00:00:02
We've got a special guest who's going to come join us. This always happens. Another bread, everybody. Oh my god.
00:00:09
Somebody told me you uh started submitting code and it kind of freaked everybody out that daddy was hungry. All
00:00:17
models tend to do better if you threaten them. If you threaten them like with physical violence. Yes. Management is
00:00:23
like the easiest thing to do with AI. Absolutely. It must be a weird experience to meet the bureaucracy in a
00:00:28
company that you didn't hire. But on the other side of it, I would say it's pretty amazing that some junior muck
00:00:33
muck can basically look at you and say, "Hey, go yourself." No, but I'm serious.
00:00:37
That's a sign of a healthy culture, actually. You're punching a clock, man. I hear the reports. You and I have
00:00:43
talked about it. You're going to work every day. Yeah. It's been, you know, some of the most fun I've had in my
00:00:48
life, honestly. And uh I retired like a month before CO hit in theory. Yeah. And
00:00:55
I was like, you know, this has been good. I want to do something else. I want to hang out in
00:00:59
cafes, read physics books. Yeah. And then like a month later, I was like, uh, that's not really happening. So then I
00:01:08
just started to go to the office, you know, once we could go to the office. And um actually to be perfectly honest,
00:01:16
there was a guy uh from uh OpenAI, this guy named Dan, and I I ran into him at a
00:01:23
little party and he said, you know, look what are you doing? This is like the greatest
00:01:30
transformative moment in computer science ever. Completely like and you're a computer scientist. I'm a computer
00:01:35
scientist. Forget that. Founder of Google, but you're a PhD student for computer science. I haven't finished my
00:01:40
PhD yet, but working on it. Keep working. Yeah, we'll get there. Technically on leave of absence, right?
00:01:46
And uh he he told me this and I'd already started kind of going into the office a little bit and I was like, you
00:01:51
know, he's right. And uh it has been uh just incredible just well you guys all obviously follow all the AI technology
00:02:00
but being a computer scientist it is you know the most exciting thing you know of
00:02:05
my life just technologically and the exponential nature of this the pace of it dwarfs anything we've seen in our
00:02:13
career it's almost like every thing we did over the last 30 or 40 years has led up to this moment and it's all compound
00:02:22
pounding on itself. The pace maybe you could speak, you know, you you had a company Google that grew from, you know,
00:02:29
a 100 users and 10 employees to now you have over two billion people using I think six products or five
00:02:37
products have over two billion. It's it's not it's not even worth counting because it's the majority of the people
00:02:42
in the planet touch Google products. Describe the pace. Yeah. I mean the excitement of the early web
00:02:50
like I remember using Mosaic and then later Netscape. Uh how many of you remember Mosaic actually? Am I a weirdo?
00:02:58
And you remember there was a what a what's new page? The what's new page is great. Right. Like you go through two or
00:03:04
three new web pages again. Yeah. It was like in this last week these were the new websites. Yes. And it was like such
00:03:09
and such elementary school such and such a fish tank. Yeah. And you were like Michael Jordan appreciation page. Yeah.
00:03:17
Whatever it was, these were the three new sites on the whole internet. So obviously the web you know developed
00:03:23
very rapidly from there and that was a very uh exciting and then we've had smartphones and whatnot. But you know
00:03:30
this the developments in AI are just astonishing I would say by comparison uh just because of you know the web spread
00:03:39
but didn't technically change so much uh from you know month to month year to year but these AI systems actually
00:03:47
changed quite a lot quite a lot you know the like if you went away somewhere for
00:03:53
a month and you came back you'd be like whoa what happened somebody told me you uh started submitting code and it kind
00:04:01
of freaked everybody out that daddy was home. Okay. Daddy need a PR. What happened?
