- 3 minutes ago
At the India AI Impact Summit, Yann LeCun, Executive Chairman, AMI Labs, and Amanda Brock, OpenUK CEO, shared insights on the future of artificial intelligence.
Category
🗞
NewsTranscript
00:00So let's begin with a big idea here, Professor Lacoon.
00:04Are we on a path to creating the smartest mind that humanity has ever known?
00:12And will that happen in our lifetime?
00:17Maybe in the lifetime of some people here, possibly not in mine.
00:22We'll see. It will take a while.
00:25But I think the more interesting thing that we're going to build is an amplifier for human intelligence.
00:32So maybe not an entity that surpasses human intelligence in all domain, although that will happen at some point.
00:41But it is something that will amplify human intelligence in ways that will accelerate progress.
00:47So then will we end up defining and redefining genius?
00:52Genius. What will a genius be?
00:56Well, you know, I think several thousand years ago or even a few centuries ago, what people identified as genius
01:03is very different from what we currently identify as genius.
01:06And I think there will be more evolution of that concept of genius.
01:15You know, in the past, perhaps, you know, genius was, you know, some act of creation or invention, but maybe
01:22not at a theoretical level like we tend to think of it today.
01:25It was, you know, it was more practical, certainly in the very ancient past, people who figured out how to
01:31cultivate crops or domesticate animals probably were seen as genius.
01:37So, you know, we have often seen, and this is a thought that you have all, you know, pretty openly
01:43shared, that AI is powerful, but not intelligent.
01:49When we make that distinction and there are conversations around LLM, where do you see intelligence and AI driven power?
01:58Yeah, I think there's a lot of confusion, really, because we tend to anthropomorphize systems that can reproduce certain human
02:08functions.
02:09So what's, I mean, LLMs are incredibly useful.
02:13There's no question about that.
02:15And they do amplify human intelligence, like computer technology going back to the 1940s.
02:22But LLMs, to some extent, except for a few domains, are mostly information retrieval systems.
02:29They can compress a lot of factual knowledge that has been previously produced by humans and can give easy access
02:36to it.
02:37In a way, it's kind of a natural evolution of, you know, the printing press, the libraries, the internet, and
02:43search engines, right?
02:44It's just a more efficient way to access information.
02:48And there are a few domains where the intelligent capabilities of those systems actually is more than that.
02:57It's more than just retrieval.
02:59So for generating code, maybe doing some type of mathematics, you know, we're getting the impression that it's beyond this.
03:08But it's still, to a large extent, domains where reasoning has to do with manipulating symbols.
03:15The problem is that, you know, why do we have systems that can pass the bar exam and win mathematics,
03:22Olympiads, but we don't have domestic robots.
03:24We don't even have cell-driving cars.
03:26And we certainly do not have cell-driving cars that can teach themselves to drive in 20 hours of practice
03:32like any 17-year-old.
03:33So we're missing something big still.
03:36So what are we teaching a 17-year-old then?
03:39Well, so the question is, how does a baby learn, or even an animal, right?
03:46Animals have a much better understanding of the physical world than any AI systems that we have today, which is
03:51why, you know, we don't have smart robots.
03:55And so, you know, we learn about the world, how the world works, mostly by observation when we are babies,
04:04a few months old.
04:05And then we learn by interaction.
04:07And we learn mental models of the world that allows us to apprehend any new situation, even if we haven't
04:14been, you know, exposed to it beforehand, we can still handle it.
04:18So a big buzzword in AI today is world models.
04:22And this is really this idea that we develop mental models of the world that allow us to think ahead,
04:32to apprehend new situations, plan sequence of actions, reason, and predict the consequences of our actions, which is absolutely critical.
04:41And LLMs don't do this, really.
04:42There is the sense, Professor, that perhaps AI will unlock an era of radical abundance.
04:53Will this abundance benefit us?
04:56Well, so, you know, if you talk to economists, they tell you, you know, if we can measure the improvement
05:05that AI will bring to productivity, which is, you know, amount of wealth produced per hour worked.
05:12It's going to add up to maybe 0.6% per year.
05:16Okay, this is from economists that actually have studied the effect of technological revolutions on the labor market and the
05:23economy.
05:24People like Philippe Aguillon or Eric Brynjofson.
05:28And so that seems small.
05:31It's actually quite big.
05:32And, you know, it's certainly going to accelerate scientific progress, progress in medicine.
05:37I do not believe there's going to be, like, a singular identifiable point where the economy is going to take
05:44off and there's going to be abundance.
