:08:58because frankly, I mean, in the
1:09:00abstract, maybe we all saw AI coming
1:09:02somewhere down the road, but we did not
1:09:04know that it was going to suddenly dawn.
1:09:06And we do have to figure out what to do
1:09:08with those people. It's not their fault
1:09:10that they've suddenly become obsolete
1:09:12and it's inconceivable that people will
1:09:16accept this. It is not. It is
1:09:18fundamentally incompatible with our
1:09:20nature. We have to have things to strive
1:09:22for and you know you can sustain life
1:09:26that way but you cannot um sustain a
1:09:30meaningful existence and so it's a
1:09:31short-term plan at best. Let's talk
1:09:33about meaning. Um on that point of job
1:09:36displacement this is already happening.
1:09:38Cler CEO who has been on this podcast
1:09:40before, a great guy, um said to on a
1:09:43blog post that they published on Cler's
1:09:45website saying that they now have AI
1:09:47customer service agents handling 2.3
1:09:50million chats per month, which is equal
1:09:52to having to hire 700 full-time people
1:09:55to do that. So, they've already been
1:09:57able to save on 700 customer service
1:10:00people by having air agents to do that.
1:10:03And they actually they actually got rid
1:10:04of those 700 jobs, right? I don't have
1:10:07that information in front of me, but
1:10:08I'll have a look. Um, I'll throw it up
1:10:10on screen for anyone that wants context
1:10:12on that. But that's already happening.
1:10:13This isn't some antologist or something.
1:10:15And these aren't high paid people in
1:10:17every case. We've done something
1:10:19similar, by the way. We've we internally
1:10:20we've
1:10:22we've replaced that function for 70%.
1:10:26Yeah. I mean, our company, we're 65
1:10:29people and, you know, we um, you know,
1:10:32we make, you know, millions per per
1:10:34head, you know. So, it's a Are you going
1:10:36to need to hire more people to get up to
1:10:38I think so, but we're we're hiring
1:10:40slowly, like, you know, we're we're
1:10:41using uh customer support, AI, and that
1:10:44meant that we we need less uh customer
1:10:47support, and we're trying to leverage AI
1:10:50as much as possible. the the person in
1:10:53HR at Replet writes software using
1:10:56Replet. So, I'll give you an example.
1:10:57She needed um Orc charts uh software and
1:11:01she looked at a bunch of them, got a lot
1:11:03of demos and they're all very expensive
1:11:06and they they're missing the kind of
1:11:08features that she want. For example, she
1:11:09wanted like version control. She wanted
1:11:11to know when when something changed and
1:11:13to go back in history. She went into
1:11:15replet in 3 days she got exactly the
1:11:18kind of software that she wanted and
1:11:20what was the cost you know perhaps $20
1:11:22you know something like that $ 20 $30
1:11:24once right and um how many employees in
1:11:27HR do we need right now we have two uh
1:11:30if if they're highly levered like that
1:11:33maybe we do not need a 20 HR team on
1:11:37this point of
1:11:38meaning I've heard so many billionaires
1:11:41in AI describe this as the age of
1:11:42abundance and I'm not necessarily sure
1:11:44If abundance is always a great thing
1:11:47because you know when we look at mental
1:11:50health and we look at why how people
1:11:53derive their meaning and their purpose
1:11:54in life much of it is having something
1:11:56to strive towards and some struggle in a
1:11:58meaningful direction to you and this is
1:12:00maybe adjacent but when there was a
1:12:03study done I think it was in Australia
1:12:04where they looked at suicide letters and
1:12:06in the suicide letters the sentiment of
1:12:08men in those suicide letters was they
1:12:11didn't
1:12:12feel worthy they didn't feel like they
1:12:14were worth it. They didn't feel like
1:12:16they were needed by their families. And
1:12:19this is much of what caused their
1:12:20psychological state. And I wonder in a
1:12:23world of abundance where we, you know, a
1:12:25lot of these AI billionaires are telling
1:12:26us that we're going to have so much free
1:12:28time and we're not going to need to
1:12:30work. If there is at all going to be a
1:12:31crisis of meaning, a mental health
1:12:33problem. I mean, there already is. And
1:12:36it doesn't require AI and it's going to
1:12:37get worse. I don't know what to do about
1:12:39it because essentially as human beings
1:12:42we are built like all organisms to find
1:12:47opportunity and figure out how to
1:12:50exploit it. That's what we do. And the
1:12:52world you're describing is really the
1:12:53opposite of that. It's one where you're
1:12:55effectively having your biological needs
1:12:59at the physiological level satisfied and
1:13:02there isn't an obvious place for your
1:13:06spare time if that's what you end up
1:13:08with to be utilized in something that
1:13:12you know there's no place to strive and
1:13:14I do imagine almost at best what would
1:13:17happen is you have people who are being
1:13:21sustained by a universal basic income
1:13:24come and then parasetized
1:13:26uh you know whatever currency they have
1:13:29to spend somebody will be targeting it
1:13:30and they will be targeting it with a an
1:13:32AI augmented system that spots their
1:13:36defects of character. I mean, again,
1:13:38we're already living in this world, but
1:13:40it will be that much worse when the AI
1:13:42is figuring out, you know, what kind of
1:13:45porn to target you with specifically.
1:13:47That's uh it's a nightmare scenario. And
1:13:50I do think it would be worth our time as
1:13:54a species to start considering if we are
1:13:57about to find ourselves in this
1:13:58situation and we find some way of
1:14:00dealing with the basic needs of the
1:14:02large number of people who are going to
1:14:04be
1:14:04sidelined. What would a world have to
1:14:07look like in order for them to have real
1:14:09meaning? Not pseudo meaning, not
1:14:10something that you know superficially,
1:14:12you know, a video game is not meaning
1:14:14even if it feels very meaningful in the
1:14:16moment. I I think that would be a a
1:14:19worthy investment for us to figure out
1:14:20how to produce it. But frankly, I'm not
1:14:22expecting us to either have that
1:14:24conversation or get very far down that
1:14:27road. I think it's much more likely that
1:14:29we will squander the wealth dividend
1:14:32that will be produced by by AI.
1:14:34Interestingly, you also see in Western
1:14:36countries that when we get more
1:14:37abundance, we start having less kids.
1:14:40And we're already seeing this sort of
1:14:41population decline in the Western world,
1:14:43which is was kind of scary. I think it's
1:14:46often associated with affluence like the
1:14:48more money someone makes the less likely
1:14:49they are to want to want to have
1:14:50children the more they try and protect
1:14:52their freedoms. But also on this point
1:14:53of
1:14:54AI relationships are hard you know my
1:14:58girlfriend is happy sometimes and not
1:15:01happy other times and I have to like you
1:15:04know go through that struggle with her
1:15:05of like working on the relationship.
1:15:07Children are hard and if we are
1:15:10optimizing ourselves and you know much
1:15:12of the reason that I sustain the
1:15:13struggle with my girlfriend is I'm sure
1:15:15from some evolutionary reason because I
1:15:16want to reproduce and I want to have kin
1:15:19but if I didn't have to deal with the
1:15:23struggle that comes with human
1:15:24relationships romantic or platonic
1:15:26there's going to be a proportion of
1:15:28people that actually choose that outcome
1:15:29and I wonder what's going to happen to
1:15:30birth rates in such a scenario because
1:15:32we're already struggling. We're already
1:15:34in a situation where we used to be
1:15:35having five children per woman in the
1:15:401950s to about two in
1:15:452021. And we're seeing a decline. If you
1:15:47look at South Korea, their fertility
1:15:48rate has fallen to 72, the lowest
1:15:51recorded globally. And if this trend
1:15:53continues, the country's population
1:15:55could half by 20, 100.
1:16:00So yeah, relationships, connections, and
1:16:03and also I guess I guess we've got to
1:16:04overlay that with the loneliness
1:16:05epidemic, which
1:16:07is they promised us social connection
1:16:09when social media came about, when we
1:16:12got Wi-Fi connections, the promise was
1:16:13that we would become more connected. But
1:16:15it's so clear that because we spend so
1:16:17long alone, isolated, having our needs
1:16:19met by Uber Eats drivers and social
1:16:21media and Tik Tok and the internet, that
1:16:23we're investing less in the very
1:16:25difficult thing of like going and making
1:16:27a friend and like going and finding a
1:16:29girlfriend. Young people are having sex
1:16:30less than ever before. Everything that
1:16:33is associated with the difficult job of
1:16:36making in real life connection seems to
1:16:39be um falling away.
1:16:41I I will make the case that everything
1:16:43that we've discussed here, all the
1:16:46negative things around loneliness, um
1:16:48around meaning, they're already here.
1:16:52And I don't think blaming technology for
1:16:55all of it is is the right thing. Like I
1:16:57think there are a lot of things that
1:16:59happened because of existing human uh
1:17:04you know, impulses and and motivations.
1:17:07Um well I I wanted to go back to where
1:17:10you started because I do think that this
1:17:12maybe is the fundamental question. Why
1:17:15is it that we are already living in a
1:17:17world that is not making us happy? And
1:17:19is that the responsibility of
1:17:21technology? And I don't think it's
1:17:22exactly technology. Human beings uh
1:17:25among our gifts are fundamentally
1:17:27technological whether we're talking
1:17:28about quantum computing or
1:17:32flintnapping an arrow
1:17:35head. What has happened to us that has
1:17:38created the growing, spreading, morphing
1:17:42dystopia is a process that Heather and I
1:17:45in our book, A Hunter Gather's Guide to
1:17:47the 21st Century, call hyper
1:17:49novelty. Hypern novelty is the fact of
1:17:54the rate of change outpacing our
1:17:57capacity to adapt to change. And we are
1:18:00already well past the threshold here
1:18:02where the world that we are young in is
1:18:05not the world that we are adults in. And
1:18:07that mismatch is making us sick across
1:18:10multiple different domains. So the the
1:18:13question that I ask is is the change
1:18:16that you're talking about going to
1:18:18reduce the rate of change in which case
1:18:21we could build a world that would start
1:18:23meeting human needs better open
1:18:25opportunities for pursuing meaningful
1:18:27work. or is it going to accelerate the
1:18:30rate of change which is in my opinion
1:18:32guaranteed to make us worse off. So if
1:18:35it was a one-time shift, right, AI is
1:18:38going to dawn. It's going to open all
1:18:39sorts of new opportunities. There's
1:18:41going to be a tremendous amount of
1:18:42disruption, but from that we'll be able
1:18:44to build a world. Is that world going to
1:18:46be stable or is it going to be just, you
1:18:48know, one event horizon after the next?
1:18:51If it's the latter, then it effectively
1:18:54says what it does to the humans, which
1:18:56is it it's going to dismantle us. When I
1:19:00look out at society, I I go, okay, it's
1:19:02having a negative impact. When I look at
1:19:04um individual use cases, it's having a
1:19:07profoundly positive impact. Including
1:19:09for me, it's having a very positive
1:19:10impact. So, it's it's one of these
1:19:13things where I wonder what is that what
1:19:15is it that we need to teach people at
1:19:17school so that they understand the world
1:19:20that we're going into? Because one of
1:19:21the biggest issues that we're having is
1:19:24that we're sending kids to school with
1:19:26this blueprint, this template that
1:19:28they're going to have this long arc
1:19:30career that no longer exists that
1:19:33essentially we're treating them like
1:19:34learning LLMs. And we're saying, "Okay,
1:19:37we're going to prompt you. You're going
1:19:38to give us the right answer. you're
1:19:40going to hallucinate it if possible. And
1:19:42you know, and and then we go, "Okay, now
1:19:44go off into the world." And they go,
1:19:45"Oh, but wait a second. I don't know how
1:19:47money works. I don't know how society
1:19:48works. I don't know how my brain works.
