0:00 You may not take interest in politics, but politics will take interest in you. So the same applies to AI many times over. >> Ilas Sutskver, the man behind the invention of open AI, gave a pretty strong speech at the University of Toronto. He expressed great concerns about the upcoming AI, which might disrupt our entire world. Watch this. >> The reason it's not going to be the most conventional convocation speech is because there is something a little different going on right now. You all leave, we all leave in the most unusual time ever. And this is something that people might say often, but I think it's actually true this time. And the reason it's true this time is because of AI, right? Obviously, I mean, from what
0:46 I hear, the AI of today has already changed what it means to be a student by a pretty considerable degree. That's I uh especially I that's what I I sense and I think it's true. But of course the impact of AI goes beyond that. What happens to the kind of work we do? Well, it's starting to change a little bit in some unknown and unpredictable ways. And some some some work may feel it sooner. Some work might feel it later. With today's AI, you can go on on uh on Twitter and you can look at what AI can do and what people say and you might feel a little bit of that. You wonder,
1:31 hey, which skills are useful? Which ones will be less useful? So, you got these questions going on. And so, you can say that the current level of challenge is how will it affect work and our careers. But the thing the real challenge with AI is that is really unprecedented and really extreme and it's going to be very different in the future compared to the way it is today. Like you know we've all seen AI, we've all spoken to a computer and a computer has spoken back to us which is a new thing. Computers would not do this in the past but now they do. So you speak to a computer and it understands you and it speaks back to you and it also does it in voice and it writes some code. It's it's pretty
2:17 crazy, but there are so many things it cannot do as well and it's so deficient. So, you can say it still needs to catch up on a lot of things, but it's evocative. It's good enough that you can ask yourself, you could imagine, okay, fine, in some number of years, some people say it's in three, some people say it's in five, 10. Numbers are being thrown around. It's a bit hard to predict the future, but slowly but surely or maybe not so slowly, AI will keep getting better. And the day will come when AI will do all of our all the things that we can do. Not just some of them, but all of them. Anything which I can learn, anything which any any one of you can learn, the AI could do as well. How do we know
3:04 this? By the way, how can I be so sure? How can I be so sure of that? The reason is that all of us have a brain and the brain is a biological computer. That's why we have a brain. The brain is a biological computer. So why can't a digital computer, a digital brain do the same things? This is the one sentence summary for why AI will be able to do all those things because we have a brain and a brain is a biological computer. And so you can start asking yourselves what's going to happen. What's going to happen when computers can do all of our jobs? Right? Those are really big questions. Those are dramatic questions. And right now, like you start thinking about it a little bit, you go, gosh, that's a little intense. But it's actually only part of the intensity because what's going to happen?
3:51 What what will we the collective V want to use these AIs for? Do more work, grow the economy, do R&D, do AI research. So then the rate of progress will become really extremely fast for some time at least. These are such extreme things. These are such unimaginable things. So right now I'm trying to pull you into that a little bit into this headsp space of this really extreme and radical future that the AI creates. But it's also very difficult to imagine. It's very very difficult to imagine. It's very difficult to internalize and to really believe on an emotional level. Even I struggle with it. And yet the logic seems to dictate that this very likely should happen.
4:37 So what does one do in such a world? You know there is a quote which is like this uh uh which goes like this. It says you may not take interest in politics but politics will take interest in you. So the same applies to AI many times over. And in particular, I think that by simply using AI and looking at what the best AI of today can do, you get an intuition. You get an intuition. And as AI continues to improve in one year, in two years, in three years, the intuition will become stronger. And a lot of the things that you're talking about now, they will become much more real. they'll
5:23 become less imaginary. In the end of the day, no amount of essays and and explanations can can compete with what we see with our own senses, with our own two eyes. And especially with AI, the very smart, super intelligent AI in the future, there will be very profound issues about making sure that they are they say what they say and not pretend to be something else. And I'm really condensing a lot into a small amount of information here in time here. But overall, by simply looking at what AI can do, not ignoring it when the time comes, that will generate the energy that's required to overcome the huge challenge that AI will pose. And the challenge that AI poses in
6:09 some sense is the greatest challenge of humanity ever. and overcoming it will also have the will also bring the greatest reward and in some sense whether you like it or not your life is going to be affected by AI to a great extent and so looking at it paying attention and then generating the energy to solve the problems that will come up that's going to be the main thing >> so that was Ilia's view but to understand the full debate I want you to watch this interview clip of Eric Schmidt where he talks about a much broader impact of AI on human lives in the coming years. Watch this. >> Okay. So, we believe as an industry that in the next one year, the vast majority of programmers will be replaced by AI
6:56 programmers. We also believe that within one year, you will have graduate level mathematicians that are at the tippy top of graduate math programs. There's lots of reasons to think this is going to happen. This is the consensus. You go, okay, well, that's pretty interesting. Now, I can't do that kind of math. Very few people can do that math. How can the computer do that math better than anybody else? To some degree, it's because math has a simpler language than human language. So, the way these algorithms actually work is they're doing essentially word prediction. So, you take you take a a sentence, you take a word out, and then it learns how to put the correct word back in. This is called the loss function, and it's optimized to do that at a scale that's unimaginable to us as humans.