05:47And there's also the question of the, you know, policies surrounding this, you know, are those benefits going to be
05:53shared across humanity or, you know, different categories of people in various countries?
05:59So that's a political question.
06:00It has nothing to do with technology.
06:02So if it's about upskilling and ensuring that you're relevant, then only perhaps you're intelligent.
06:08Will that then mean that the countries that adopt AI and the, you know, the pace at which India and
06:16the scale at which India has adopted AI,
06:18the challenge would be to create talent which is upskilled and reskilled and have the required skills for this?
06:27Absolutely.
06:29So the relationship that we're going to have with intelligent AI systems is going to be similar to the relationship
06:37that a leader in business politics or academia or some other domain has with their staff.
06:48Okay.
06:49AI is going to be our staff.
06:50Every one of us is going to be a manager of a staff of, you know, intelligent machines.
06:57They'll do our bidding.
06:58They might be smarter than us.
07:00But, you know, certainly if you are an academic or a politician, you work with staff that are smarter than
07:05you.
07:06In fact, that's the whole point.
07:07You need to, you know, attract people who are smarter than you because that's what makes you more productive, you
07:13know.
07:13So for an academic, it's students who are smarter than their professor and teach, you know, it's not the professor
07:19that teaches graduate students.
07:21It's the other way around, actually.
07:23And certainly we have a lot of examples of politicians who are surrounded by people who are smarter than them.
07:30Earlier today, when Prime Minister Modi addressed the gathering, he said that India doesn't fear AI.
07:37We are seeing this as our destiny future, which is bhaagya.
07:41Do you see that with a summit of this nature being hosted in India, it's a message to the global
07:50south?
07:50And that's where perhaps the next big innovation in AI could be coming from.
07:58Well, long term, it's going to come from countries that have favorable demographics.
08:04And that means India, Africa.
08:08You know, the youth is the most creative part of humanity.
08:11And there's sort of a deficit of that in the north, largely.
08:19So, you know, the scientists, the top scientists of the future, in fact, many of the present are from India
08:27and in the future will be from mostly Africa.
08:32So what does that mean, though?
08:36Right.
08:36It means having incentives for young people to kind of study, first of all.
08:44So the idea that somehow we don't need to study anymore because AI is going to do it for us.
08:49And, you know, that's completely false.
08:51Absolutely, completely false.
08:53And it's not because I'm a professor, OK, that I'm saying this.
08:57On the contrary, we're going to have to study more.
09:00We see, for example, a trend where in industry in the past, in certain countries, it's certainly true for India,
09:06but it's also true in European countries.
09:09And certainly in the U.S., we see more demand for people with more education at the PhD level, for
09:17example.
09:17The number, the demand for PhD level scientists in industry has grown in the last 15 years, in part because
09:28of AI, but because of everything, because technological progress and, you know, hinges on scientific progress.
09:35And scientific progress is brought about by scientists and, you know, scientists mostly have done PhDs.
09:40And so there is more demand for education, not less.
09:44And so for countries in the global South, that means, you know, investing in education and youth.
09:53But when you say about education, will AI assist education in terms of making students or youth of the country
10:02more literate or will they become more AI dependent?
10:08Well, I mean, we're dependent on technology, right?
10:11I'm dependent on this pair of glasses, otherwise I don't see you.
10:14So that has been with us for centuries.
10:18Yeah, we'd be dependent on AI, of course, but AI will facilitate access to knowledge and thereby kind of be
10:27a tool for education.
10:29I think the effect on society, you know, might be extrapolated from what was observed in the 15th century when
10:37the printing press started enabling the production of printed matters and the dissemination of knowledge, right?
10:46It had a huge effect on society worldwide, at least in countries that allowed it to flourish.
10:54And I think it's going to be a similar transformation with AI, of course, in the modern world, just more
11:00access to knowledge.
11:02The Internet played also a similar role.
11:04And, you know, I think this is just going to make people more informed, smarter, able to make a more
11:11rational decision if it's deployed in the proper way.
11:14So if you were to define this moment, which we are witnessing in history, we are living it.
11:20How will you say it?
11:22Is it like the advent of electricity?
11:25Yeah, people have made that claim.
11:27They have made that claim, yes.
11:28Including economists.
11:29It's the new electricity.
11:31I think it's more like the new printing press, really.
11:33But, again, in the sort of this vision of, you know, kind of more dissemination and sharing of knowledge and
11:40amplification of human intelligence.