1:19:50I don't know how I meant to handle this
1:19:52novelty problem. I'm not sure how to
1:19:54approach someone in a in a social
1:19:56situation and ask if they want to go on
1:19:57a date." Um so all the important things
1:20:00that actually are the important
1:20:01milestones that people want to be able
1:20:03to hit and that technology can actually
1:20:06have an impact on we get no user manual.
1:20:09So I think one of the biggest things
1:20:10that has to happen is we have to equip
1:20:14uh young people all through school that
1:20:17to actually prepare them for the world
1:20:18that's coming or the world that's here.
1:20:22Well on the one hand I think you you
1:20:25outline the problem very well.
1:20:27effectively we have a a model of what
1:20:29school is supposed to do that you know
1:20:30at best was sort of a match for the 50s
1:20:33or something like that and it woefully
1:20:36misses the mark with respect to
1:20:38preparing people for the world they
1:20:39actually
1:20:41face if we were going to prepare them I
1:20:44would argue that the only toolkit worth
1:20:46having at the moment is a highly general
1:20:50toolkit the capacity to think on your
1:20:53feet and pivot as things change is the
1:20:55only game in town with respect to our
1:20:57ability to prepare you in advance. Maybe
1:21:00the the other auxiliary component to
1:21:02that would be teaching you what we know
1:21:05which is frankly not enough about how to
1:21:08live a healthy life. Right? If we could
1:21:11if we could induce people into the kinds
1:21:14of habits of behavior and the
1:21:18consumption of food and then train them
1:21:20to think on their feet, they might have
1:21:22a chance in the world that's coming. But
1:21:24uh the fly in the ointment is we don't
1:21:28have the teachers to do it. We don't
1:21:29have people who know. And that is the
1:21:32question is could the AI actually be
1:21:34utilized in this manner to actually
1:21:38induce the right habits of mind for
1:21:40people to live in that world. I I I
1:21:42spent a lot of time in education
1:21:44technology. One thing that is as we say
1:21:47on the internet a black pill about
1:21:49education in general, education
1:21:51intervention is there's a lot of data
1:21:54that shows that there are very little
1:21:57interventions you can make in education
1:21:59to generate better outcomes. Um and so
1:22:03you know uh there's been a lot of
1:22:05experiment around pedagogy around you
1:22:07know how to configure the the the
1:22:08classroom that have resulted in very
1:22:11marginal improvements. There's only one
1:22:14intervention and this this has been uh
1:22:16reproduced many times that creates two
1:22:19sigma two standard deviation
1:22:22uh positive outcomes in education
1:22:24meaning you're better than 99% of uh of
1:22:27of everyone else and that is one-on-one
1:22:31tutoring I thought so I was going to say
1:22:33smaller classrooms and personalization
1:22:34one-on-one tutoring yeah and and but by
1:22:36the way if you look someone also did a
1:22:38survey of all the geniuses the
1:22:40understanders of the world and found
1:22:42that they all had one-on-one tutoring.
1:22:44They all had someone in their lives that
1:22:46took interest in them and tutored them.
1:22:48So, what can create one-on-one tutoring
1:22:51opportunity for every child in the
1:22:53world? AI. AI. My kids use it and it's
1:22:56incredible. Yeah. As in like um they're
1:22:59interacting and and it's adapting to
1:23:01their speed. Yes. And um it's giving
1:23:04them different analogies to work with.
1:23:06So, like, you know, my son was learning
1:23:08about division and it's asking him to
1:23:10smash glass and how many pieces he
1:23:13smashes it into with this hammer and,
1:23:15you know, and it's saying things like,
1:23:16"No, Xander, go for it. Really smash
1:23:18it." And um and he's loving it, right?
1:23:21Is that synthesis? Yeah. Yeah. I'm an
1:23:23investor in this company. Oh, well, it
1:23:24was it was it's great to watch that
1:23:27simulated one-on-one tutoring because
1:23:29it's talking to him. It's asking him
1:23:31questions. Brett, you're an educator.
1:23:33you uh spent much of your life teaching
1:23:35people in universities. How do you
1:23:38receive all of this? Well, on the one
1:23:40hand, I agree that the uh the closer to
1:23:43one to one you get, the better. But I
1:23:46also personally believe that 0ero to one
1:23:50is best. And what I mean by that
1:23:54is part of what's gone wrong with our
1:23:56educational system is that it is done
1:24:01through abstraction.
1:24:03And effectively the arbiter of whether
1:24:07you have succeeded or failed in learning
1:24:09the lesson is the person at the front of
1:24:11the room. And that's okay if the person
1:24:14at the front of the room is truly
1:24:15insightful. And it's terrible if the
1:24:18person at the front of the room is
1:24:19lackluster, which happens a lot. So what
1:24:23doesn't work that way is interaction
1:24:26with the physical world in which nobody
1:24:28has to tell you whether you've succeeded
1:24:29or failed. If you're faced with an
1:24:31engine that doesn't start, you can't
1:24:34argue it into starting. You have to
1:24:36figure out what the thing is that has
1:24:37caused it to fail, and then there's a
1:24:40great reward when you alter that thing
1:24:42and suddenly it fires up. So, I'm a big
1:24:45fan of being as light-handed as possible
1:24:49and as concrete as possible in teaching.
1:24:51In other words, uh, when I've done it,
1:24:53and not just with students, but with my
1:24:55own children, I like to say as little as
1:24:58possible, and I like to let physical
1:25:00systems tell the person when they've
1:25:03succeeded or failed. And that creates an
1:25:06understanding. You can extrapolate from
1:25:08one system to the next. And you know
1:25:10that you're not just extrapolating from
1:25:11one person's misunderstanding. You're
1:25:13extrapolating from the way things
1:25:15actually work. So, I don't know if AI
1:25:18can be leveraged in that context. My
1:25:21sense is there's probably a way to do
1:25:22it, but one would have to be deliberate
1:25:24about it, especially with robotics and
1:25:26humanoid robots. Actually, that is that
1:25:28is the place uh where where you can do
1:25:31this is with robotics that
1:25:34um it seems to me. Yeah. Well, robotics
1:25:38will teach you the physical computing
1:25:40part of it. And then the question is how
1:25:42do you infuse this with AI so that um it
1:25:45is that it you know it provokes you out
1:25:49of some eddy where you're caught and
1:25:51moves you into the ability to solve some
1:25:53next level problem uh that you you
1:25:55wouldn't have found on your own. What do
1:25:57what do you think should be taught in
1:25:58the classroom with everything that you
1:26:00now know? Well, you're all fathers here.
1:26:03You all have your own children. So, it's
1:26:05a good question for you. How old are
1:26:08your kids? How old are your kids? Uh
1:26:09three and five. 19 and 21 and six,
1:26:13seven, and 10. My children are very
1:26:15young, but uh we already do use AI and I
1:26:18sit down with them in front of replet
1:26:19and we generate ideas and make make
1:26:21games. And um I would say, you know,
1:26:23what Brett said about generality is very
1:26:25important. The ability to pivot and kind
1:26:27of learn skills quickly. Being
1:26:30generative is very very important.
1:26:33Having a you know a fast pace of
1:26:35generating ideas and iterating on those
1:26:37ideas. We sit down in front of Chad GPT
1:26:41and my kid imagines scenario. Oh, what
1:26:43if you know there's a there's a cat on
1:26:45the moon and then you know what if the
1:26:47moon is made of cheese and what if
1:26:49there's a mouse inside it or and so we
1:26:51keep generating these um variations of
1:26:55these different ideas and I and I find
1:26:57that you know makes them more
1:26:58imaginative and and creative. Uh rule
1:27:01number one that I tell my kids is stay
1:27:04away from porn at all costs. I'd rather
1:27:07you have a drug problem than a porn
1:27:08problem. And I actually mean that. I
1:27:10think it's I think porn is more
1:27:11dangerous to the to the human being as
1:27:13as bad as a drug problem is. But when we
1:27:16get to the question of how to confront
1:27:18the world and uh the things that you're
1:27:20going to be um expected to to do in the
1:27:23workplace and all of that, my point to
1:27:26them is you are facing the uh the
1:27:30dawning of the age of complex systems
1:27:36that you are going to have to interact
1:27:37with. And in the age of complex systems,
1:27:39you have to understand that you cannot
1:27:42blueprint a solution. And you have to
1:27:44approach these systems with a upgraded
1:27:49toolkit of humility because the ability
1:27:52of the system to do something you don't
1:27:54predict is much greater than a highly
1:27:55complicated system. So you have to
1:27:58anticipate that and be very sensitive to
1:27:59the fact that what you intended to
1:28:02happen is not what's going to happen. So
1:28:04you have to monitor the unintended
1:28:06consequences of whatever your action is
1:28:08and that there are really two tools
1:28:10which work. One of which you just
1:28:12mentioned which is the prototyping. You
1:28:14prototype things. You don't imagine that
1:28:16I know the solution to this and I'm
1:28:18going to build it. You imagine I think
1:28:19there's a solution down there. I'm going
1:28:21to make a proof of concept and then I'm
1:28:23going to discover what I don't know and
1:28:25I'm going to make the next version.
1:28:26Discover what I don't know and
1:28:27eventually you may get to something that
1:28:29actually truly accomplishes the goal. So
1:28:32prototyping is one thing.
1:28:34And also instead of using the blueprint
1:28:36as the metaphor in your mind uh navigate
1:28:40you can navigate somewhere. And you know
1:28:42that the way I think of it is a surfer
1:28:47is in some ways mastering a complex
1:28:50system but they're not doing it by
1:28:52planning their days surfing down the
1:28:54waves. You can't do that. What you can
1:28:56do is you can be expert at absorbing
1:28:59feedback and navigating your way down
1:29:01the wave. and that that's the right
1:29:03approach for a complex system. Nothing
1:29:04else is going to work. And so I guess
1:29:07the final piece is uh general tools
1:29:10always no specialization. This is this
1:29:13is the age of generalists and um invest
1:29:16in those tools and they will pay.
1:29:19So the guiding philosophy for me is uh
1:29:21to produce high agency generalists. So
1:29:24um ultimately I want them to be
1:29:26motivated self-starters and have a wide
1:29:29general toolkit. I imagine them very
1:29:31much what you imagine which is
1:29:33instructing robots, instructing agents,
1:29:36coming up with ideas. Um, and I imagine
1:29:39them having a very high velocity life
1:29:41where they may be writing a book,
1:29:43organizing a festival, having a podcast,
1:29:46starting a business, and being part of
1:29:47somebody else's business all at once as
1:29:49they are of the ADHD. Yeah. Right.
