Co-Founder & Director, Google
Short Preview of Sabrina debating AGI with Google founder Sergey Brin
Aspen Institute January 14, 2026. Thank you to the Aspen Center for Physics for use of this 7-minute preview. To view the entire 41-minute debate click: https://www.youtube.com/live/Rs5_Q7nU...
Thank you to Sergey Brin, Adam Brown and Ethan Dyer!
(Sorry for the audio/video sync issues.)
Sabrina Pasterski
Physics Girl
About Sergey Brin
In May and June 2025, Sergey Brin made several public appearances discussing the future of AI and Google's work. At Google IO in May 2025, Brin said he did not think "we really know what the world looks like in 10 years" due to the rate of progress in AI. He also commented on the simulation hypothesis, stating that if humans are in a simulation, the argument would apply recursively, leading to an "infinite stack of simulations" or a need for a "stopping criteria." He said he does not believe humans are "really equipped to reason about sort of one level up in the hierarchy." At an AGI House event in June 2025, Brin discussed the convergence of specialized models into general ones, noting that "increasingly our main Gemini LLMs can be the state-of-the-art for math" and other scientific questions. He defined AGI as "the idea of the AI can actually improve itself," while acknowledging that others define it as an AI that can "do anything a person can do." Brin said he felt "very good about where Gemini is" but acknowledged that Google was "a little bit late on really focusing deeply" on coding, giving kudos to competitors like GPT-5.5 for overnight coding tasks while pitching Gemini 3.5 Flash for speed. On the Moonshot Podcast, Brin reflected on X's work, saying the company was "premature at productizing" some projects like Google Glass, but that it served as a "phenomenal learning platform." He also expressed interest in Von Neumann machines, describing the concept of building something that can "build itself" and "exponentially build something."
Transcript (23 segments)
H
Host0:06
Thanks. Yeah, we're trying to go for the maximum number of introductions in one of these public talks. And the panelists are welcome to come up. Sergey, Sabrina. Yeah. So our panelists in addition to Adam who you've met are going to be Sabrina Perski, who's a theoretical physicist and faculty at the Perimeter Institute. She works on high energy physics, celestial holography, and thinks about how AI can accelerate physics. She was selected for Forbes 30 under 30 for science and is an all-star. And Sergey Brin, who co-founded Google and has recently come out of retirement, you know, Jordan style, to help Google's AI effort. You might know him from his role in the major motion picture, The Internship. So yeah, I think feel free to take a seat.
Um yeah, let's see. So I think to start things off, can try to stoke a little controversy and see if Sergey or Sabrina, there's anything that you take umbrage with from Adam's talk that you think the future is not going to look the way Adam has painted it or that the past didn't look like the way he told it.
S
Sergey Brin1:34
Well, I'll just mention that in the graph where he has sort of the utopia or extinction, technically speaking, the GDP per capita will go up in either case.
H
Host1:49
Okay. I think a strongly worded letter to the Financial Times would be in order.
S
Sergey Brin1:54
Yeah, it's an extremely competitive field. I mean, you know, if you were to go away on a, I don't know, ocean crossing for a month or something, you would be shocked at what happened in the month.
S
Sabrina Perski2:12
Yeah, I guess for me, I'm not so scared of maybe the fact that things are moving fast, but just how things are synced up in the right way. So the thing that would be scariest to me is, okay, so say everybody comes out of this convinced within five years, you know, like Google, like Gemini or all these other chat bots are going to be better than any academic. Why, if our data is going to be available to these companies to train on anyway, is it clear that our grant money won't be going to those companies? So the whole vision of a future where all this stuff is automated sounds great, but how that actually gets deployed scares me.
S
Sergey Brin2:43
I mean, when I was a graduate student, I don't feel I was trained per se. I mean, there were maybe a few classes that I had and a couple advisers here and there who gave advice, but mostly I felt like it was a time and space to have a lot of academic freedom without lots of responsibilities, for example, taking care of graduate students or things like that. So I think from the point of view of having academic freedom, I find the LLM very empowering.
H
Host3:18
You're worried that progress will slow down or just that humans will be alienated from their progress?
S
Sabrina Perski3:24
I would rather have the progress happen very quickly than have it everybody believe it's going to happen really quickly and happen to the extent that some people then devalue like education or things like that, and then maybe it didn't actually happen at the rate you needed and then something kind of falls out. I think it'd be awesome to be collaborating more with industry, especially just at the level of capabilities. I mean, maybe you're saying vibe coding can get us to being able to code, but if you've ever been really crappy at coding and playing with these large language models, you just get caught being dumb yourself, at least as far as I don't think we have access to maybe the better versions that have that.
S
Sergey Brin4:01
Yeah, I guess, well, starting from search, which obviously we got best known for, to now communication, be it Gmail, Android, so forth, and now advanced AIs. I mean, the exciting thing is that these things are available to pretty much everybody. I mean, I guess you need to buy a smartphone, but, you know, Android phones go for under a hundred bucks around the world. So they're accessible to almost everybody. I mean, it's the great equalizer. We all walk around as kings in the world, you know, kings in our pocket. Basically, we have access to the best information, the best communication, and now the best ability to think and reason. I mean, that's incredible. So, and kudos to the whole industry. But it's exciting times we live in.
S
Sabrina Perski4:51
But I do think that there's a fun sense in which, okay, so say you give me an automated grad student. I mean, it's not about the cost. It's like I don't have to train it. It's going to be already smarter than me. It can do all this work and I can automate all these steps. Then we get to do cooler things like we get to think about designing experiments. That's going to cost a hell of a lot more than anything, I mean, like a theorist would cost right now.
H
Host5:11
How much physics have you learned with Gemini?
S
Sergey Brin5:15
I think the question was when will these be the superhuman teachers and, you know, when will it mostly be humans learning from the models rather than humans teaching the models?
S
Sabrina Perski5:27
Um, that's a good question. I think if I'm being just practical today, the reason I find the models to be great teachers is because they're available at the time that's convenient for me, which is for example while I'm driving and then, you know, obviously use an audio-to-audio model. But I don't know, I guess in my life I don't have, you know, time to schedule for a teacher to show up if I wanted to learn Korean or something.
S
Sergey Brin6:01
Quantum field theory back in the day.
H
Host6:04
Now you have Gemini.
S
Sergey Brin6:05
Yeah, Leonard Susskind would show up. Actually, Adam arranged that. That was pretty impressive to teach quantum field theory. That was when I briefly retired. In fact, nowadays I don't know if I would have so much time for that. So I find LLMs to just be great teachers because they're available and essentially free.
H
Host6:30
You're saying that time is already here?
S
Sergey Brin6:32
I think from a practical standpoint, yes. I don't think they're necessarily as good as if you did have the best teacher in the world at the time that was convenient for you. But presumably they will get better. I think when I have a tough question these days, you know, I'll be honest, obviously I go to Gemini, but I'll try some of the others too to just get a diverse range of answers and, you know, sometime you never know, one got trained in a somewhat different way, took a different path, and I think you kind of want that kind of diversity.
S
Sabrina Perski7:04
I think there's going to be a role for human experts for a long time. One thing you might...
H
Host7:08
How long?
S
Sabrina Perski7:10
Um, we'll see.
H
Host7:14
Um, yeah, I guess let's wrap up and thank the panelists.