7:42 So you do the same thing for math, but there you use a conjecture and then a proof format through a protocol called lean. In programming, it's pretty simple. You just keep writing code until you pass the programming test. So strangely, the first question I always ask programmers is what language do you program in? And the correct answer is it doesn't matter because you're trying to design for an outcome. You don't care what code is generated by the computer. This is a whole new world. Okay. So that's one year. Okay, what happens in two years? Well, I've just told you about reasoning and I've told you about programming and I've told you about math. Programming plus math are the basis of sort of our whole digital world. So, the evidence and the claims from the research groups in OpenAI and
8:28 and anthropic and so forth is that they're now somewhere around 10 or 20% of the code that they're developing in their research programs is being generated by the computer. That's called recursive self-improvement is the technical term. So what happens when this thing starts to scale? Well, a lot. One way to say this is that within three to five years, we'll have what is called general intelligence, AGI, which can be defined as a system that is as smart as the smartest mathematician, physicist, you know, artist, writer, thinker, politician, maybe not in the same level. Um, but you get the idea. Uh, just the creative industries and so forth. But
9:14 imagine that in one computer. Okay. Well, that's pretty interesting. I call this, by the way, the San Francisco consensus because everyone who believes this is in San Francisco. It may be the water. What happens when every single one of us has the equivalent of the smartest human on every problem in our pocket? So, it means you have the best architect when you have an architecture problem. Another thing that's going on is the development of agentic solutions and agents are refer to systems that have input and output in memory and they learn. An example here is that I want to uh buy another house. Uh I happen to like Virginia. I grew up in Virginia. I say find me a house in the greater MLAN area. Look at the that's one agent. Look at all the rules. Figure out how big a house I can build. That's another agent.
10:02 Do the transaction to buy the land. That's another agent. design the house with a human architect, right? But sort of ignore them for most of the thing, but they have to sign it off and then I approve it and then find the contractor, right? Hire the contractor, pay the bills, and at the end sue the contractor for lack of performance. Okay? Now, I just gave you the stupidest possible explanation. I just described every business process, every government process, and every and every sort of academic process in our nation. >> So, it isn't just the programmers that are going to be out of work. We're all going to be out of work. >> No, that's not a consequence. I'll come to that. But, but the reason I want to I want to make the point here is that in the next year or two, this foundation is
10:49 being locked in and it's not we're not going to stop it. gets much more interesting after that because remember the computers are now doing self-improvement. They're learning how to plan and they don't have to listen to us anymore. We call that super intelligence or ASI artificial super intelligence. And this is the theory that there will be computers that are smarter than the sum of humans. The San Francisco convent consensus is this occurs within six years just based on scaling. Now, in order to pull this off, you have to have an enormous amount of power. I was here yesterday testifying about this, you know, and we need like I
11:34 can talk at some length about how many gigawatts and how many nuclear power plants and all the kind of stuff we can talk about separately. This path is not understood in our society. There's no language for what happens with the arrival of this. I wrote a book on this with Henry Kissinger called Genesis which you know I recommend obviously um because I wrote it >> available >> available available in your usual places um but the important point is this is happening faster than our human that our society our democracy our laws will address and there's lots of implications that's why it's underhyped people do not understand what happens when you have intelligence at this level which is largely free that's the How do we get
12:20 ready for it? >> Well, we start by talking about it. And by the way, on the jobs thing, everyone assumes that automation will replace will eliminate jobs. If you look at the history of automation ever since the the looms and uh in uh 300 years ago, the jobs are changed, but more jobs are created than destroyed. In this case, you'd have to convince me that this time is different. If you look in Asia where they for whatever reason are choosing not to have children, the Asian reproduction rate is in the order of 1.0 or lower. So they're rapidly disappearing. So the Asian countries are very very quickly automating. The tools that I'm describing will allow the few humans that will be working very hard in