1:29:51Exactly. Um, so the high agency
1:29:54generalist is the kind of guiding
1:29:56philosophy. Some of the things that we
1:29:58do is like we do chess, we do Brazilian
1:30:00jiu-jitsu, we do dancing, we do acting
1:30:03classes, playing in nature, uh
1:30:05entrepreneurship, understanding that you
1:30:07can start a lemonade. We just did
1:30:08lemonade stands which was amazing. U we
1:30:11sold lots of lemonade on the street. So
1:30:13those kind of things and jumping from
1:30:15one thing to the next thing, but also
1:30:17trying to avoid too many screens and
1:30:19forcing them into making stuff from
1:30:22what's going on around the house. Um,
1:30:25some distinctions that we try and give
1:30:26them is the difference between creating
1:30:28and consuming because I think AI has
1:30:30this superpower of making you a hyper
1:30:32consumer or a hyper creator. Um, and if
1:30:34you don't understand the distinction
1:30:36between creation and consumption, you
1:30:38end up falling into the consumption
1:30:39trap, whether it be porn or just news or
1:30:44um, thing, you know, things that feel
1:30:45like you're productive, but you're
1:30:47actually just consuming stuff. Won't
1:30:48that be the most successful AI? the one
1:30:51that plays with my dopamine the most.
1:30:54Yeah. And and makes you and makes you
1:30:56think that you're achieving something
1:30:58when you're actually just consuming
1:31:00something. So trying to give them the
1:31:02understanding that there is this
1:31:04difference in their life between
1:31:05creation and consumption and to be on
1:31:07the creation side. I started my first
1:31:09business at 12 years old and I started
1:31:11more businesses at 14, 15, 16, 17 and
1:31:1418. And at that time, what I didn't
1:31:17realize is that being a founder with no
1:31:19money meant that I also had to be the
1:31:21marketeteer, the sales rep, the finance
1:31:23team, customer service, and the
1:31:25recruiter. But if you're starting a
1:31:27business today, thankfully, there's a
1:31:29tool that wears all of those hats for
1:31:31you. Our sponsor today, which is
1:31:33Shopify. Because of all of its AI
1:31:35integrations, using Shopify feels a bit
1:31:37like you've hired an entire growth team
1:31:40from day one, taking care of writing
1:31:42product descriptions, your website
1:31:44design, and enhancing your products
1:31:46images, not to mention the bits you'd
1:31:48expect Shopify to handle, like the
1:31:50shipping, like the taxes, like the
1:31:51inventory. And if you're looking to get
1:31:53your business started, go to
1:31:56shopify.com/bartlet and sign up for a $1
1:31:59per month trial. That's
1:32:01shopify.com/bartlet.
1:32:04The thing that we I think all agree on
1:32:07is that this is inevitable. Do you agree
1:32:09with that, Brett? I think it's sad that
1:32:12it is inevitable, but at this point it
1:32:14is. What part of it do you find sad?
1:32:18We have squandered a long period of
1:32:24productivity and peace in which we could
1:32:27have prepared for this moment. and our
1:32:31narrow focus on competition
1:32:36has created
1:32:38a a fragile world that I'm afraid is not
1:32:42going to survive the disruption that's
1:32:43coming. And it didn't have to be that
1:32:45way. This was foreseeable. I mean,
1:32:47frankly, the movie 2001, which came out
1:32:51the year before I was born, anticipates
1:32:54some of these problems. And you know we
1:32:58treated it too much like education. I
1:33:00mean like entertainment and not enough
1:33:03like education. So we are now you know
1:33:07we've had the AI era opened without a
1:33:11discussion about its implications for
1:33:12humanity. There is now for game
1:33:14theoretic reasons no way to slow that
1:33:18pace because as you point out if we
1:33:20restrain ourselves we simply put the AI
1:33:23in the hands of our competitors. That's
1:33:24not a solution. So, I don't advocate it,
1:33:27but there's a lot more preparation we
1:33:29could have done. We could have
1:33:30recognized that there were a lot of
1:33:31people in jobs that were uh about to be
1:33:34obliterated and we could have thought
1:33:36deeply about what the moral implications
1:33:39were and what the solutions at our
1:33:42disposal might have been. And having not
1:33:45prepared, it's going to be a lot more
1:33:46carnage than it needed to be. Amjad, I
1:33:49heard you say a second ago that what we
1:33:50should be talking about is how we deal
1:33:51with job displacement. Do you have any
1:33:54theories if you were prime prime
1:33:56minister or president of the world and
1:34:00you your job was to deal with job
1:34:01displacement let's just say in the
1:34:03United States how would you go about
1:34:05that the first thing I would do is uh
1:34:09teach people about these systems whether
1:34:11it's um programs on on the TV or
1:34:16outreach or or what have you just trying
1:34:18to get people to understand how chat GBT
1:34:21works how these algorith algorithms work
1:34:24and as the new jobs arrive um I think
1:34:28you know there's going to be an
1:34:29opportunity for people to be able to
1:34:31detect that you know the this job
1:34:34requires this set of skills and I I I
1:34:37have this this kind of experience and
1:34:39although my experience are potentially
1:34:41outdated I can repurpose that experience
1:34:43to do that job I'll give you an example
1:34:46a teacher his name is Adil Khan you know
1:34:49he started using at the time GPT3 and uh
1:34:53felt like it does amazing work as a
1:34:55tools for teachers or even potentially a
1:34:58teacher itself. So he learned a little
1:35:00bit of coding and he he went to
1:35:01Unreallet and he built uh this company
1:35:04and um just two years later they're
1:35:07worth hundreds of millions of dollars.
1:35:08Obviously, not everyone will be able to
1:35:10create businesses of that scale, but
1:35:13because you have an experience in a
1:35:15certain domain, you'll be able to build
1:35:18the next iteration of of of that using
1:35:22technology. So, even if your job was
1:35:23displaced, you'll be able to figure out,
1:35:27you know, what's what's potentially what
1:35:30potentially comes after that. So, so I I
1:35:32I think people's expertise that they
1:35:35built, I don't think they're all for
1:35:37waste. Even if your job went away,
1:35:41you can never really predict what jobs
1:35:42are coming. I mean, I think of this
1:35:44crazy situation where I tell my
1:35:47grandfather, what is a personal fitness
1:35:50trainer? And he would his mind would be
1:35:53blown by this idea that well, okay, I
1:35:56don't really want to go to the gym, so I
1:35:58have to make an appointment and pay
1:35:59someone to go to the gym and meet with
1:36:01me there. And then he stands there and
1:36:02tells me to lift heavy things that I
1:36:04don't really want to lift. And then he
1:36:06counts them and tells me that I've done
1:36:08a good job and then I put the heavy
1:36:10things down and then at the end of that
1:36:12I feel really good and I pay him a bunch
1:36:14of money. My grandfather would be like
1:36:16what on earth have you been scammed? Is
1:36:18this so we can never predict what what
1:36:21this uh future of jobs would look like.
1:36:23Even just 20 30 40 years apart the jobs
1:36:26rapidly and convincingly just morph into
1:36:29something else. I think it's very
1:36:31dangerous the idea that we need to focus
1:36:34on skills. I think the future is not in
1:36:36skills. Skills are being replaced. It's
1:36:38this idea that the education system has
1:36:40to stop being compartmentalized and has
1:36:42to be a lifelong learning approach. The
1:36:44department of education needs to be
1:36:47seeing people as lifelong learners who
1:36:49are constantly disrupted and need
1:36:51re-education. Interesting. That that's
1:36:53going to be a thing. The Department of
1:36:54Education needs to start as a kid and go
1:36:58right through to maybe 70. Does the
1:37:00Department of Education have a role
1:37:01anymore at all? Depends on your
1:37:03definition of education. I think if
1:37:04you're trying to teach kids or if you're
1:37:06trying to teach kids to, you know,
1:37:09remember facts and figures from a
1:37:11history book, then no. But if it's about
1:37:14coaching, mentoring, being displaced,
1:37:16finding the next thing, and maybe if
1:37:18it's AIdriven and all of those kind of
1:37:20things, then it's a different paradigm
1:37:22shift around what education is and what
1:37:23its purpose is. And if we see it as a
1:37:26fluid thing where we wave into an
1:37:28opportunity and then wave back into
1:37:30education, spotting a new opportunity
1:37:32and then back here. If we're learning
1:37:35rather than skills, but we're learning
1:37:36tools. So it's a tools-based education
1:37:39as opposed to a skills-based education.
1:37:41The the purpose of education for most of
1:37:43human history was about virtue, about
1:37:45becoming a great person who had good
1:37:47judgment and who had good values. And we
1:37:49don't really do much of that anymore.
1:37:51But I think if we essentially said if we
1:37:53get back to what is if we ask the
1:37:56question what is the purpose of
1:37:57education and where does it fit in our
1:37:58lives and at what time frame does it go
1:38:01for and then we just trust that people
1:38:04are going to come up with weird and
1:38:06wonderful jobs. You know this is sounds
1:38:09crazy but also and this is a weird
1:38:13analogy. My cat is incredibly happy. How
1:38:17do you know? Well, it it demonstrates
1:38:20all the characteristics of of being a
1:38:22happy cat and it lives in a world of
1:38:25super intelligence as far as it's
1:38:27concerned. So, there's this house and
1:38:29food just magically happens. It has no
1:38:31idea that there's this Google calendar
1:38:33that runs a lot of things that happen
1:38:35around it. The food gets delivered. The
1:38:38money is magically made by something
1:38:40that is inconceivably more intelligent
1:38:42than the cat. And yet the cat has
1:38:44evolved to be living this life of
1:38:46purpose and meaning inside the house.
1:38:48And as far as it's aware, it it's got a
1:38:50great life. But you have the power at
1:38:53any moment if you're having a bad day to
1:38:55do something not so pleasant to that
1:38:57cat. And it can't really reciprocate
1:38:59that. Exactly. But but what's in it for
1:39:02me to hurt the cat?
1:39:04Because the in this analogy, you might
1:39:08want to move house and the landlord
1:39:09doesn't allow cats. So you've got a
1:39:11decision to make. Yeah, that there are
1:39:13things that the cat is highly disrupted
1:39:15by due to no fault of the cat. I get it.
1:39:17But as far as cat existence goes and the
1:39:20and the history of cats, if you were to
1:39:23ask that cat, do you want to tra trade
1:39:25places with any of the other cats that
1:39:27came before you? It would probably say,
1:39:28I don't want to take the risk because
1:39:30all the other cats had to fend for
1:39:32themselves in a way that I don't have
1:39:33to. It's very possible that we live end
1:39:35up living in a life a lot like the house
1:39:38cat in the sense that from our
1:39:41perspective we're extremely like we're
1:39:44having very interesting lives and
1:39:46purpose meaning and just there's this
1:39:48massive higher intelligence that's just
1:39:50running stuff and we don't know how it
1:39:53works but it doesn't really matter
1:39:54whether how it works. We we we are the
1:39:57beneficiaries of it and it it's doing
1:40:00important things and we're enjoying
1:40:01being house cats in in its in its life.
1:40:03I have a few things to say about this.