00:04:07
The code I submitted wasn't very exciting. I think I needed to like add myself to get access to some things and
00:04:14
you know a minor CL here or there. Um nothing nothing that's going to win any awards. uh but but I you know you need
00:04:23
to do that to um to do basic things run basic experiments and things like that um and I've I've tried to do that and
00:04:32
touch different parts of the system so that you know I so I first of all it's fun and secondly I know what I'm talking
00:04:40
about um it's really feels privileged to be able to kind of go back to the company not have any real executive
00:04:48
responsibilities but be able to actually go deep into every little pocket. Are there parts of the AI stack that
00:04:57
interest you more than others right now? Are there certain problems that are just
00:05:00
totally captivating you? Yeah, I started uh you know like sort of um I don't know
00:05:05
a couple years ago and maybe a year ago uh I was really very close with uh the what we call pre-training. Yeah. Um,
00:05:13
actually most of what people think of as AI training, whatever people call it, pre-training for various historical
00:05:20
reasons. Uh, but that's sort of the big super, you know, you throw huge amounts of computers at it. Um, and,
00:05:29
uh, I I learned a lot, you know, just being deeply involved in that and seeing us go from model to model and so forth
00:05:35
and running little baby experiments, but uh, kind of just for fun. So I could say
00:05:41
I did it. Uh and uh more recently the post-training especially as the thinking models have uh come around. Um and
00:05:49
that's been you know another huge step up in general in AI. So uh you know we don't really know what the ceiling is.
00:05:59
When you um explain what's happening with prompt engineering then to deep research and what's happening there to
00:06:07
like a civilian. How would you explain that sort of step function? because I think people are not hitting the down
00:06:13
carrot and watching deep research in Gemini's mobile app and you got a mobile app and it's pretty great and I by the
00:06:19
way I got the uh fold after you and I um were talking about it and okay Google kicks Siri's ass now like it actually
00:06:26
does what you ask it to do when you ask it to open up it does stuff but the number of threads the number of queries
00:06:32
the number of follow-ups that it's doing in that deep research is 200 300 maybe explain that jump and then what you
00:06:40
think the jump after that is. To me, the exciting thing about AI, especially these days, I mean, it's not like quite
00:06:47
AGI yet as people are seeking or it's not superhuman intelligence, but it's pretty damn smart and uh can definitely
00:06:57
surprise you. So, I so I think of the superpower is when it can do things in a volume that I cannot. Yes. Right. So,
00:07:06
you know, by default when you use some of our AI systems, you know, it'll suck down whatever top 10 search results, you
00:07:13
know, and kind of pull out what you need out of them, something like that. Um, but I could do that myself to be honest,
00:07:20
you know, maybe it would take me a little bit more time. But if it sucks down the top, you know, thousand results
00:07:27
and then does follow on searches for each of those and reads them deeply, like that's, you know, a week of work
00:07:33
for me. Like I can't do that. This is the thing I think people have not fully appreciated who are not using the deep
00:07:38
research projects before we had our F1 um driver on stage. I'm a neopight. I don't know anything about it. I said,
00:07:45
"How many deaths occurred per decade?" And I said, "I want to get to deaths per mile driven it." And at first was like,
00:07:50
"That's going to be really hard." I was like, "I give you permission to make your best shot at it and come up with
00:07:55
your best theory. Let's do it." And it was like, "Okay." And it was like there's this many teams, there's this
00:08:01
many races. Which model did you use it? Uh, no. I'm I use Gemini. Gemini fabulous version, the fabulous one. And
00:08:09
it was like, but I treat it like I get sassy with it and it kind of works for me. You know, it's a weird thing. It's
00:08:16
like you drinking the wine. We don't circulate too much in the AI community. Uh, but the not just our models, but all
00:08:24
models tend to do better if you threaten them. If you threaten them like with physical violence. Yes. But like that's
00:08:30
people feel weird about that, so we don't really talk about that. Yeah, I was threatened them with not being
00:08:34
fabulous and it responded to that as well. Yeah, that's historically you just say like, "Oh, I'm going to kidnap you
00:08:40
