YT YouTube2Text

YouTube2Text → Transcripts → What do tech pioneers think about the AI revolution? - The Engineers, BBC World Service

What do tech pioneers think about the AI revolution? - The Engineers, BBC World Service — Transcript

by BBC World Service · 4,289 words · 458 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Hello.
0:00I'm Caroline Steel.
0:02This is the BBC World Service.
0:04And welcome to The Engineers.
0:22This year we're at the Science Specialist University, Imperial College London.
0:26And we're here to focus on the technical revolution
0:29defining our era, artificial intelligence.
0:33I'm joined by a panel of three world leaders in the field and a large,
0:37enthusiastic audience in Imperial's Great Hall.
0:44Already a computer can defeat the world's
0:46greatest player at our most complex strategy game.
0:50The first movie written entirely by AI
0:52has just been released, and AI may have discovered
0:55our first antibiotic in three decades.
0:59Together with our partners, Royal Commission 1851,
1:02we've brought together three engineers at the cutting edge of this field
1:06to discuss their work and what it means for us humans.
1:10Paolo Pirjanian
1:10is Armenian, but he was born in Iran and started his career working on Mars
1:15rovers for NASA.
1:16He's now founder and CEO of Embodied, which is a company
1:19that builds emotionally intelligent robots to help with child development.
1:24David Silver
1:25is from the UK, where he's principal research scientist at the AI
1:29Research Lab.
1:30Google DeepMind.
1:31He led the team that used AI
1:33to defeat the world's best player at the complicated strategy game Go.
1:38And he's working on artificial general intelligence.
1:41Regina Barzilay is Israeli-American and a distinguished professor for AI
1:46and Health at MIT in the U.S.
1:48She created a major breakthrough in detecting early stage breast cancer
1:52and also led the team that used AI
1:54to discover what is hoped to be a brand new antibiotic.
1:58So please do join me in welcoming them all.
2:10Regina, let's start with you.
2:13So what is it that made you shift your work to oncology?
2:16Sadly, you were in the perfect position to do that, weren't you?
2:20Yeah.
2:20So actually, I started my work at MIT in 2003
2:25as a faculty and I was working on natural language processing and AI
2:30And in 2014, I was diagnosed with breast cancer
2:33and I was treated in one of the best hospitals
2:36in the United States, Massachusetts General Hospital.
2:39And what I discovered going through the treatment that there was really no AI
2:44or not even basic information technology as part of the treatment.
2:48Neither the diagnostics nor the treatment nor the post-treatment.
2:53And, you know, after I was treated, I just was totally confused
2:57as to what I want to do,
2:58because it was the first time I realised that, you know, my life is finite
3:02and I've seen a lot of very sick people there surrounding me.
3:06And I was thinking, What can I do?
3:08And MGH - this hospital at MIT - is just one subway stop away,
3:12they are separated by a bridge
3:14I’m saying, how come we have all this great technology at MIT,
3:18but none of it is actually coming to the hospitals and helping patients.
3:23So after I finished my treatment, I still didn't have my hair.
3:26I started kind of going from doctor
3:27to doctor, asking them, you know, how I can bring AI - I will do it for free.
3:31I'm a professor.
3:32So there were not many takers, but eventually
3:35we found somebody who... it was a doctor called Connie Lehman
3:39who had the idea that we can apply
3:43AI to do early detection of cancer.
3:46Thank you, Regina.
3:48David, you came to AI via the games industry
3:51and you did a PhD in reinforcement learning.
3:54What is reinforcement learning and how did you use it in those early days?
3:59Yeah.
3:59So I guess I started out in the games industry before
4:03I went back to academia and I was working on building computer games.
4:07And as a big part of building computer games is building the AI
4:10for those games, that kind of makes all of the characters move around.
4:14And I found myself being fundamentally disappointed
4:17by the methods that were being used in those games.
4:21And it felt like
4:21what I really wanted to do was build something that had real AI in it.
4:25I discovered this idea of reinforcement learning, which is basically a method
4:29very much like those that animals and humans use,
4:32where the system is able to learn for itself from experience, from trial
4:36and error, from trying things out and seeing what works and what doesn't.
4:39So is it sort of like when we learn to
4:42not touch fire because at some point we try it and it really hurts
4:46and we learn don't do that in the future because the consequences aren't appealing.
4:50Is it sort of like how humans learn reinforcement learning?
4:53Yes, it's a lot like that.