1:40:04One, I'm pretty sure your cat's not as
1:40:06impressed with your capacity as you are
1:40:10or as you think he is. Um I just know
1:40:12cats well enough to be pretty sure of
1:40:14that. But oh, it looks down on me. Yeah,
1:40:15you're right. I think it I think it's a
1:40:17fair it's a fair point that there is an
1:40:19existence and actually, you know, pets
1:40:21really do have it. If they have loving
1:40:22owners, they really do have it pretty
1:40:24great. And I would also point out that
1:40:25there's a way in which we already are
1:40:27this way. Most of us do not understand
1:40:30the process that results in electricity
1:40:33coming out of the walls of our house or
1:40:35the water that comes out of the tap. And
1:40:37we're pretty much okay with the fact
1:40:39that somebody takes care of that and we
1:40:41can busy ourselves with whatever it
1:40:42might be. But the place that I find
1:40:45something troubling in your description
1:40:48is that you say that the nature of what
1:40:51we do is to deal with the fact that jobs
1:40:54are always being upended. That's a very
1:40:57new process. That is the hypern novelty
1:41:00process. It used to be that it was only
1:41:03very rarely that a population had a
1:41:06circumstance where you didn't
1:41:07effectively do exactly what your
1:41:09immediate ancestors did. Right? Um, in
1:41:12general, you took what the jobs were,
1:41:15you picked something that was suited to
1:41:18you, and you did that thing
1:41:20intergenerationally.
1:41:21Intergenerationally. And the point is,
1:41:23we've now gotten to the point where even
1:41:26within your lifetime, what is possible
1:41:29to get paid for is going to shift
1:41:31radically in ways that nobody can
1:41:33predict. And that is a dangerous
1:41:35situation. Like probably every two
1:41:37years, like two or three years, right?
1:41:39And so maybe there's some model by which
1:41:41we can surf that wave and you can learn
1:41:43a generalist toolkit and you know that
1:41:46your survival doesn't depend on your
1:41:49being able to you know switch up every
1:41:51two years and never miss a beat or maybe
1:41:54we can't but I do think it is worth
1:41:56asking the question if the rate of
1:42:00technological change has taken us out of
1:42:03the normal human circumstance of being
1:42:07able to deduce what you might do for a
1:42:09living based on what your ancestors did
1:42:11and put us in a situation where what
1:42:13your ancestors did is going to be
1:42:15perfectly irrelevant no matter what. But
1:42:17that is effectively a choice that has
1:42:19been made for us. And we could choose to
1:42:22slow the rate of change so that we would
1:42:26live in some kind of harmony where our
1:42:29developmental environment and our adult
1:42:32environment were a match. Now, as a
1:42:33biologist, I would argue if we don't do
1:42:35something like that, this is a matter of
1:42:38time. Yeah. How would we change? How do
1:42:40we slow the rate of change? Well, I I
1:42:42mean, you can you can you can be the
1:42:43Amish, right? You can be the Amish and
1:42:45live in your own communities and and I
1:42:47would assume some people would would
1:42:49want that. Well, I'm you know, when
1:42:52Heather and I wrote our book, I wanted
1:42:54the first chapter to be, are the Amish
1:42:56right? And the answer is they can't be
1:42:59exactly right because they picked an
1:43:01arbitrary moment to step off the
1:43:03escalator. But are they right that
1:43:04there's something dangerous about this
1:43:06continuing pattern of technological
1:43:07change? Clearly they are. What do the
1:43:09Amish do for anyone that doesn't know?
1:43:11The Amish live as if it was what 1850 or
1:43:18something. So they live in a they don't
1:43:21use cars. They I think they do have
1:43:23phones but they do not have electricity.
1:43:27Basically they they voluntarily accept a
1:43:30techn they're basically a lite community
1:43:33and they uh have turned out to fare
1:43:37surprisingly well against many of the
1:43:39things that have upended modern one of
1:43:42them right yeah co they did beautifully
1:43:44quite happy people very low autism rates
1:43:46they they have all sorts of advantages
1:43:48so anyway I'm not arguing that we should
1:43:50live like the Amish I don't see that but
1:43:51I do think the idea that they had an
1:43:53insight which was you need to step off
1:43:56that escalator because you're just going
1:43:57to keep making yourselves sicker is
1:43:59probably right now. Maybe this is a
1:44:02one-time shift. We've stepped over the
1:44:05event horizon. We are going to be living
1:44:07in the AI world. And maybe if we're
1:44:09careful about it, we can figure out how
1:44:11to turn that landscape of infinite
1:44:14possibility that you're describing into
1:44:17a place that doesn't change. That you
1:44:20always have the opportunity to decide
1:44:23what needs to be done. But that none
1:44:26that the that living over that event
1:44:28horizon is not an everchanging process.
1:44:31It's just the next frontier. I do want
1:44:33to also propose or ask the question when
1:44:36we talk about our hyperchanging world.
1:44:38Isn't it harder for older people to
1:44:41learn because of the the way that the
1:44:43brain works in terms of processing speed
1:44:44and memory flexibility?
1:44:47So I was wondering if you're going to
1:44:48get a situation where like my father h
1:44:50because of his brain and the reduced
1:44:52memory flexibility and processing speed
1:44:54that happens when you're older is going
1:44:55to struggle significantly more than my
1:44:57niece who can seem to learn I mean my
1:45:00niece knows five languages and she's
1:45:01seven or something crazy like that five
1:45:04languages but I mean the brain is much
1:45:05more plastic isn't it? So the it's and
1:45:07that's goes back to our evolutionary
1:45:09psychology which you know much
1:45:10evolutionary history which you know much
1:45:11more than I do about of we're meant to
1:45:14learn our lessons when we're young. use
1:45:16that information for a lifetime. But if
1:45:18that information is changing quickly,
1:45:20well, that's I mean this is exactly what
1:45:21I'm pointing to. It is not normal for
1:45:24your developmental environment to fail
1:45:27to prepare you for your adult
1:45:29environment. The normal thing is as a
1:45:31young person, you take on ever more of
1:45:34the responsibilities of the adult
1:45:37environment. And then at some point, you
1:45:40know, in a properly functioning culture,
1:45:42there's a right of passage. You go into
1:45:44the bush for 10 days, you come back
1:45:46with, you know, a large uh, you know,
1:45:49game animal and now you're an adult and
1:45:52you take that program that you've been
1:45:53building and you activate it. And that
1:45:55is normal. And, you know, you're a lot
1:45:59happier person. You're a lot more
1:46:00fulfilled if your life has that kind of
1:46:02continuity to it. And you know, I'm not
1:46:05against the idea that we have enabled
1:46:07ourselves to do things that can't be
1:46:09done if that's the the limit, but we
1:46:13have also harmed ourselves gravely. And
1:46:16I would like to somehow pry apart our
1:46:21ability to improve our well-being from
1:46:26our self-inflicted wounds that come from
1:46:29this
1:46:30neverending pace of change. And I don't
1:46:32know if it's possible, but I think it's
1:46:34a worthy goal. Something amusing. I
1:46:36don't know if it's exactly a
1:46:37counterpoint, but um during co
1:46:41especially and you know through the the
1:46:44recent technological change uh some
1:46:47people have started living closer to the
1:46:50more ancestral environment. Um so uh
1:46:54people whose jobs are online, some of my
1:46:58friends like went and built communities
1:47:00like collectives where they you know
1:47:02live and they they create farms and they
1:47:05they eat and then they eat and they have
1:47:07like an email job. They do their email
1:47:09jobs for five hours and and go out and
1:47:11they all have children and it's it's a
1:47:14fascinating life. And there was so much
1:47:16rethinking in in Silicon Valley about
1:47:18how we live. And there's a bunch of
1:47:19startups that are trying to create um
1:47:21cities where they're like, okay, we know
1:47:25that we've we're suffering because our
1:47:28cities are not really walkable. And
1:47:31there's so many reasons why we're
1:47:32suffering. First, we're not getting the
1:47:34movement. Second, there's a social
1:47:36aspect of walkable city where you're
1:47:38able to interact with people. You'll
1:47:40make uh friends by just happening to be
1:47:42in the same place as others. let's
1:47:44actually build uh walkable cities and if
1:47:47we want to you know uh transport faster
1:47:50we'll have these self-driving cars on
1:47:52the perimeter of the city that are going
1:47:53around and I think there are ways in
1:47:56which technology can afford us to uh to
1:47:59live uh in a way that
1:48:02reverses I I guess in a more local way.
1:48:06I I like that vision, but I also am
1:48:08aware that there's a different vision,
1:48:10right? You see people in Palo Alto, for
1:48:12example, actually exerting, you know,
1:48:15very strong controls on how much their
1:48:17children are exposed to, uh, you know,
1:48:20to phones. And I live in Palo Alto.
1:48:22Yeah. So, so you see that. On the other
1:48:24hand, what I am worried about is that
1:48:28the the elites of PaloAlto don't realize
1:48:32that what they're doing is they're
1:48:34figuring out how to reduce the harm to
1:48:36their own families as they're exporting
1:48:38the harm to the world of these
1:48:40technologies that for everybody else are
1:48:42unregulated. And so the question is, can
1:48:45we bring everybody along? If the AI
1:48:47revolution is going to alter our
1:48:49relationship to work and everything
1:48:52else, can we bring everybody along so
1:48:54that at the end of this process instead
1:48:56of saying well you know it's a shame
1:48:58that uh you know three billion people
1:49:01were sacrificed to this transition but
1:49:03progress is progress we can really say
1:49:05well we figured it out and everybody now
1:49:07is living in a style that is closer to
1:49:10their programming and closer to the
1:49:12expectations of their physical bodies.
1:49:15You know, if that were true, then I
1:49:16would I would be I would love to be
1:49:18wrong in my fears about what's coming.
1:49:21Um, but unfortunately, the market is not
1:49:24going to solve this problem without our
1:49:25being deliberate about forcing it to.
1:49:30What's your biggest fear? Like when you
1:49:32say my fears about what's coming, what
1:49:33do you like what's what's the picture
1:49:35that comes in your mind? Oh, it's a
1:49:36whole different topic actually. Um my my
1:49:39fear coming stemming from technology
1:49:43uh and AI is that this is a runaway
1:49:48process and that that runaway process is
1:49:50going to interface very badly with some
1:49:53latent human programs. that in effect
1:49:56the need for
1:49:58workers largely disappears and the
1:50:01people who are at the head of the
1:50:03processes that result in that
1:50:05elimination for the need for workers
1:50:07start talking about useless eaters.
1:50:08Maybe they come up with a new term this
1:50:10time. Thin the herd. Yep. Or they allow
1:50:12it to be thinned or something. Right.
1:50:14I've heard you talk about the five key
1:50:17concerns you have or the five key
1:50:19threats you have before. Could you name
1:50:21those five? So the first one is the one
1:50:23I worry least about. I don't worry zero
1:50:25about it, but I worry least about it,
1:50:27which is the malevolent AI uh that the
1:50:29doomers are so focused on. The second
1:50:33one is the idea that you know an AI can
1:50:36be misaligned not because it has
1:50:38divergent interest but because it just
1:50:40misunderstands what you've asked it.
