AI beyond chatbots in 2026: Stanford Computer Science’s Jure Leskovec

? (?) 00:00.000

All right. So, one of the things that you say you're you're really amazed about in 2025 is the advancement of AI and some of these chatbots. I have to ask you, what's the benchmark? What were you expecting?

Jure Leskovec (Computer Science Professor) 00:10.320

I think the the speed of how quickly AI research and advancements is going moving forward is just unprecedented. And uh every year we think we'll hit some wall but something else opens up. I think in 2025 the the big breakthrough was in terms of reasoning and allowing AI to

Jure Leskovec (Computer Science Professor) 00:29.520

reason autonomously over multiple steps and self-correct. Um and you know, I'm very excited about what's going to come in 2026.

? (?) 00:37.800

All right. So Yuri, as we look at 2025, one of the big stories were chips. Whether it was Nvidia chips and they had a mode or other companies like Alphabet coming out with their TPU's, what are you expecting in the new year? Especially as this morning we're looking at a deal

? (?) 00:49.960

where SoftBank uh is reportedly looking to buy a private equity company really focused on data centers.

Jure Leskovec (Computer Science Professor) 00:57.120

I think what the demand for AI is huge. and the computational complexity and computational needs for these models is humongous. So it's really all about bringing down the bringing down the cost, bringing the down the cost of inference and the new chips, new data centers and new

Jure Leskovec (Computer Science Professor) 01:14.000

algorithmic techniques are key here. That's why we see Nvidia moving in with partnership with with Grock as well as other companies because they are realizing AI is going to be the next productivity layer and making it cheap, making it computation You're

? (?) 01:33.360

talking about the next frontier, Yuri. You're looking at agents and chatbots. I know they're not exactly the same, but you see them actually advancing quite a bit more in 2026. What happens with these agents and chatbots in 2026? How are they used differently? And I got to ask

? (?) 01:46.680

you, do you think it impacts the job market?

Jure Leskovec (Computer Science Professor) 01:49.360

Uh, great question. I think we are going to move beyond chatbots, right? Today, you know, kind of we think of AI us asking questions and it providing answers. And we are really going to move to not to AI we chat with, but the AI that makes decisions and is making those decisions

Jure Leskovec (Computer Science Professor) 02:04.960

autonomously. That's I think where the where the next frontier is. Basically, these autonomous agents that are talking to other agents and they are taking tasks autonomously, making decisions autonomously and optimizing all kinds of workflows across a number of different

Jure Leskovec (Computer Science Professor) 02:21.920

industries. I think it's really about going beyond AI that is chatting to AI that is doing. So,

? (?) 02:28.560

Yuri, one other thing you're looking at are the visas that allow really high-skilled workers, the H1-B visas, to come into the country. You're obviously an advocate of that according to our notes. I do want to ask you a question. AI is supposed to also be an equalizer. We talk a

? (?) 02:40.120

lot about it potentially impacting the job market and things like that, but a lot of other people describe it as leveling the playing field. Why do we need these visas? Why isn't AI able to upskill some of the people we have here in the US and bring out the talent of people here

? (?) 02:52.840

in the US? You often hear that opportunity is not equal, but talent is equally distributed.

Jure Leskovec (Computer Science Professor) 02:58.640

I think the key for US to to remain the leader and to be ahead of everyone else is to remain the global talent magnet. And I think in that respect it is crucial to have strong, well-funded, forward-looking research institutions, universities that are educating this talent. And

Jure Leskovec (Computer Science Professor) 03:18.520

of course then the next step is as this talent is being educated is that there is an easy and a streamlined way for this talent to join the workforce and start contributing back to the economy. For example, um you know, at Stanford, it takes more than half a million dollars

Jure Leskovec (Computer Science Professor) 03:33.680

worth of investment, five years of my involvement to educate a single PhD student. And it is just not a bad investment if after this person has been educated, they have all the all the knowledge and are ready to start contributing to the economy. If it's hard for them to stay,

Jure Leskovec (Computer Science Professor) 03:48.840

if it's hard for them to fund the next startup, the next Nvidia, the next large thing. So I think the key is for US to to remain the global talent magnet, to have the most best forward-looking universities, research institutions, development centers, and to make it seamless to

Jure Leskovec (Computer Science Professor) 04:06.640

transition from educational research point of view to the to the industry, to the economy, start contributing, and start building this next generation leaders and companies. Alright.

? (?) 04:17.240

Speaking of next generation of leaders and companies, you're a founder yourself, cumo.ai. We were just talking about it a short time ago. What's the outlook for your company? Your company is actually in the manufacturing space.