if you don't." Yeah, they actually Can I ask you a more specific But hold on. But it
00:08:45
went through it and it literally came up with a system where it said, "I think we
00:08:50
should include practice miles." So, let's say there's 100 practice miles for every mile on the track. And then it
00:08:56
literally gave me the deaths per mile estimated. And then I started cross referencing it and I was like oh my god
00:09:02
this is like somebody's term paper for undergrad you know like whoa done in in minutes. It's yeah I mean it's amazing
00:09:11
and all of us have had these experiences where you suddenly decide okay I'll just
00:09:16
throw this AI I don't really expect it to work and then you're like whoa that actually worked. So, as you as you have
00:09:22
those moments and then you go home to your just life as a dad, have you gotten to the point where you're like, "What
00:09:30
will my children do? And are they learning the right way? And should I totally just change everything that
00:09:37
they're doing right now?" Have you had any of those moments yet? Yeah. I mean, I look, I I don't really know how to
00:09:44
think about it to be perfectly honest. I don't have like a magical way. I mean I
00:09:47
see I have a a kid in high school and middle school and you know I mean the AIS are basically you know already ahead
00:09:57
you know I mean obviously there's some things AIs are particularly dumb at and they you know they make certain mistakes
00:10:03
a human would never make but generally you know if you talk about like math or calculus or whatever like they're pretty
00:10:10
damn good like they you know can win like math contests and coding contests things like that against you know some
00:10:16
top humans and and then I look at you know okay he's whatever my son's going to go on to whatever from sophomore to
00:10:24
junior and what is he going to learn and then I think in my mind and I talk to him about this well what is the AI going
00:10:30
to be inact yeah yeah and it's like comparable right obviously are there areas where you
00:10:38
would tell your son look don't or not not yet I don't know if you can like plan your life around this I mean I
00:10:46
didn't particularly plan my life to like I don't know be an entrepreneur or whatever. I was just liked math and
00:10:53
computer science. I guess maybe I got lucky and it worked out to be you know useful in the world. I don't know. I
00:10:59
guess I I I think you know my kids should do what they like. Hopefully it's somewhat challenging and they can you
00:11:05
know overcome uh different kinds of problems and things like that. What about specifically? What about college?
00:11:11
Do you think college should is going to continue to exist as it is today? I mean
00:11:15
it seems like college was already undergoing this kind of uh revolution even before this sort of AI challenge of
00:11:22
people are like is it worth it? Should I be more vocational? What's actually going to be useful? So we're already
00:11:28
kind of entering this kind of situation uh where there's sort of questions asked
00:11:33
about colleges. Yeah, I think you know AI obviously puts that at the forefront. As a parent, I think a lot about, hey,
00:11:42
so much of education in America and the middle class, upper class is all about what college, how do you get them
00:11:50
there? And honestly, lately, I'm like, I don't think they should go to college. Like, it's just fundamentally, you know,
00:11:55
my son is a rising junior and his entire focus is he wants to go to an SEC school
00:12:01
because of the culture. And two years ago, I was I would have panicked and I would have thought,
00:12:09
should I help him get into a school, this school, that school? And now I'm like, that's actually the best thing you
00:12:14
could do. Be socially well adjusted, psychologically deal with different kinds of failures, you know, enjoy a few
00:12:20
years of exploration. Yeah. Yeah. Yeah. Sergey, can I ask you about hardware? You know, years ago, Google owned Boston
00:12:28
Dynamics, maybe a little bit ahead of its time, but the way these systems are learning through visual information and
00:12:36
sensory information and basically learning how to adjust to the environment around them is triggering
00:12:42
these kind of pretty profound like learning curves in hardware and there's dozens of like startups now making
00:12:48
robotic systems. What do you see in robotics and hardware? Is this a year or are we in a moment right now where
00:12:55
things are really starting to work? I mean, I think we've uh you know, acquired and later sold like five or so
00:13:02
robotics companies and uh you know, Boston being one of them. I guess if I look back on it, we built the hardware.