1:50:41these autonomous agents. You know, the
1:50:43famous example is you ask them to
1:50:45produce as many paper clips as possible
1:50:47and they start liquidating the universe
1:50:48to make paper clips and you know, it's a
1:50:50it's a sorcerer apprentice kind of
1:50:52issue. The third one I would say
1:50:56is actually all of the remainder of them
1:50:58I would say are guaranteed and
1:51:02um the third of them is the derangement
1:51:07of human intellect that we are already
1:51:11living in a world where it's very
1:51:12difficult to know what the facts even
1:51:16mean. Right? We the facts are so
1:51:18filtered and we are so persuaded by
1:51:21algorithms that it's you know our
1:51:24ability to be confident even in the
1:51:25basic facts even within our own
1:51:27discipline sometimes is uh at an
1:51:29all-time low and it's getting worse and
1:51:32that problem takes a giant leap forward
1:51:35at the point that you have the ability
1:51:39to generate undetectable deep
1:51:42fakes. Right? that's going to alter the
1:51:45world very radically when the fact that
1:51:47you're looking at videotape of somebody
1:51:49robbing a bank doesn't mean that they
1:51:51robbed a bank or that a bank was even
1:51:53robbed. Um, so anyway, I call this we
1:51:56deal with this a lot by the way. I I
1:51:57think every single week, every single
1:52:00week I send my I have a chat people that
1:52:02just are now basically spending I'd say
1:52:0430% of their time dealing with deep
1:52:06fakes of me doing crypto scams, inviting
1:52:09people to Telegram groups, and then
1:52:11asking them for credit card details. We
1:52:13had one on X. I think you probably saw
1:52:15it, Dan, didn't you of me? But that
1:52:17someone was running ad deep fake ads on
1:52:18X of me. And it wasn't just one ad. It
1:52:21was like there were it was like swatting
1:52:22flies. There was 10 of them. And I
1:52:24messaged them to X and there was 10
1:52:26more. Then the day after there was 10
1:52:27more. Then the day after there was 10
1:52:29more. Then it started happening on on
1:52:30Meta. So it's a video of me basically
1:52:33asking you to come to a Telegram group
1:52:34where people are being scammed and
1:52:36audience members of mine are being
1:52:37scammed. And when I send them to Meta,
1:52:39they thankfully remove them. But then
1:52:41there's five more. And I went on
1:52:42LinkedIn yesterday and my DMs are,
1:52:44"Steve, by the way, there's this new
1:52:45scam." And I actually at this point I c
1:52:47I'd need someone fulltime just sending
1:52:50this over to Meta. I'm the I'm the same
1:52:52but on a smaller scale. Every week it's
1:52:55did you really message me on Facebook
1:52:57asking me for my crypto wallet and blah
1:52:59blah blah. My my least favorite ones are
1:53:01when the the single mother messages me
1:53:03saying that she just paid £500 of her
1:53:05money and how devastated she is and I
1:53:07feel this moral obligation to give her
1:53:09her money back um because she's fallen
1:53:11for some kind of scam. That was me. It
1:53:13was my voice. It was a video of me
1:53:14telling her something. Yeah. And I don't
1:53:16know I don't know how you deal with that
1:53:17but sorry do continue. Well, I mean
1:53:19that's actually on the list here. the
1:53:21massive disruption to the way things
1:53:22function both because people are going
1:53:24to be unemployed in huge numbers and
1:53:27because those who are not abiding by our
1:53:30social contract are going to find
1:53:32themselves empowered more than the
1:53:34people who do. So in this case, not only
1:53:38is this poor woman, you know, now out
1:53:40500 bucks for whatever the scam was, but
1:53:44you've also been robbed whether or not
1:53:46you pay her back for the thing that she
1:53:48thought she purchased. your credibility
1:53:51is being stolen by somebody and you have
1:53:53no capacity to prevent it. This has
1:53:55happened to me also and it is profoundly
1:53:58disturbing and it is only one of a dozen
1:54:00different ways that AI enables those who
1:54:06are absolutely willing to shrink the pie
1:54:11from which we all derive in order to
1:54:14enlarge their slice. you know, there
1:54:16there are innumerable ways that this can
1:54:18happen and um I think people do not see
1:54:21it coming. They don't understand how
1:54:22many different ways they are going to be
1:54:26robbed every bit as surely as if
1:54:27somebody was printing money. Um and then
1:54:30the last one is that this just simply
1:54:33accelerates demographic
1:54:36uh processes that do potentially result
1:54:39in the unleashing of technologies that
1:54:43pre-existed AI. you know, this this can
1:54:45easily result in an escalation
1:54:48uh into wars that turn nuclear. Um, so
1:54:52anyway, I think that list could probably
1:54:55be augmented at this point now that
1:54:57we've, you know, spent a little time in
1:54:59the AI era. We can begin to put a little
1:55:01more flesh on the bones both of what is
1:55:04possible in this era and what we should
1:55:06fear. One of those you you mentioned uh
1:55:08truth, you know, the problem of truth.
1:55:11Would you say just a thought experiment,
1:55:14someone today like an average person,
1:55:17college educated say
1:55:18person, are they more
1:55:21propagandized or led astray
1:55:26than someone in Soviet Russia?
1:55:30Well, I don't know because I didn't live
1:55:32in Soviet Russia, but my understanding
1:55:34from people who did was that there was a
1:55:39wide awareness that the propaganda
1:55:42wasn't true. Doesn't mean they knew what
1:55:43to believe, but there was a cynicism,
1:55:46which is one of my fears here, is that
1:55:48the, you know, you're really stuck
1:55:50choosing between two bad options in a
1:55:54world where you can't tell what is true.
1:55:55You can either be overly credulous and
1:55:57be a sucker all the time, or you can
1:55:59become a cynic and you can be paralyzed
1:56:01by the fact that you just don't believe
1:56:02anything. But neither of those is a
1:56:05recipe for do you think Google search
1:56:06first and maybe now chat GPT has helped
1:56:09people more or
1:56:11less to find truth? I
1:56:15think it's not chat GPT exactly but all
1:56:18of the various AI engines that we're
1:56:20starting with Google have briefly
1:56:22enhanced our capacity to know what's
1:56:24true because in fact they allow us to
1:56:26see through the algorithmic manipulation
1:56:29because the AI is not well policed. you
1:56:33can get it to recognize patterns that
1:56:35people will swear are not true. Um, and
1:56:38so anyway, a lot of us have found it
1:56:40useful in just simply unhooking the
1:56:42gaslighting. Um, so that's been very
1:56:44positive. But I also remember the early
1:56:47days of search and search used to be a
1:56:50matter of there are some pages out
1:56:53there. I don't know where they are.
1:56:54Here's a mechanized something that's
1:56:56looked through this stuff and just point
1:56:58me at the direction of things that
1:56:59contain these words. right before the
1:57:02algorithmic manipulation started
1:57:03steering us into believing pure nonsense
1:57:06because somebody who controlled these
1:57:08things decided it was useful for us to
1:57:10believe those things. So my guess is at
1:57:13the moment AI is enhancing our ability
1:57:16to see more clearly but that really
1:57:18depends on some kind of agreement to
1:57:21protect that capacity that I'm not aware
1:57:24of us having are you implying there that
1:57:27AI will protect us from AI i.e. the
1:57:30woman that got scammed in my audience,
1:57:33the platforms would have a tool built in
1:57:34which would be able to identify shortly
1:57:37that that is not me and the ad is been
1:57:40launched by someone in another country
1:57:42potentially and then also when she
1:57:43starts being asked for her credit card
1:57:45details in such a way on Telegram 10
1:57:48minutes later the system will able to
1:57:49understand there that this is probably a
1:57:51scam at that touch touch point too and
1:57:53it will also be the defense not just the
1:57:55offense.
1:57:56First thing uh question is Meta
1:57:59incentivized to solve this problem? Yes.
1:58:03Yes. And so Meta is probably actively
1:58:06working on AIs and again it's going to
1:58:08be a cat and mouse game like every abuse
1:58:10that happens out there. So I I think
1:58:13that the market will naturally respond
1:58:15to things like that in the same way that
1:58:18you know we installed antiviruses as you
1:58:21know annoying as they are. I think we'll
1:58:23install uh AIS on our computers that
1:58:27will allow us to at least help us kind
1:58:29of sort the the fake from from the
1:58:31truth. Well, but let's let's take the
1:58:33example you say. Is Meta incentivized to
1:58:36solve this problem? Superficially, it
1:58:38seems that it should be, but how many
1:58:40times in recent history have we watched
1:58:42a corporation cannibalize its own
1:58:45business over what at best is the
1:58:47bizarre desires of its shareholders,
1:58:50right? Why was X throwing off people
1:58:53with large accounts or Facebook or
1:58:58Google? It would seem that you would
1:59:00expect based on the market choosing
1:59:03search engines or social media sites,
1:59:05you would expect these companies to be
1:59:08absolutely mercenary and say, you know,
1:59:10if Alex Jones has a big audience, who
1:59:12are we to say? That's what I would have
1:59:14expected. Instead, you had these
1:59:18companies policing the morality of
1:59:21thought even though it reduced the size
1:59:25of the population using the platforms. I
1:59:27have a hard time explaining why that
1:59:29happened, but I have every reason to
1:59:30expect the same thing will happen with
1:59:32AI. What are you excited about with AI?
1:59:34What's your your optimistic take?
1:59:37Because at the start of this
1:59:37conversation, you said that there's
1:59:39infinite ways that it could improve our
1:59:41lives and there's 10 times more ways
1:59:43that it could hurt our lives. But let's
1:59:45investigate some of those ways that it
1:59:46could drastically improve our lives.
1:59:48There's a couple of different ways. One,
1:59:50we have, as we mentioned before, a der
1:59:53of competent teachers and professors.
1:59:57And that is a problem that will take
1:59:59three generations at least to solve if
2:00:02what we're going to do is start tomorrow
2:00:04and start educating people in the right
2:00:06way that would make them competent to
2:00:07stand at the front of a room and
2:00:08educate. But if we can augment that
2:00:10process, if we can leverage a tool like
2:00:13AI so that you know a small number of
2:00:16competent teachers can maybe reach a
2:00:18larger number of pupils, that's
2:00:20plausible I think. Second thing is we
2:00:24have a tremendous number of problems
2:00:26that are obstacles to us living well on
2:00:30this planet that AI might be able to
2:00:33manage that human intellect alone
2:00:35cannot. Right? Just in the same way that
2:00:39you know compute power can calculate
2:00:41things at a rate that human beings can't
2:00:43keep up and there are certain things you
2:00:44want calculated very well. There are
2:00:46also some reasoning problems. You could
2:00:49imagine that instead of having um static
2:00:54laws that govern behavior poorly because
2:00:59they get gamed that you could have a
2:01:01dynamic interaction. You could specify a
2:01:05an objective of something like a law and
2:01:08then you could monitor whether or not a
2:01:10particular intervention successfully
2:01:13moved you in the direction that you were
2:01:14hoping to go or did something
2:01:16paradoxical which happens all the time
2:01:18and you could have you could basically
2:01:19have governance that is targeted to
2:01:22navigation and prototyping rather than
2:01:24to specifying a blueprint for how we are
2:01:27to live. So we wouldn't need
2:01:29politicians. Um, at the moment we're
2:01:32stuck with, you know, constitutional
2:01:35protections that are as good as has been
2:01:38constructed and still inadequate to
2:01:41modern realities.