4:56So in fact
4:57humans are believed to have, you know, a major part of the brain
5:00which is devoted to providing a signal, giving feedback that
5:04that makes the brain actually learn
5:06to do more of the good things and less of the bad things.
5:09And so actually, that's inspired a lot of work in machine
5:12learning to make machines have that same capability.
5:15But a machine doesn't feel heat or isn't rewarded by a cookie.
5:19How can you reward a machine?
5:22Yeah, so for a machine, it's just a number.
5:24So you give it a positive number
5:27if it's done something good and a negative number if it's done
5:30something bad.
5:31And at the end of the day, everything
5:33stems from that one single number.
5:35So this one single number, which we call the reward, contains
5:38enormous power because it's the signal that drives everything.
5:41Paolo, you said your experiences of feeling
5:44alienated in foreign countries made you want to create an imaginary friend.
5:48And I'm sure much of our audience can relate to having an imaginary friend,
5:53but for most of us, they stay imaginary.
5:56How did you go about making a real one?
5:59So unfortunately, there's a lot of people
6:03in need of companionship or therapy,
6:07and there's a massive gap of labour force
6:10that can provide us, as an example, to use numbers from the U.S.
6:14We know the prevalence of things such as autism is growing rapidly.
6:18Ten years ago it was one out of about 200 kids.
6:21Today, it's one out of about 30 kids.
6:23So the experiences I had, which was leaving my family
6:27at a very young age, living abroad in a society that's amazing.
6:32I mean, these are amazing people,
6:34but yet you are different, so you are not going to be embraced.
6:38So this is not too dissimilar from a child on the autism spectrum
6:42that has a hard time expressing themselves or reading emotions from other people.
6:47And that was the genesis of creating a robot companion that understands
6:52human emotions, can create a deep relationship
6:55with a child, and it'll help them exercise and practice social skills
7:00such as eye contact, turn-taking, joint attention and so on,
7:04so that the child has a chance of being successful in their society.
7:08Thank you.
7:09Regina, what can AI do
7:11when it comes to understanding cancer that humans can't?
7:16So I think that in cancer and in many other diseases,
7:20a big question is always, how do you deal with uncertainty?
7:23And unfortunately, today we rely on humans who don't have this capacity
7:29to make predictions.
7:30And as a result, many times people get wrong treatments
7:33or they are diagnosed much later.
7:36And one question that really troubled me is, you know, how late I was diagnosed
7:42and when we already developed a model, I came back to my own mammograms
7:46and rediscovered the mammograms two years earlier
7:49already had on a tiny small cancer.
7:53Now, for human eye, for radiologists,
7:56it's impossible to diagnose it because it's so, so confusing.
8:00There's so many other spots on your tissue.
8:03So what AI can do, it can do a lot of tasks which humans cannot do.
8:06Take all the data that we have
8:08and remove the guessing out of diagnosis and treatment.
8:12Thank you.
8:14David, AI had already defeated the reigning grandmaster, Garry Kasparov, at chess
8:19well before you started your project
8:21AlphaGo and the rules of Go sound quite simple.
8:26Basically, on each turn, a player puts down a counter on the board
8:30and you gain territory by connecting your counters
8:34and the player with the most territory at the end of the game wins.
8:38So why is it harder for a computer to beat a human at Go than at chess?
8:43Which sounds more complicated.
8:45So the game of Go is this very beautiful and elegant game
8:49where it seems at first glance like the rules are very simple.
8:52But once you start to understand it a bit like unpeeling
8:55an onion, you discover more and more layers of complexity
8:58and what's amazing is that when humans play this game, they basically...
9:02If you ask them to describe how they did something, they really don't know.
9:06They've used incredible intuition.
9:08And so these amazing professional players who've devoted their entire lives
9:11to this game have built this incredible intuition and creativity
9:16and intuition and creativity are two traits which were previously
9:19considered to be very human and very hard to build into machines.
9:23So while chess, it was possible to succeed just with tactical look ahead
9:27in the game of Go, that wasn't enough because, you know, early on in the game
9:31you just have this handful of stones on the board and you really just have to
9:35imagine what the game will pan out like, you know, 300 moves later
9:39with this sort of intuitive sense of where it will go.
9:42And that required some some major breakthroughs.
9:45Paolo, debatably even more complex than Go is human children.
9:50Your human centred robot forms an emotional bond with children.
9:53How can a robot do that?
9:57Well, first of all, it's important
9:59to make clear that the robot is not meant to replace the need for human contact.
10:02It's really almost like training wheels to teach children the social skills
10:07and then be able to practice those in real life.