11:45But, you know, the impact on society and the way countries need to be run is very difficult to predict
11:58at this point.
11:59I'm sort of an optimist in the sense that, you know, I think societies would figure out how best to
12:06use the technology for the benefits of their population.
12:11While I am an optimist, nevertheless, I'm going to ask this question to you, Professor.
12:16Are we overestimating the change or underestimating what has struck us?
12:23So usually in technological shifts of this type, we are overestimating the changes in the short term and underestimating them
12:31in the long term.
12:34I think for AI, it's a little bit different because there's been a huge amount of hype and expectations that,
12:40you know, the transition to human level AI, superhuman level AI is going to be an event and is going
12:46to happen within the next few years.
12:48And people have been making that claim for the last 15 years and it's been false.
12:52In fact, they've been making it for the last 60 years or 70 years and it's been false.
12:56Every time in the history of AI that scientists have discovered kind of a new paradigm on AI, of AI,
13:03how you build intelligent machines.
13:05People have claimed, you know, within 10 years, the smartest entity on the planet will be a computer.
13:11And that just proved to be wrong, you know, four or five times in the last 70 years.
13:16It's still wrong.
13:17We're still very far from that.
13:19We're not very far, we're getting close, right?
13:21We're seeing the end of the tunnel, but it's not like, you know, we're going to have super intelligent systems
13:27within two years.
13:28It's just not happening because of this gap, right?
13:31You know, where is the robot that can learn to drive in, you know, in 20 hours of practice like
13:36a 17-year-old?
13:37Even though we have millions of hours of training data of people driving cars around, we should be able to
13:42train an AI system to just imitate that.
13:44That doesn't actually quite work.
13:46It's not reliable enough.
13:47Okay, so let's try and wrap up this conversation with who gets to define intelligence now onwards.
13:54Will it be actually humans, machines, or both together?
13:59Probably both together, but mostly humans.
14:02We set the agenda, and the biggest difficulty is not to get fooled into thinking that a computer system is
14:15intelligent simply because it can manipulate language.
14:18We tend to think of language as the epitome of human intelligence, right?
14:23But in fact, it turns out language is easy to deal with because language is really a sequence of discrete
14:32symbols of which there is only a finite number.
14:35And that turns out to make things easy when you train a system to predict what the next word is
14:40in a text, which is what LLMs are based on.
14:43It turns out the real world is much, much more complicated, and it's been known in computer science for many
14:49years.
14:49It's called the Moravec Paradox, after a roboticist, Hans Moravec.
14:54And so the company I'm building and the research program I've been working on for the last 15 years or
15:00so is intelligence for the real world.
15:03You know, how to deal with high-dimensional, continuous, noisy signal that the real world is, which your house cat
15:10is perfectly able to deal with, or a squirrel, or whatever, but not computers yet.
15:17That's the big challenge for the next few years in AI, dealing with the real world.
15:22And that's the point of the company I'm building.
15:24So AI has to deal with the real world, or real world has to deal with AI?
15:28AI has to deal with the real world.
15:30The messiness of the real world, the unpredictability of the real world.
15:35All right.
15:36Thank you so much for this conversation, Professor.
15:45Amanda Brock, she is the CEO of Open UK, whose famous pizza analogy has now gone viral.
15:53Before I bring in the pizza analogy, how has your stay been in India so far?
15:59And tell us about the India AI Summit.
16:03The fact that this is titled as Impact Summit.
16:07So certainly there is the sense that we have moved ahead from action to impact.
16:13So I've been here since Sunday.
16:15It feels like a lot longer, as we were just discussing.
16:18It's been an incredibly busy week.
16:20Very, very long days, starting early in the morning, going late at night, meeting after meeting.
16:25So many discussions going on.
16:26So much activity and so much energy here.
16:30I didn't know the pizza analogy that you knew what that was.
16:33So, okay.
16:35So the pizza analogy essentially is that topping get all the glory, but the base is everything.
16:41That's right.
16:42So how do we finally pay and sustain the open source maintainers?
16:48So I think we need to see a shift.
16:51And there's been a lot of talk across the summit about digital public good, about public good, about civic tech.
16:56And I think we need to be very clear on the meanings of things.
16:59And what we're really talking about is people's gift to society when they contribute this open source, whether it's software,
17:06whether it's AI that builds that infrastructure for us.
17:09And I think we need to see a shift in governments and a recognition from governments that they are benefiting
17:15from the work of those individuals.