2:01:43Dan, what are you excited about with AI
2:01:45from an individual level, but also from
2:01:47a societal level? Yeah. Well, the big
2:01:49ones are healthcare and education. I
2:01:51mean, it's ridiculous that you uh are
2:01:54sitting there in pain, having had an
2:01:56MRI, and there just hasn't been someone
2:01:58to look at that MRI yet. and and tell
2:02:00you what to do. Um and that could easily
2:02:03be solved there's all sorts of
2:02:04healthcare issues where um and also not
2:02:07only that throughout the entire world
2:02:09there are places that just don't have
2:02:10general practitioners and they don't
2:02:12have you know medical advisers and and
2:02:15you know the breakthroughs in global
2:02:16healthcare will be phenomenal and the
2:02:18breakthroughs in global education could
2:02:20be transformational um on the planet. I
2:02:23I'm excited at an individual level that
2:02:26I think the industrial age created a
2:02:28bunch of jobs that are very dehumanizing
2:02:30and we've just kind of gotten used to
2:02:32them and put up with them. The idea that
2:02:33work should be repetitive and you know
2:02:36you just repeat the same loop over and
2:02:38over and over over again and over a
2:02:4010-year period of time you might get you
2:02:42know graduated up one gear and all that
2:02:44kind of stuff. I don't think that's very
2:02:46human. Um the idea that you could be
2:02:49simultaneously writing a book, launching
2:02:51a business, running a team, launching a
2:02:54festival, having an event. Um that that
2:02:57that you could actually be doing this
2:02:59kind of like mini kingdom work where
2:03:01you've got this little, you know, uh
2:03:04ecosystem around you of fun things that
2:03:06you're involved in that is actually made
2:03:08possible for a vast majority of people
2:03:10if they embrace these kind of tools. um
2:03:13you can live an incredibly fulfilling
2:03:15and amazing and impactful existence or I
2:03:19know that I do as a result of having
2:03:21these tools in my life. Like I'm I'm
2:03:23doing things that I could have only
2:03:24dreamed about uh as a kid. And what
2:03:26would you say to entrepreneurs? I know
2:03:28you you work with thousands of
2:03:29entrepreneurs. What are you telling them
2:03:31in terms of their current businesses or
2:03:33business opportunities that you're
2:03:34foreseeing? So I think that small teams
2:03:36have infinite leverage now and that when
2:03:39you have a team of say five to 10 people
2:03:43who share an incredible passion for a
2:03:45meaningful problem in the world and they
2:03:47want to see that meaningful problem
2:03:48solved and they come together in the
2:03:50spirit of entrepreneurship to solve that
2:03:52problem. That little 5 to 10 person team
2:03:56armed with the technology that we now
2:03:57have available, you can have a a big
2:04:01impact. You can make a lot of money. You
2:04:03can have a lot of fun. you can solve
2:04:04meaningful problems in the world. You
2:04:06can scale solutions. You can probably do
2:04:08more in a three-year window than most
2:04:10people did in a 30-year career. Uh and
2:04:13then that little band of 5 to 10 people
2:04:15could either go together onto a new
2:04:18meaningful problem or they could disband
2:04:20and you know work on other meaningful
2:04:22problems with different teams. In such a
2:04:25world where you have this sort of
2:04:26infinite leverage but everyone else has
2:04:29access to the same infinite leverage.
2:04:31What becomes the USP? Going back to this
2:04:33idea of the moat, like what is the thing
2:04:34of value when we've all got access to
2:04:36$20 infinite leverage? Well, first of
2:04:39all, the first thing you need to do you
2:04:41need to understand is that this moment
2:04:43of time is the least competitive uh
2:04:47moment. Like if you understand how to
2:04:49use these tools, you can start making
2:04:51money tomorrow. like we, you know, I see
2:04:53countless examples of people making
2:04:56thousands of dollars with these hustles
2:04:57that I that I talked about or building
2:05:00businesses that generates millions of
2:05:01dollars in the first couple of months of
2:05:03existence. So, I would say start moving
2:05:05now. Start building things. So, it's an
2:05:08unprecedented time of of of wealth
2:05:11creation. Clearly at some point as the
2:05:15market gets more efficient as people
2:05:17more and more people understand how to
2:05:18use these tools um there's less
2:05:22potential for uh you know creating these
2:05:25massive businesses quickly and we've
2:05:27seen this like the dawn of the internet
2:05:29or dawn of the web you know it was a lot
2:05:31easier to create Facebook than it is now
2:05:34then we had mobile and for three four
2:05:37five years it was very easy to create
2:05:40massive businesses and then it became
2:05:42harder. Being just at the edge of what's
2:05:44possible is going to be very very
2:05:46important over the next couple years.
2:05:48And that's that gets me really excited
2:05:49because the entrepreneurs who are paying
2:05:51attention are going to are going to be
2:05:53having the most amount of fun, but
2:05:55they're also going to be able to make a
2:05:56lot of money. How many applications have
2:05:59been built on Replet to date? So, you
2:06:02know, I can talk about the millions of
2:06:03things that have been built since the
2:06:05since we started the company, but just
2:06:07since uh September when we launched
2:06:10Replent, there's been about 3 million
2:06:12applications built purely in natural
2:06:16language with no uh with no coding at
2:06:20all, purely natural language. uh of
2:06:22those I think uh 300 400,000 of them
2:06:25were deployed in a real um in the site
2:06:30was deployed and it is having people are
2:06:33using it some kind of business some kind
2:06:35of internal tool. I built one last night
2:06:37by the way an internal tool or I uh
2:06:39built an application to track um how my
2:06:43kids earn pocket money. Amazing. So, I
2:06:46just told it that I wanted to track the
2:06:48tasks that are happening around the
2:06:49house and put an assign a value to them
2:06:51and I want to be able to at the end of
2:06:52the week push a button and get a summary
2:06:54of how much to pay each child um for
2:06:57their pocket money. we are so screwed.
2:07:02And within 15 minutes, it had created
2:07:04this application and it was amazing.
2:07:06Like you could toggle between like
2:07:08here's the place where you have the kids
2:07:10and here's the weekly reports and here's
2:07:12the um how much per task and you can
2:07:15tick off the tasks or remove tasks or
2:07:17add tasks. So then I now have this
2:07:19application for which took 15 minutes of
2:07:23just talking about what I wanted and now
2:07:25I have an application to run the pocket
2:07:27money uh situation in the in the house.
2:07:29And this by the way it's something
2:07:31having run an IT agency years ago.
2:07:34That's something that we would have
2:07:35charged five to10,000 pounds to create
2:07:38or5 to 10,000 US dollars to to create
2:07:42and how much time probably talking
2:07:44something that would have been a three
2:07:46four week project and we're at the start
2:07:49of the S-curve now that you're
2:07:50describing and it's already if you if
2:07:53it's a $20 replet's roughly $20 a month
2:07:5525 for the for the base case you did one
2:07:58day of usage let's say it's a dollar it
2:08:01cost you and it cost you minutes and a
2:08:03dollar now and we're at the start of the
2:08:04scurve and and you talk to it like
2:08:06you're chatting to a developer. So one
2:08:08of the things that slows down the
2:08:10development process is you have to send
2:08:12the information to a developer and they
2:08:14need to understand it and then they need
2:08:15to create something and then come back
2:08:16to you. This just happens in front of
2:08:18your eyes uh while you're watching it
2:08:20and it's actually showing you what's
2:08:21being built and it's it's really wild.
2:08:25This one change has transformed how my
2:08:27team and I move, train and think about
2:08:29our bodies. When Dr. Daniel Lieberman
2:08:31came on the DEO. He explained how modern
2:08:34shoes with their cushioning and support
2:08:36are making our feet weaker and less
2:08:38capable of doing what nature intended
2:08:40them to do. We've lost the natural
2:08:42strength and mobility in our feet and
2:08:44this is leading to issues like back pain
2:08:46and knee pain. I'd already purchased a
2:08:49pair of Viva barefoot shoes. So, I
2:08:50showed them to Daniel Lieberman and he
2:08:52told me that they were exactly the type
2:08:54of shoe that would help me restore
2:08:55natural foot movement and rebuild my
2:08:57strength. But I think it was
2:08:58plantficitis that I had where suddenly
2:08:59my feet started hurting all the time.
2:09:01And after that I decided to start
2:09:02strengthening my own foot by using the
2:09:04Vivo barefoot. And research from
2:09:06Liverpool University has backed this up.
2:09:07They've shown that wearing Vivo barefoot
2:09:09shoes for 6 months can increase foot
2:09:11strength by up to
2:09:1360%. Visit vivo
2:09:15barefoot.com/doac and use code diary 20
2:09:19from my sponsor for 20% off. A strong
2:09:21body starts with strong feet. This has
2:09:25never been done before. A newsletter
2:09:28that is ran by 100 of the world's top
2:09:31CEOs. All the time people say to me,
2:09:34they say, "Can you mentor me? Can you
2:09:35get this person to mentor me? How do I
2:09:37find a mentor?" So, here is what we're
2:09:39going to do. You're going to send me a
2:09:41question. And the most popular question
2:09:42you send me, I'm going to text it to 100
2:09:46CEOs, some of which are the top CEOs in
2:09:49the world running a hundred billion
2:09:50dollar companies. And then I'm going to
2:09:52reply to you via email with how they
2:09:54answered that question. You might say,
2:09:57"How do you hold on to a relationship
2:09:59when you're building a startup? What is
2:10:00the most important thing if I've got an
2:10:02idea and don't know where to start?" We
2:10:04email it to the CEOs. They email back.
2:10:05We take the five, six top best answers.
2:10:07We email it to you. I was nervous
2:10:09because I thought the marketing might
2:10:10not match the reality. But then I I saw
2:10:12what the founders were replying with and
2:10:14their willingness to reply and I thought
2:10:15actually this is really good and all
2:10:17you've got to do is sign up completely
2:10:20free. I don't think we've spent a lot of
2:10:23time talking about autonomous weapons.
2:10:25This is the thing that really worries
2:10:27me. And the thing that worries people
2:10:30about AI is this idea is that it is uh
2:10:34this you know emergent system and
2:10:36there's no one thing behind it and it
2:10:38can be it can act in a way that's uh
2:10:41unpredictable and not really guided by
2:10:43humans. also think it's true of
2:10:44corporations of governments and so I
2:10:47think individual people uh can often
2:10:50have the best intentions but the
2:10:52collective can land on doing things in a
2:10:56way that's harmful or morally irrepant
2:11:00and I think um we talked about China
2:11:03versus the US and that creates a certain
2:11:06race dynamics where um they're both
2:11:09incentivized to cut corners and
2:11:11potentially do do harmful things and in
2:11:14the world of geopolitics
2:11:17um and wars, you know, what really
2:11:20scares me is is autonomous weapons. And
2:11:23why does it scare you? Because
2:11:28uh you know you can imagine
2:11:32uh autonomous drones being trained on
2:11:36someone's face and you can send a a
2:11:39swarm of of drones and they can be this
2:11:43um sort of autonomous killing
2:11:45assassination machine and it can sort of
2:11:48uh function as a you know country verse
2:11:51country technology in in the world of
2:11:54for which is still crazy but it can also
2:11:58become a tool for governments to
2:12:01subjugate the citizens and and people
2:12:04think we're we're safe in the west but I
2:12:08think the experience with co showed that
2:12:11even the systems in the west can very
2:12:16quickly become draconian. Yeah.