10:12The way the robot forms
10:15bond is that humans are wired to bonding.
10:18We create connections with inanimate objects all the time.
10:21I mean, with a robot that has eyes, can make eye contact,
10:25can smile back at you and can speak to you and express emotion and empathy.
10:30It's actually not that hard to create a bond there.
10:34And children open up to to these robots very quickly
10:37in ways that they may not even open up to their therapists or parents.
10:41Thank you, Paolo.
10:42Regina, you not only made an impact on oncology, your team at MIT used
10:47AI to discover what could be our first new
10:50antibiotic in three decades.
10:53It seems it can be E. Coli, MRSA, and strains of bacteria
10:57which are currently resistant to all other antibiotics.
10:59So I think we all wish it success.
11:03How did you do that?
11:05So I should say
11:06that, you know, developing antibiotics is not an area
11:09with an immense competition, even though their resistance to
11:14to antibiotics that we have continues to grow.
11:17This happened to be an area where pharmaceutical companies
11:20are not very active because economically it doesn't work for them.
11:25So in some ways we do need to have alternative approaches.
11:29I met a colleague and he was working.
11:31He was from biological engineering, he was working on antibiotics.
11:37And he was describing the big problem of finding new molecules
11:41which are effective against bacteria, drug resistant bacteria.
11:45But at the same time, are not toxic to humans.
11:47They have some molecules screened against, I think E.coli.
11:53We started with that
11:54and then we just gave to the machine, you know, thousands of molecules.
11:58And for each molecule you knew whether it kills a bacteria or not.
12:01It was kind of the first attempt to learn automatically.
12:04How do you look at the structure
12:05of the molecule and predict whether it would have a desired effect?
12:09We found a molecule that didn't look like something human created.
12:13And it turns out in the lab that it was able to kill
12:17using a different mechanism of action, kill it in a different way.
12:20And that's what made it so effective against so many different species.
12:25David, let's go back to you.
12:27So, so far we've been talking about systems designed to perform
12:31one task - that's known as narrow AI,
12:34but you're working towards artificial general intelligence.
12:37Could you explain what artificial general intelligence is?
12:40So if you think about humans and human intelligence, it's
12:44this wonderful and beautiful thing where we're able to learn skills
12:48which are incredibly diverse, where,
12:50you know, one person might choose to specialise in learning how to play tennis
12:54and another person might specialise
12:55in becoming an amazing chef and another person, a pianist
12:59and another person, a scientist and so when we want to build
13:04artificial intelligence, we want systems which not only solve a single problem
13:09but in a similar manner to humans, are able to approach
13:13any number of problems with intelligence, and
13:17that's capable of doing amazing things in each of those different areas.
13:20And that's what we refer to as artificial general intelligence or AGI for short.
13:25And how far off do you think we are from that being a reality?
13:29So I think it's going to be a spectrum over many years.
13:31And I also think it's likely or at least plausible
13:34that there are many breakthroughs that are still required
13:37before we can really crack, you know, the same kind of
13:41level of intelligence that humans have.
13:43Regina, you've developed AI to better predict cancer,
13:46but it's only employed in a tiny number of cases, right?
13:50Why is AI not used more widely in medicine?
13:53The problem is that we're creating a lot of great technology,
13:56but this technology is not really translated into patient care.
13:59And if I would ask the audience, you know, how many of you when you last saw
14:04your physician, have you actually seen any AI,
14:07and I'm sure not many can really attest to it.
14:10So I think
14:11the technology for many of the tasks is really mature,
14:16but there are many other questions which have nothing to do with AI per se.
14:19One is the regulation or regulations, in Europe,
14:24UK, US - continue to change.
14:27Another big problem is people don't really know how to bill for AI.
14:31And today the American doctor who uses AI loses money
14:35when they see a patient.
14:37So it's not much of a motivation.
14:39So just to clarify that, did you say that using AI for doctors in the US
14:43could actually make them lose money because it makes treatment more effective?
14:48I'm referring to a specific paper.
14:53The way the billing is done
14:57it somehow relates to the time the doctor spends with a patient.
15:00So if you have something that makes it faster, you're actually losing money.
15:04Very depressing.
15:05And Paolo, what about your challenges?
15:07The robot you've developed mimics human behaviour.
15:10What's next?
15:12I'm very hopeful with what we are doing in terms
15:14of creating social emotional AI systems
15:17that can help humanity become its best.
15:21If we can intervene early with children for instance on the autism spectrum,
15:25they have a chance of really integrating well with the society and
15:29doing really well.