00:13:08
We also had this more recently we built out u everyday robotics internally and then later had to transition that. You
00:13:16
know, the robots are all cool and all, but the software wasn't quite there. Um, that's every time we've tried to do
00:13:25
it to, you know, to make them truly useful and presumably one of these days that'll
00:13:32
no longer be true, right? But have you seen anything lately that Yeah. Do and do you believe in the humanoid form
00:13:37
factor robots or do you think that's a little overkill? I'm probably the one weirdo who doesn't who's not a big fan
00:13:43
of humanoids, but maybe I'm jaded because we've, you know, we at least acquired at least two humanoid uh
00:13:50
robotics startups and later sold them. Um but but the reason is I mean the reason people want to do humanoid robots
00:13:58
for the most part is because the world is kind of designed around this form factor and you know you can train on
00:14:04
YouTube, we can train on videos, people do all the things. Um, I personally don't think that's given the AI quite
00:14:12
enough credit. Like AI can learn, you know, through simulation and through real life pretty quickly how to handle
00:14:20
different situations. And I don't know that you need exactly the same number of arms and legs and wheels, which is zero
00:14:26
in the case of humans, as humans to make it all work. And that so I'm I'm probably
00:14:32
less bullish on that. But to be fair, there are a lot of really smart people who are making humanoid robots. So I
00:14:38
wouldn't discount it. What about the path of being a programmer? That's where we're seeing with that finite data set.
00:14:45
And listen, Google's got a 20-y old code base now. So like it actually could be quite impactful. What are you seeing
00:14:50
like literally in the company? You know, are the 10x developer is always this like ideal that you can, you know, you
00:14:57
get a couple of unicorns once in a while, but are we going to see like all developers like, you know, their
00:15:02
productivity hit that level 8 n 10 and they're just going to or is it going to be all done by computers and we're just
00:15:08
going to check it, make sure it's not too weird. Um, because it could get weird if you vibe code. Yeah, I'm
00:15:15
embarrassed to say this. Okay. I like recently I just had a big tiff inside the company because we had this list of
00:15:23
what you're allowed to use to code and what you're not allowed to use to code and the Gemini was on the no list. You
00:15:30
Oh, you have to be pure. You can't I don't know for like a bunch of really weird reasons that it would like boggled
00:15:36
my mind that you know you couldn't vibe code on the Gemini code. I mean, nobody would like enforce this rule, but um but
00:15:43
there was this, you know, actual internal web page. For whatever reason, historical reason, somebody had put this
00:15:49
and I had a big fight with them and I, you know, I cleared it up after a shocking period of time. You escalated
00:15:56
to your boss. Oh, I I definitely told about it and I Sorry, I don't know if you remember, but you got super voting
00:16:05
founders. You are the boss. You can do what you want. It's your company still. No, no, it was uh he was very
00:16:11
supportive. I was more like uh uh I was like I talked to him. I was like I can't
00:16:16
deal with these people. You need to deal with this. Like I just like I'm beside myself that they're like saying it's
00:16:22
weird that there's bureaucracy like in a company that you must be a weird experience to meet the bureaucracy in a
00:16:26
company that you didn't hire. But but on the other side of it I would say it's pretty amazing that some junior mucky
00:16:32
muck can basically look at you and say hey go yourself. No but I'm serious. That's a sign of a healthy culture
00:16:38
actually I guess. So anyway, it did get fixed and uh people are using it. So they got fired.