17:17And we need to see a contribution back from every nation of a percentage of its innovation fund.
17:23This is what we've recommended in the UK.
17:25And that they then collaborate to make sure that the money gets spread across everybody that needs it.
17:30So when you say about AI must serve humanity, okay, and not the reverse, the humanity should not serve AI,
17:40then what urgent policy or funding intervention do you think should come from Global South?
17:48Oh, that's interesting.
17:49Specifically from the Global South.
17:51I think you need to start by learning from our mistakes.
17:55And we've tried to do things like have an uneven playing field where open source was given advantages and it
18:01doesn't work.
18:02Open source is strong enough that it can be on the same footing, the same level playing field as other
18:06things.
18:07I think you need to look at what we've done, how we've got here, and then think, how can you
18:12do better?
18:13They were talking in the plenaries this morning about UPI and how that leapfrogged.
18:18And I think that leapfrog is something that you can do again and again with foresight.
18:23I think you have an enormous population of innovators here.
18:27You have so many skilled individuals.
18:29The research we did, I think it was something like 25 million off the top of my head.
18:33And those people have...
18:3425 million innovators in India?
18:3625 million who have GitHub accounts.
18:38Double-check my figure for you, but 25 million who have GitHub accounts and who collaboratively even engage in technology.
18:46Now, those people make you the second biggest in the world in terms of collaborative innovation, but you're not leading.
18:53And I think you have the skills to lead.
18:55You just have to look ahead and work out how to do it.
18:59Why do you think we are not leading?
19:01Is it also because we are starting perhaps late?
19:05No, I don't think so.
19:06I think you are not leading because most countries don't recognize the people who do the work because the U
19:14.S. has dominated it so much.
19:16We talk about a submarine in the U.K.
19:18We talk about the yellow submarine, the Beatles, and we talk about a submarine under our digital economy.
19:24And in that submarine, we say we've got all our innovators because they all collaborate with the U.S.
19:28and you don't see them in your home country, in your home geography.
19:31India is one of the few countries that's of the same sort of scale as the U.K.
19:35So we don't have anything like the level of people you have, obviously.
19:39We're a tiny country.
19:40But we have 7.1% of our population doing that, which is higher than you per capita.
19:46So we're two similar nations where we have this massive innovation that's unseen because it's globally collaborative.
19:52You know, since you're talking about that act, we have seen now what is being termed as New Delhi Frontier.
20:02And that's essentially the nature of commitments that New Delhi AI Impact Summit is making.
20:07And it's also about the outcomes.
20:09What they are whooping certainly is that this will ensure some kind of evaluation of multilingual capabilities on a subset
20:19of languages.
20:21Do you think that hyper-local focus is the way forward?
20:24We have a strapline at Open UK, which is UK leadership, global collaboration.
20:30And I think you have to do something very similar, which is focus locally, build what you need locally.
20:36And then with those languages, you're serving a huge population outside India as well, even with those languages.
20:41But I think you have to take that local focus and then take it out to the world and have
20:46a global perspective.
20:46So it's a balance.
20:47So just some time back, I questioned Professor Jan Lekoon and I had asked him.
20:54I met him.
20:55I met him this week.
20:56I wanted to, you know, find an answer to a question I'm very curious about.
21:02That are we in the process of perhaps creating the smartest mind that all of us know about?
21:10And what will be your response?
21:12Depends who you ask.
21:13I don't know if I knew I'd probably be worth a lot of money, like he is.
21:18It's a process.
21:20And none of us know when it's going to speed up and slow down.
21:23The things that I'm seeing come to market now, you're nowhere near it.
21:26But then we don't know what's going to happen tomorrow.
21:28So nobody can give you a concrete answer.
21:30So are we living in the times that which are still uncertain with regards to AI, a kind of a
21:37bubble or a hype?
21:38Or are you of the opinion that it's so real and unpredictable at the same time?
21:44I think there's a piece where, yes, it is unpredictable.
21:46Never say never.
21:47But the reality of what we have right now is that we have productivity tools.
21:52We have things that can make us better at jobs, make our jobs easier if we train them appropriately.
21:57You know, that garbage in, garbage out.
21:58There's a lot of work to do to get an agent to do what you want it to do the
22:01way you want it.
22:03But I think if you get through that, then you have the potential to do well from AI.
22:10But it's mundane tasks that it's going to do for you.
22:13It's not brilliant.
22:14Amanda Brock, really appreciate your time and thank you for this conversation.
Comments