5:40
There will be no apocalypse in the labor market due to AI. This idea of 50% unemployment, maybe an uprising in the streets, I don't know. This.I don't see any realistic way to do this. In fact, early data, even from some organizations that have been inflating scenarios of 50% unemployment, shows that AI is not having the negative impact on jobs that everyone feared. Since AI automates only a portion of someone's work, it turns out that we still really need humans to do the rest of the tasks. And if what's happening in software development is an indicator of what's to come in other sectors, I think it's going to be great. On the contrary, we can't find enough developers right now, and I can already see that we lack marketers, sales specialists, or operations
managers who work with AI. Ahem. But the challenge is upskilling, because for all these roles we need to help people acquire new skills, and I do n't just mean AI skills. But it turns out that. maybe I'll give an example from software development. Many people were, and perhaps still are, front-end developers, meaning they create a website or app to make it look good. But AI is doing a good job with front-end development. So many front-end developers are now becoming full-stack developers: they not only create what the user sees, but also what runs in the background. They create mobile applications and deal with security issues. Therefore, many developers are now taking on a wider range of tasks in their work. This means that instead of just
focusing on the visual part, they have to learn how to build all these other elements as well. And they have to master the skills of working with AI. When they do this, they become extremely valuable in the job market. We just can't find enough people like that. I see the first signs that a similar transformation is taking place in other professional fields. I'm curious, though: you say there I'm curious what this means for your vision for the further progress of AI. Because the fear of the apocalypse is not about the AI we have in 2026, but the AI we will have in 2030, or, if your predictions are short-term, the AI we will have in 6 months, right? Few would argue that modern AI can perform any task in a field like
software development, but the fear is that if we keep this pace of progress going for another year or two, it will be hard to predict where it will stop. Are you secretly saying that you think he will stop? That we will hit a hard limit to some of your scaling laws, or some architectural changes will be needed, and there will be tasks that AI systems of the future simply cannot perform, and probably never will. Well, I don't think we'll hit a hard limit, but progress will be gradual for a long time to come. So I would really like AI to be able to perform any intellectual task that a human is capable of. Then we'll go on vacation or something. But realistically, this
will require numerous groundbreaking breakthroughs, and I would like to see it happen in our lifetime. But whether it will take decades, or even more, I don't know. Um, and until AI can do 100% of what humans can, which seems like an incredibly distant future. I don't think we'll get to that anytime soon. Of course, not in 6 months. It is almost highly unlikely that we will achieve this by 2030. There is still a lot of work for people. And you used the term " cumulative progress". One of the hype terms in Silicon Valley was cumulative progress or recursive improvements and so on. And at least right now it's more hype than reality. For decades, computers have been making computers better. I mean, we used the last generation of semiconductor microchips to
build the current generation. And yes, in leading AI labs, they write all the code, or almost all the code. So, of course, AI helps us create better AI. But it turns out that if we look at the bottlenecks in AI progress, a lot of them are that we need to build new GPU servers. We need to build power plants. Um, we need to do a lot of experiments to see what works. So, unfortunately, even superintelligent AI, if we ever get to it, must obey the laws of physics. And the bottlenecks are the time to build data centers and run experiments to figure out what to do, rather than just, you know, thinking in the box. Um, then these are fundamental bottlenecks, because the barrier to greater intelligence is not simply getting smarter at something.
There are other bottlenecks that seem very difficult to overcome. If you're not worried about the job market apocalypse, are you worried about any of the narrower implications? I think primarily about the impact on the early stages of a career. You mentioned your colleague at Stanford, Brynjolfsson, who did a great study last year. It shows that the more a sector depends on AI, the worse the employment prospects for those who enter it. I found this research very compelling, and it seems to be borne out in practice: getting an entry-level software engineer job without skills and experience becomes quite difficult, because you can't offer much, as you used to in roles where you just write code, learning for a few years.
Yes, Eric's article was very informative. I think he has new data that paints a more optimistic picture. But, again taking software development as an example, I believe that the problem is not the lack of vacancies. The problem is the slowness of our education system, which cannot adapt to changes to prepare specialists for jobs that are currently in demand. Therefore, it is very unfortunate that today, in many universities, you can still get a bachelor's degree in computer science without ever making a call to the API of an advanced AI model. And today I wouldn't hire a software engineer who's never done this. At some universities, you can still earn a bachelor's degree in computer science without using AI agents to write code. And, frankly, no
one on my team would hire a developer who doesn't know how to use an AI agent for programming. So for me the problem is not the jobs apocalypse." We don't need junior specialists. Last summer, my team had many interns, including high school students, college students, and graduates. And they were wonderful. They did a lot of work. I'm so glad I had these interns. But the secret is that I only hire those for whom AI is their native environment. So there are jobs, but we need to fix the education system.