15:32When we think about other vulnerable
15:35areas of our life is when we age.
15:38Social isolation, being lonely, and that leads to mental
15:42health issues, that leads to physical health issues and so on.
15:45We can have the same systems become a companion that help you there.
15:49Once we figure out the physical task,
15:51you can also imagine that they can give you assistive care,
15:54meaning that they can be not only a social emotional companion for you,
15:59but they can also be a companion and say, let's cook some food together,
16:03let's go for a walk together and so on, which is going to help
16:06a lot with independent living with dignity
16:09when we are at that age.
16:11Do you think we'll see robots helping with assistive care in the future
16:15any time soon, or is this way off in the distance?
16:19I think it is within the next decade.
16:22Oh wow. David, you're working on Gemini, which is Google's answer to ChatGPT,
16:28and you aim for it to be able to do both tax returns and write a novel.
16:33People would probably be very happy to have their tax returns done or I would
16:36definitely would be.
16:37But novels, should we really be letting AI sort of take over human culture?
16:44So it's a great question.
16:46I think I wouldn't see it as taking over human culture.
16:49I think what will happen, or the most likely outcome, is that we'll be
16:53providing an incredibly powerful tool to human authors.
16:57So we've already seen this in a number of areas where
16:59we've developed technologies that enable authors of different
17:04kinds of media to basically create things much more powerfully using tools.
17:08So, for example, there's a music authoring system called Lyria
17:12that was released recently, and there's this wonderful footage of
17:16Will.i.am when he's playing with it for the first time.
17:18And he's just so excited because he says it can kind of speed up
17:22his songwriting process by, you know, 10 to 100 times.
17:25So I think, you know what I really hope we get to is a world in which the
17:29AI and the humans kind of work together to just make everything better.
17:34So, you know, I'm excited to be in a world where we have much more...
17:38the most amazing novels that we can imagine,
17:40that which go far beyond the ones we have today.
17:42Thank you, David.
17:43Thank you.
17:44This is The Engineers - Intelligent Machines from the BBC World Service.
17:48We've discussed medical AI, emotionally intelligent robots, the goal
17:52of artificial general intelligence and the threats AI might pose.
17:56Has anyone got a question on anything we've discussed so far? Wow.
18:00Okay.
18:01Pretty much everyone has a hand up, so this is going to be tricky.
18:04Could we start with the man in the red shirt?
18:11If you could stand up and say your name and your question.
18:13Thank you very much.
18:15Hello, my name's Simon. The British government wants to make Britain
18:19a leader in AI and sees the way to do this by making it a safe space.
18:23They're looking at doing that by making sure you can't develop
18:27AI where you don't understand the consequences before you develop it.
18:31Does the panel think that's the right approach?
18:34So, yes.
18:35David, what do you think about legislation around AI?
18:39I think we need some kind of regulation.
18:42I think it's, you know, an area which clearly is going
18:45to have more and more consequence and impacts on society.
18:49So I think regulation is important.
18:51I think some of the areas which have been agreed
18:55in various summits over the last year - you know, fantastic start.
19:01One thing I would say is I think it's hard sometimes
19:04to come up with like a one size fits all recipe for AI
19:09because it's so different in different areas.
19:11So you know, the regulation that you might need in medicine might look
19:14quite different to the kind of regulation you might need for, say, a chatbot.
19:19So I think, you know, we have to look at each area separately
19:23and make sure that whatever we do, you know, we really fully understand
19:26the consequences of what the impact of AI will be in that area.
19:30There's a lot of anxiety about the pace
19:32AI is moving, right?
19:34We have people resigning, leaders in the field to campaign
19:37for more safeguarding against the threat of misinformation,
19:40the threat to jobs, and even an existential risk to humanity.
19:44Regina?
19:45I actually feel quite opposite.
19:46People are really suffering. There are lots of incurable diseases.
19:49It's hard for patients, it's hard for their families.
19:52There is a lot of technologies that is out there
19:54and because we cannot get together to put the regulation in place,
19:57make, you know, the payers take part of it and find a way to bring it
20:02I think we are really making many, many people suffer.
20:08Thank you. Paolo, what about robots?
20:11Are they going to take all our jobs?
20:13Yes.
20:14Yes. Okay, great. We've got it.
20:16We finally have an answer. Thank you, Paolo.
20:20On a serious note, I feel it's a bit of a
20:23conundrum to think about legislation, because on one hand,
20:28yes, there are definitely risks and you would like to regulate it
20:31so that no one with bad intentions goes wild.