2:12:18Apparently, I've heard in um Iran that
2:12:22uh they have facial recognition cameras
2:12:25that detect where the women are wearing
2:12:27hijabs in their own cars and it
2:12:30automatically detains the car. If you're
2:12:34driving and you're not wearing a hijab
2:12:36and if you're certainly if you're
2:12:38walking down the street, it just picks
2:12:40that up and immediately you're you're in
2:12:42trouble. uh you can like it acts as a
2:12:45police officer and a judge and you know
2:12:49a law m lawmaker it's the judge jury and
2:12:52executioner essentially and it's just
2:12:54happens instantaneously what happened in
2:12:56Canada with the truckers uh uh sort of
2:13:00protest where they froze their bank
2:13:02account by virtue of just being there
2:13:03just by being in that location and just
2:13:06to confirm that Iran has implemented a
2:13:08comprehensive surveillance system to
2:13:09enforce its mandatory hijab laws
2:13:12utilizing various ious technologies, one
2:13:13of which is cameras and facial
2:13:15recognition. So they've put cameras in
2:13:17public spaces to identify women who are
2:13:19not adearing to the hijab dress code.
2:13:22Yeah. And just on that, London has just
2:13:25put those face cam facial recognition
2:13:27systems into London and also all
2:13:29throughout Wales. um and they're being
2:13:31rolled out at speed
2:13:34and like all you would need is a change
2:13:37of government that wanted to implement
2:13:39something similar and all the base layer
2:13:42technologies already in there. It gets a
2:13:43little bit worse in Iran because they
2:13:45have this new app called the Nazar app
2:13:46where the government has introduced the
2:13:48Nazar mobile application which allows
2:13:49you as a citizen to report another
2:13:51citizen who is not wearing their hijab
2:13:53and it it logs their location, their
2:13:55time um when they weren't wearing it and
2:13:57the vehicle license plate with the
2:13:59crowdsource data. It can then go after
2:14:01that individual. I would also just point
2:14:04out that I think we're not being
2:14:05imaginative enough. I agree with you. I
2:14:07have the same concern about these
2:14:09autonomous weapons, but I also think
2:14:11this doesn't have to occur in the
2:14:13context of war or even governmental
2:14:16oppression that it is perfectly
2:14:17conceivable that effectively this allows
2:14:20this drops the price of a an
2:14:24undetectable or an unprosecutable crime.
2:14:27And maybe economic moes return in the
2:14:30form of people taking out their
2:14:32competitors or anybody who attempts to
2:14:33compete with them using an autonomous
2:14:35drone that can't be traced back to them.
2:14:38You know, that follows facial
2:14:39recognition. And you know, you don't
2:14:41have to kill very many people for others
2:14:42to get the message that uh this is a a
2:14:45zone that uh you shouldn't mess around
2:14:47in. So, I could imagine, you know,
2:14:50effectively a new high-tech organized
2:14:53crime that uh protects rackets and makes
2:14:58tons of money and subjugates people who
2:15:00haven't done anything wrong. I had
2:15:02Mustafa Sullivan on the podcast in 2023
2:15:04when this all of this stuff started
2:15:06kicking off and he is the CEO of
2:15:08Microsoft AI. You're familiar with
2:15:10Mustafa? Of course. Yeah. Um and he one
2:15:12of the things he said to me at the time
2:15:13was one of my fears is a tiny group of
2:15:15people who wish to cause harm are going
2:15:17to have access to tools that can
2:15:19instantly destabilize our world. That's
2:15:21the challenge. How to stop something
2:15:23that can cause harm or potentially kill.
2:15:25That's where we need containment. And it
2:15:27sounds a little bit like what you're
2:15:28saying amad that we will now have these
2:15:31these tools. you were talking in the
2:15:32context of the military, but as Brett
2:15:35said there, even smaller groups of
2:15:36people that might have been, I don't
2:15:38know, cartels or gangs can do similar
2:15:41harm. And at the moment, in terms of
2:15:43autonomous weapons, both the US and
2:15:45China are investing heavily in AI
2:15:47powered weapons, autonomous drones, and
2:15:48cyber warfare because they're scared of
2:15:50the other one getting it first. And we
2:15:53we talked about how much of our lives
2:15:54run on the internet, but cyber weapons
2:15:56and cyber AI agents that could be
2:15:59deployed to take down China's X, Y, or
2:16:02Zed or vice versa are real concern.
2:16:05Yeah. Yeah. I I think all of that is is
2:16:10um is a real concern. You know, unlike
2:16:12Mustafa, I I don't think containment is
2:16:14is possible. Part of the reason why this
2:16:18game theoretic system uh of competition
2:16:21between the US, China, corporations,
2:16:26individuals makes it so that this this
2:16:28technology is already, you know, is
2:16:31already out and really hard to put it
2:16:33back in the in the bag. I did ask him
2:16:36this question and I remember the answer
2:16:37because it was such a stark moment for
2:16:38me. I said to Mustafa, "Do you think
2:16:40it's possible to contain it?" And he
2:16:41replied, "We must." So I asked him
2:16:43again. I said, "Do you think it's
2:16:44possible to contain it?" and he replied
2:16:45we must and I asked him again I said do
2:16:47you think it's possible we must so the
2:16:49problem with that uh uh chain of
2:16:51thinking is that it might lead to an
2:16:53oppressive system
2:16:55uh there is uh one of the say doomers or
2:16:58philosophers of of AI which I I respect
2:17:01his work his name is Nick Bostonramm and
2:17:03he he uh he's he was trying to think of
2:17:08ways in which we can contain AI and the
2:17:13thing that he came up with is perhaps
2:17:16more oppressive than something that the
2:17:18AI would come up with is like total
2:17:20surveillance state. You need total
2:17:22surveillance on compute on people's
2:17:25computers on people's ideas to not
2:17:27invent AI or AGI. It's like taking the
2:17:30guns or something or Right. Exactly. I
2:17:32mean there's always there's always this
2:17:34problem of containing any sort of
2:17:36technology is that you do need um
2:17:39oppression and draconian policies to do
2:17:42that. Are you scared of anything else or
2:17:44concerned about anything else as it
2:17:45relates to AI outside of autonomous
2:17:47weapons? You know, we talked at the
2:17:49about
2:17:50the birthight crisis and I think a more
2:17:54generalized problem there is creating
2:17:58virtualized environments
2:18:01uh via VR where everyone is living in
2:18:04their own created universe and uh it's
2:18:09so enticing and even create simulates
2:18:12work and simulates struggle uh such that
2:18:15you don't really need to leave this this
2:18:17world and so every one of us will be
2:18:20solopcystistic, you know, similar to the
2:18:21Matrix. Ready Player One. Ready Player
2:18:24One. We're all kind of uh plug even
2:18:26worse than Ready Player One. At least
2:18:28that's a massivelyworked
2:18:30environment. I'm talking about AI
2:18:32simulating everything uh for us and
2:18:36therefore you're literally in the
2:18:37matrix. You know, maybe that this is I
2:18:40was about I had that same thought. I've
2:18:42enjoyed this great simulation. Yes. and
2:18:45and and so I mean are you familiar with
2:18:48the Fermy's paradox? No, I'm not. So
2:18:50Fermy's paradox is um the question uh
2:18:53the you know professor his name is uh
2:18:56Fermy he asked the question uh if the
2:19:00universe is is that vast then where are
2:19:03the aliens? the fact that humans exist,
2:19:07you can deduce that other civilizations
2:19:10exist. And if they do exist, then why
2:19:14don't we see them? And then that spurred
2:19:17a bunch of Fermy solutions. So there's
2:19:20uh I don't know, you can find hundreds
2:19:21of solutions on the internet. One of
2:19:24them is the uh sort of house cat on a
2:19:26thought experiment where actually aliens
2:19:29exist, but they kind of put us in an
2:19:31environment like the Amish in a certain
2:19:34time and do not expose us to what's
2:19:37going on out there. So they we're pets.
2:19:39Maybe they're watching us and kind of
2:19:40enjoying uh what we're doing, stopping
2:19:42us from hurting ourselves, stopping us
2:19:44from hurting ourselves. There are so
2:19:45many things, but one of the things that
2:19:48I think is potentially a solution to the
2:19:51phrases paradox and one of the saddest
2:19:54outcomes is that civilizations progress
2:19:58until they invent technology that will
2:20:01lock us into infinite pleasure and
2:20:03infinite simulation such that we we
2:20:07don't have the motivation to go into
2:20:11space to seek out the explor
2:20:14exploration, potentially other alien
2:20:17civilizations. And perhaps that is a
2:20:20determined outcome of humanity or like a
2:20:24highly likely outcome of any species
2:20:27like humanity. We like pleasure.
2:20:29Pleasure and pain is the main
2:20:30motivators. And so if you create an
2:20:33infinite pleasure machine, does that
2:20:35mean that we're just at home in our VR
2:20:38environment with everything taking care
2:20:40for us and literally like the matrix and
2:20:42the world the real world would suck in
2:20:44such a scenario? Yes. Be terrible. I
2:20:46mean the other simpler explanation of
2:20:48the Fermy paradox is that you generate
2:20:51sufficient technology that you can end
2:20:53your species and it's only a matter of
2:20:55time from that point which you know we
2:20:57can have that discussion about nuclear
2:20:59weapons. We can have it about AI, but
2:21:02does some technology, if if we stay on
2:21:04that escalator, does some technology
2:21:06that we generate ultimately whatever
2:21:08allows you to get off the planet allows
2:21:10you to blow up the planet? There you go.
2:21:12I want to get everyone's closing
2:21:13thoughts and closing remarks. And
2:21:16hopefully in your closing remarks, you
2:21:17can capture something actionable for the
2:21:20individual that's listening to this now
2:21:22on their commute to work or the single
2:21:24mother, the average person who maybe
2:21:25isn't as technologically advanced as
2:21:27many of us at this table, but is trying
2:21:29to navigate through this to figure out
2:21:31how to live a good life over the next
2:21:3310, 20, 30 years. Yeah. Take as long as
2:21:37you need. I think we live in the most uh
2:21:40interesting time in human history. So
2:21:43for the single mother that's listening,
2:21:45for someone who wouldn't be the
2:21:46stereotype of a tech row, don't assume
2:21:49that you can't do this stuff. It's never
2:21:52been more accessible today within your
2:21:54work. You can be an entrepreneur. You
2:21:56don't have to take massive risk to go
2:21:58create a business um by you quit your
2:22:02job and go create a business. There are
2:22:04countless examples. We uh we have a user
2:22:06who's a product manager at a larger real
2:22:09estate business and he built something
2:22:12that created 10% lift in conversion
2:22:15rates which generated millions and
2:22:17millions of dollars of that business and
2:22:19that person became celebrity at that
2:22:21company and became someone who is
2:22:23lifting everyone else up and teaching
2:22:25them how to use these tools and
2:22:27obviously that that is like a really
2:22:29great for for anyone's career and you're
2:22:31going to get a promotion and your
2:22:33example of building a piece of software
2:22:36for your family for your kids to to
2:22:39improve and and to learn more to be
2:22:42better kids uh as an example of being
2:22:44entrepreneur in your family. So I really
2:22:48want people to break away from this
2:22:51concept of entrepreneurship being this
2:22:54is your podcast a diary of a CEO. You
2:22:56started this podcast by talking to CEOs
2:22:59I assume right and over time uh it
2:23:02changed to everyone can be a CEO
2:23:05everyone is some kind of CEO in their
2:23:07life and so uh I think that we
2:23:12have unprecedented access to tools for
2:23:15that vision to actually come to reality.