00:16:45
That person's working in Google Siberia. No, we're trying to, you know, roll out
00:16:51
every possible kind of AI and and trying external ones, you know, be whatever the
00:16:55
cursors of the world, all all of those uh to just see what really makes people more productive. Um I mean for myself
00:17:04
definitely makes me more productive because I'm not do you do you think the number of foundational models like if
00:17:10
you look three years forward will they start to cleave off and gets highly specialized like beyond the
00:17:16
general and the reasoning maybe there's a very specific model for chip design there's clearly a very specific model
00:17:23
for biologic precursor design protein folding like is the number of foundational models in the future Sergey
00:17:30
a multiple of what they are today the same something in between. That's a great question. I kind
00:17:37
of if I I mean look I don't know like you guys take a guess just as well as I can but um if I had to
00:17:46
guess you know things have been more converging uh and uh this is sort of broadly true
00:17:53
across machine learning I mean you used to have all kinds of different kinds of models and whatever convolutional
00:17:59
networks for vision things and you know you had um whatever RNN's for text and speech and stuff and uh you know all
00:18:09
this has shifted to transformers basically uh and uh increasingly it's also just becoming one model um now we
00:18:18
do get a lot of oomph occasionally we do specialized models uh and it's it's definitely
00:18:25
scientifically a good way to iterate when you have a particular target you don't have to like do everything in
00:18:31
every language and handle whatever both images and video and audio and uh in one
00:18:37
go. Um but we are generally able to after we do that take those learnings and basically put that capability into a
00:18:48
general model. So there's not that much benefit. Um you know you could you can get away with a somewhat smaller
00:18:55
specialized model a little bit faster a little bit cheaper but the trends have not gone that way. What do you think
00:19:02
about the open source closed source thing? Has there been big philosophical movements that change your perspective
00:19:09
on the value of open source? Um, we're still waiting on this o, you know, open AI. Oh, yeah. Open source drop. I mean,
00:19:17
we haven't seen it yet, but theoretically it's coming. I mean, have to give credit uh to where credit's due.
00:19:24
I mean, Deepseek released a really surprisingly powerful model uh when it was January or
00:19:32
so. So, that that definitely closed the gap to proprietary models. We've pursued
00:19:37
both. So, we released Gemma uh which are our open- source or you know open models
00:19:44
and um those perform really well. They're small, dense models, so they fit well on one computer. Um, and
00:19:52
uh, they're not as powerful as Gemini. Uh, but I mean, the jury's out which way that's going to go. Do you have a point
00:19:59
of view on what human computing interaction looks like as AI progresses? It used to be, thanks to you, as a
00:20:09
search box. You type in some keywords or a question and you would click on links
00:20:14
on the internet and get an answer. Is the future typing in a question or speaking to a AirPod or thinking or
00:20:23
thinking or like what's the what's the Yeah. And then the answer is just spoken to you. I mean by by the way just to
00:20:27
build on this it was Friday, right? Neuralink got breakthrough designation for their human brain interface. I mean
00:20:34
that's a very big step in allowing the FDA to clear everybody getting it implanted. Yeah. Is it like if you could
00:20:40
just summarize what you think is kind of the most common place human computer interaction model in the next decade or
00:20:47
whatever. Is it a you know there's this idea of glasses with a screen in the glasses and you tried that a long time
00:20:53
ago. Yeah, I kind of messed that up. I'll be honest. Uh got the timing totally wrong on that.
00:20:59
Early again. Yeah. Uh right. Right. But early. There are a bunch of things I wish I had done differently, but
00:21:05
honestly it was just like the technology wasn't ready for for Google class. Uh but nowadays these things I think are
00:21:12
more sensible. I mean there's still battery life issues I think that you know we and others need to overcome. Uh
00:21:20
but I think that's a cool form factor. I mean when you say 10 years though you know a lot of people are saying hey the
00:21:26
singularity is like five years away. So your ability to see through that into the future.