13:08 30 or 40 years. If these trends continue, the rest of us will be dependent on those hardworking humans. It'll make their productivity more much greater. >> Now, here's another clip of Eric Schmidt. Here he shares much deeper concerns about upcoming AI technology. Watch this. >> One way to think about the AI that you all know is think of it as language to language. You ask a question, the answer comes back. You ask a question, it can even write code. Nowadays, the models are multimodal. So, for example, you can take a picture and say, tell me what's in this picture. uh technically there are APIs which allow uh one firm to call an open AAI or Gemini API or anthropic for etc and do
13:53 the classification of the picture and so forth. These are all tactics that increase the intelligence of the underlying system. There are three things going on right now this year. So less than the time frame you gave are really interesting. One is called infinite context windows. Infinite context men windows means that you can keep feeding the answer back in is the question. So it allows you to do stepby-step planning. You know, how do I uh how do I build a house? Well, the first is I have to find a contractor. I found a contractor. What do I have to talk to them? Then I have to have an architect. How do I find an architect? Then I have to tell the architect what to do and then design me the house. I'll give it to the architect. He can redesign it. You know, it's a series of steps. Uh the next one are called agents. And agents is a generally overused term
14:41 and most people think that agents will essentially act as memory sources. So an agent can be understood as it's watching something and when it sees it, it takes an action. It does that by knowing what to do based on what it's seen. The specs for how agents work are completely undefined in the industry. The dominant companies want to have their own agents and they don't want the agents to interact because they want control for obvious reasons. Many people think that there will be an agent store that you will download like we see with apps but not this year. And the third one is text to text to code. Now I don't know about you all but I've programmed and managed programmers for more than 40 years and they never do what I want. So can you imagine if the computer you said write me a program to do this and it actually
15:27 writes the code. In our case, uh the program would be search through all the literature, find out who is working on energy policy, who has a technological background or a role in which they have to be technologically liberate literate. Identify them, rank them, score them based on whatever our goal is. Um and and then automatically invit send them an invitation. If they say yes, say congratulations. If they say no, why not? and call them and with a synthetic voice tell them that they're idiots for not coming. That's the kind of program I would write. Thank god I'm not doing that. But but you see how easy it would be to automate tasks. So that's I think the first step. The next step is not as clear. There are uh there's sort
16:12 of huge contest um there's a huge set of contests going on now which are at a scale that's unimaginable. You have the big three in the US. Anthropic uh which is allied with Amazon, Gemini obviously from Google, OpenAI, Microsoft and let's assume they all do really well. It looks that they're doing really well. I can talk about what their problems are but fundamentally they're they're doing well. You have Facebook which has chosen open- source path for the 400 billion model that has a lot of implications strategically, right? Which we can discuss. um all of these are vying for the best reasoning, the best answers and then the best predictive analytics, the best image classifiers and the best multimodal. That technology then diffuses or the technical term is distilled into more
16:58 specialized models. And I think that's the action you'll see in the next one to two years. You did not mention artificial general intelligence. First, for those of us who aren't necessarily um totally up to speed on AI, what is it and where are we? >> There are multiple definitions of AGI, but the it's the term has been around for 15 years. The basic idea is what is the point where you have the flexibility of a human in your intelligence system. So, one way to understand it today is that we these are called narrow AI approaches, although they're not certainly not narrow. you basically they're they're initiated by a human at what point is the question can the computer generate its own objective function its own goal and how will that
17:43 emerge uh the there's what I call the San Francisco school because they're all in San Francisco uh which is a separate set of issues and they all talk to each other and they've all convinced themselves that if within two to three cranks of the systems the crank is about 18 months you get to AGI And they define AGI as intelligence greater than the sum of human intelligence. I personally think that that's likely but not in three years, not in >> what is the time frame, do you? We don't know.