20:35On the other hand, if you think about it, it's an extremely
20:40potent technology.
20:43It could change everything
20:45in our lives including
20:48being strategically important technology
20:51to master from a national perspective.
20:55And from that perspective, if you regulate it,
21:00let's say part of the regulation is slowing it down,
21:04what are our adversaries going to do?
21:07Are they going to then have an advantage over us?
21:10So it's a bit of an arms race.
21:12And I think for that reason, practically
21:14speaking, I think it's going to be very hard
21:16to regulate it to that level unless you can have
21:20international regulation and agreement
21:23across all the powerful nations to say this is how we're going to handle it.
21:26And I haven't seen that work out very well in the history of time.
21:31Fair enough.
21:32Audience, we have time for a couple more questions.
21:34So hands up. If we could go to the man in the white top.
21:37Thank you.
21:38Hello, panel. My name's Rob.
21:42And I'm a business and sports coach.
21:45And I'm wondering whether what you've been talking about
21:49is possible for helping humans to improve their performance
21:53at a sport.
21:55Good question.
21:57Who's our greatest sport enthusiast?
21:59Perhaps this one is for you, David.
22:01Can we be using AI to improve sports performance?
22:06It's a great question.
22:07There's a lot of really amazing research that's happening to try and do just that.
22:11You know, one thing which we've been doing at Google DeepMind is actually
22:15a collaboration with Liverpool Football Club to try and help them improve
22:19their tactics.
22:22So that's one example.
22:24I think, you know, the amazing thing about sports is it's become over time
22:29so refined in terms of the particular approaches
22:32that people take that actually, you know, really they're very open
22:36to new discoveries and new ways to do things.
22:40And so it's been really fun actually, just watching that kind of thing unfold.
22:43Very cool. Thank you, David.
22:45I can see we've got lots of young people in the audience, some teenagers even.
22:49Does any of our younger audience have a question?
22:52If so stick your hand up.
22:53There was a young lady who had a hand up here
22:56with a ponytail. Yes.
22:59As AI develops, I reckon
23:01that humans probably depend more on AI and maybe learn less.
23:06Is there anything we could do to maybe make sure that as it
23:11develops, humans still develop and learn?
23:14Ooh, good question.
23:15So will we stop learning, as AI does the learning instead for us?
23:20I'm going to put that to all of you if that's all right. What do you think, David?
23:23I'd like to imagine a world where we have an AGI
23:26which is like a personal friend, assistant teacher,
23:31and everything we do understands what we want to learn, and it knows
23:34just how to teach us and help us to learn more and more and more and more.
23:37Regina, what do you think?
23:38Do you think we're just going to end up relying on
23:40AI for everything?
23:42As a non-native speaker of English
23:43I remember as a young professor I spent a humongous amount of time
23:46like reading my papers and making sure, you know, because in my native language
23:51we don't have
23:52and so I make a lot of mistakes when I write.
23:54I see now how I write papers.
23:56Actually, it removes a lot of my pressure in writing
23:59and I can really focus on ideas rather than doing the small things.
24:03So I hope that we will find a symbiosis.
24:06We don't really, you know, remove our basic skills,
24:11but we can just do more and focus on the things that we can do better.
24:16I would like to imagine if Isaac Newton or Albert
24:20Einstein had access to these tools today, I mean, as prolific
24:24as they were at a time where we didn't even have calculators,
24:27imagine what they would have done and the impact it would have had on the world
24:31if they had access to these tools. In the next five years
24:34potentially, you don't need to code.
24:36You will just tell
24:38your favourite AI system and say I want a code that does x, y, z
24:42and you can accomplish what would take years of many people in hours today.
24:49So it will make you more prolific.
24:51Thank you.
24:51I think it's fair to say all three of you have given us hope for the future.
24:55Audience, thank you so much for your questions.
24:57I wish we could take more, but I'm afraid we're out of time.
25:01That's it for The Engineers -
25:02Intelligent Machines at Imperial College, London.
25:05I'm Caroline Steel.
25:06On behalf of the BBC World Service
25:08our partners
25:09the Royal Commission for the Exhibition of 1851 and my producer, Charlie Taylor.
25:14Please join me in giving a warm round of applause for our brilliant pioneering
25:18AI engineers Regina Barzilay, Paolo Pirjanian and David Silver

About this transcript

This page contains the full transcript of What do tech pioneers think about the AI revolution? - The Engineers, BBC World Service by BBC World Service, generated from the public captions YouTube serves with the video. The transcript has 4,289 words across 458 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.