2:23:19Well, it is obviously a moment of a kind
2:23:22of human phase transition. Something
2:23:25that I believe will be the equal of a
2:23:28discovery of farming or writing or
2:23:34electricity. And the darkness that I
2:23:38think is valid in looking at all of the
2:23:40possible outcomes of this scenario is
2:23:43actually potentially part of a different
2:23:45story as well. In evolutionary biology,
2:23:47we talk about an adaptive landscape in
2:23:50which a niche is represented as a peak
2:23:54and a higher niche, a better niche is
2:23:57represented as a higher peak. But to get
2:23:59from the lower niche to the higher
2:24:00niche, you have to cross through what we
2:24:02call an adaptive valley. And there's no
2:24:04guarantee that you make it through the
2:24:05adaptive valley. And in fact, the
2:24:08drawing that we put on the board, I
2:24:09think, is overly hopeful because it
2:24:11makes it in two dimensions. It looks
2:24:13like you know exactly where to go to
2:24:15climb that next peak. And in fact, it's
2:24:17more like the peaks are islands in an
2:24:20archipelago that is in fog where you
2:24:23can't figure out what direction that
2:24:25peak is and you have to reason out it's
2:24:27probably that way and you hope not to
2:24:29miss it by a few degrees. But in any
2:24:32case, that darkness is exactly what you
2:24:36would expect if we were about to
2:24:38discover a better phase for humans. And
2:24:42I think we should be very deliberate
2:24:43about it this time. I think we should
2:24:45think carefully about how it is that we
2:24:47do not allow the combination of this
2:24:50brand new extremely powerful technology
2:24:53and market forces to turn this into some
2:24:57new kind of enslavement. And I don't
2:25:00think it has to be. I think the
2:25:02potential here does allow us to refactor
2:25:05just about everything. Maybe we have
2:25:07finally arrived at the place where
2:25:10mundane work doesn't need to exist
2:25:12anymore and the pursuit of meaning can
2:25:14replace it. But that's not going to
2:25:16happen automatically if we don't figure
2:25:18out how to make it happen. And I hope
2:25:21that we can recognize that the peril of
2:25:24this moment is best utilized if it
2:25:27motivates us to confront that question
2:25:30directly.
2:25:32Each one of us has two parents, four
2:25:34grandparents, eight great-grandparents,
2:25:3716, 32, 64. You've got this inc like
2:25:41this long line of ancestors who all had
2:25:44to meet each other. They all had to
2:25:45survive wars. They all had to survive
2:25:48illness and disease. Everything had to
2:25:51happen for us. One, each individual, all
2:25:53of this this stuff had to happen for us
2:25:55to get here. And if we think about all
2:25:57the people in those thousands and
2:25:59thousands of people, every single one of
2:26:01them would trade places in a heartbeat
2:26:03if they had the opportunity to be alive
2:26:05at this particular moment. They would
2:26:07say that their life was struggle,
2:26:11disease, that their life was a lot of
2:26:14mundane and meaningless work. It was
2:26:16dangerous. You know, every single one of
2:26:18us has probably got ancestors that were
2:26:20enslaved, probably got ancestors that
2:26:24died too young, uh, probably got
2:26:27ancestors that worked horrific
2:26:29conditions. We all have that. And they
2:26:32would all just look at this moment and
2:26:34say, "Wow." So, are you telling me that
2:26:36you have the ability to solve meaningful
2:26:38problems, to come up with adventures, to
2:26:41travel the world, to pick the brains of
2:26:44anyone on the planet that you want to
2:26:45pick the brains of? You can just listen
2:26:47to a podcast. You can just watch a
2:26:49video. You can talk to an AI. Like, are
2:26:52you telling me that you're alive at this
2:26:53particular moment? Please make the most
2:26:56of that. Like, do something with that.
2:26:59You know, you can sit around
2:27:00pontificating about society and how
2:27:03society might work. But ultimately, it
2:27:05all boils down to what you do with this
2:27:07moment. and solving meaningful problems,
2:27:10being brave, having fun, making your
2:27:14little dent in the universe. You know,
2:27:16that's that's what it's all about. And I
2:27:18feel like there's an obligation to your
2:27:20ancestors to make the most of the
2:27:22moment.
2:27:24Thank you so much to everybody for being
2:27:26here. I I've learned a lot and I've
2:27:28developed my thinking, which is much the
2:27:30reason why I wanted to bring us all
2:27:31together because I know you all have
2:27:32different experiences, different
2:27:34backgrounds in education. and you're
2:27:35doing different things, but together it
2:27:37helps me sort of pass through all of
2:27:38these ideas to figure out where I land.
2:27:40And I I I ask a lot of questions, but I
2:27:43am actually a believer in humans. I'm I
2:27:46I was thinking about this a second ago.
2:27:48I was thinking, do I am I optimistic
2:27:50about humans ability to navigate this
2:27:52just because I have no other choice?
2:27:54Because as you said, the alternative
2:27:56actually isn't worth thinking about. And
2:27:57so I do have a optimism towards how I
2:28:00think we're going to navigate this in
2:28:02part because we're having these kinds of
2:28:03conversations and we in history haven't
2:28:06always had them at the birth of a new
2:28:08revolution when we think about social
2:28:09media and the implications that had.
2:28:10We're playing catchup with the with the
2:28:12downstream
2:28:14consequences. And I am hopeful. Maybe
2:28:17that's the entrepreneur in me. I'm
2:28:18excited. Maybe that's also the
2:28:20entrepreneur in me. But at the same
2:28:22time, to many of the points Brett's
2:28:23raised and Amjud's raised and Dan's
2:28:25raised, there are serious considerations
2:28:27as we swim from one island to another.
2:28:29And because of the speed and scale of
2:28:32this transformation that Brett
2:28:33highlights and you look at the stats of
2:28:35the growth of this technology and how
2:28:37it's spreading like wildfire and how
2:28:38once I tried Replet, I walked straight
2:28:41out and I told Cozy immediately I was
2:28:42like, "Cussie, try this." And she was on
2:28:44it and she was hooked. And then I called
2:28:45my girlfriend in Bali who's the breath
2:28:47work practitioner and I was like, "Type
2:28:48this into your browser. R E P L I T."
2:28:51And then she's making these breath work
2:28:53schedules with all of her clients
2:28:55information ahead of the retreat she's
2:28:56about to do. It's spreading like
2:28:58wildfire because we're internet native.
2:29:00We were native to this technology. So
2:29:02it's not a new technology. It's
2:29:03something on top of something that's
2:29:04intuitive to us. So that transition, as
2:29:07Brett describes it, from one peak to the
2:29:08other or one island to another, I think
2:29:10is going to be incredibly destabilizing.
2:29:12And I've having interviewed so many
2:29:14leaders in this space from Reed Hoffman
2:29:15who's the founder of LinkedIn to the CEO
2:29:17of Google to Mustafa who I mentioned
2:29:20they don't agree on much but the thing
2:29:22that they all agree on and that Sam
2:29:23Alman agrees on is that the long-term
2:29:26future the long-term way that our
2:29:28society functions is radically
2:29:29different. People squab squabble over
2:29:31the short term. They sometimes even
2:29:33squabble over the midterm or the
2:29:34timeline but they all agree that the
2:29:37future is going to look completely
2:29:38different. Amjud, thank you for doing
2:29:40what you're doing. you're what we didn't
2:29:42get to spend a lot of time on it today
2:29:43and this is typically what I do here but
2:29:45your story is incredibly inspiring
2:29:48incredibly inspiring from where you came
2:29:50from what you've done what you're
2:29:51building and you are democratizing and
2:29:53creating a level of playing field for
2:29:55entrepreneurs in Bangladesh to Cape Town
2:29:58to San Francisco to be able to turn
2:30:00their ideas into reality and I do think
2:30:02just on the surface that that's such a
2:30:04wonderful thing that you know I was born
2:30:06in Botswana in in Africa and that I
2:30:08could have the same access to turn my
2:30:11imagination into something to change my
2:30:13life because of the work that you're
2:30:14doing at Replet. And I highly recommend
2:30:16everybody go check it out. You you
2:30:17probably won't sleep that night because
2:30:19it's so it's so for someone like me it
2:30:21was so addictive to get to be able to do
2:30:23that because it's been the barrier to
2:30:24creation my whole life. I've always had
2:30:25to call someone to build something. Dan,
2:30:28thank you again so much because you
2:30:29represent the voice of entrepreneurs and
2:30:30you've really become a titan as a a
2:30:31thought leader for entrepreneurs in the
2:30:32UK and that perspective that balance is
2:30:35incredibly important. So, I really
2:30:36really appreciate you being here as
2:30:38always and you're a huge fan favorite of
2:30:40our show and Brett, thank you a
2:30:41gazillion times over for being a a human
2:30:45lens on complicated challenges and you
2:30:48do it with a fearlessness that I think
2:30:50is imperative for us finding the truth
2:30:52in these kind of situations where some
2:30:54of us can run off with optimism and we
2:30:56can be hurtling towards the mouse trap
2:30:58because we love cheese and I think
2:31:00you're an important
2:31:01counterbalance and voice in the world at
2:31:04this time. So, thank all of you for
2:31:05being here. I really really appreciate
2:31:06it and um we shall
2:31:11see. These things live forever.
2:31:17So, this has always blown my mind a
2:31:19little bit. 53% of you that listen to
2:31:22this show regularly haven't yet
2:31:23subscribed to the show. So, could I ask
2:31:25you for a favor? If you like the show
2:31:26and you like what we do here and you
2:31:28want to support us, the free simple way
2:31:29that you can do just that is by hitting
2:31:31the subscribe button. And my commitment
2:31:33to you is if you do that, then I'll do
2:31:35everything in my power, me and my team,
2:31:36to make sure that this show is better
2:31:38for you every single week. We'll listen
2:31:40to your feedback. We'll find the guests
2:31:41that you want me to speak to, and we'll
2:31:43continue to do what we do. Thank you so
2:31:45much.
2:31:48[Music]
2:32:05[Music]

About this transcript

This page contains the full transcript of AI AGENTS DEBATE: These Jobs Won't Exist In 24 Months! by The Diary Of A CEO, generated from the public captions YouTube serves with the video. The transcript has 27,465 words across 3,903 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 choose a caption track available on the video.

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, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.