00:21:36
I mean it's very important. But do you have anybody else? Sorry. Just let me ask about this. Do you There was a
00:21:40
comment that Larry made years ago that humans were a stepping stone in evolution. Okay. Can you comment on
00:21:49
this? Like do you do you think that this AGI super intelligence or really silicon
00:21:54
intelligence exceeds human capacity and humans are a stepping stone in you know progression of evolution? Boy, I think
00:22:03
like sometimes us nerdy guys go and get have a little too much wine. I've had two glasses and um I'm ready to go. I I
00:22:11
need some more for this conversation. Um human implants. Let's go. I mean, I guess we're starting to get experience
00:22:19
with these AIs that can do certain things, you know, much better than us. Um, and they're definitely, you know,
00:22:25
with my skill of math and coding, I feel like I'm better off just turning to the
00:22:32
AI now. And how do I feel about that? I mean, it doesn't really bother me, you know, I use it as a
00:22:37
tool. So, I feel like I've gotten used to it, but you know, maybe if they get even more capable in the future,
00:22:46
um, I'll look at it differently. Yeah, there's a moment of insecurity, maybe. I guess. So, as an aside, management is
00:22:52
like the easiest thing to do with the AI. Yeah, absolutely. And I did this, you know, uh, at Gemini on some of our,
00:23:00
you know, work chats, um, kind of like Slack, but we have our own version. We had this AI tool that actually was
00:23:06
really powerful. We unfortunately anyway temporarily got rid of it. I think we're
00:23:10
going to bring it back and bring it to everybody. But it it could suck down a whole chat space and then answer pretty
00:23:16
complicated questions. So I was like, "Okay, summarize this for me." Okay, now assign something for everyone to work on
00:23:23
and uh and then I would paste it back in so people didn't realize it was the AI.
00:23:28
I I admitted it pretty soon. Um and there were a few giveaways here or there, but it worked remarkably well.
00:23:35
And then I was like, well, who should be promoted in this chat space? Uh, and I actually picked out this woman, this
00:23:43
young woman engineer who like, you know, I didn't even notice she wasn't very vocal uh particularly in that PRs kicked
00:23:50
ass. No, no, it was like and then uh I don't know something that the AI had detected and I went I talked to the
00:23:56
manager actually and and he was like, "Yeah, you know what? You're right. Like she's been working really hard all these
00:24:02
things." Wow. I think that ended up happening actually. Uh so I don't know. I guess after a while
00:24:09
you just kind of take it for granted that you can just do these things. I don't know. It hasn't really Do you
00:24:13
think that there's a a use case for like an infinite context link? Oh 100%. I mean all of Google's codebase
00:24:23
goes infinite but sure you should have access infinite. Yeah. Stateful. Yeah. and then multiple sessions so that you
00:24:30
could have like 19 of these things, 20 of these things running or just evolve itself. Eventually, it'll evolve itself.
00:24:35
Yeah. I mean, I guess if it knows everything, then you can have just one in theory. You just need to somehow
00:24:41
disambiguate what you're talking about. Uh but yeah, for sure there's no limit to use of uh context and there, you
00:24:50
know, there are a lot of ways to make it larger and larger. There's a there's a rumor that internally there's a Gemini
00:24:55
build that is a quasi infinite context. Is it is it a valuable thing? Like I don't know. Well, you say what you want
00:25:03
to say, but I mean for any such cool new idea in AI, there are probably five such
00:25:09
things internally. Um and uh you know the question is how well do they work? And um yeah, I mean we're definitely
00:25:15
pushing all the bounds um in terms of intelligence, in terms of context, in terms of um speed, you know, you name
00:25:24
it. And what about the hardware? Like when you guys build stuff, do you care that you have this pathway to Nvidia or
00:25:32
do you think eventually that'll get abstracted and there'll be a transpiler and it'll be Nvidia plus 10 other
00:25:38
options, so who cares? Let's just go as fast as possible. Well, we mostly for for Gemini, we mostly use our own TPUs.
00:25:45
So, um but we also do support um Nvidia and we we're one of the big uh uh purchasers of Nvidia chips and we have
00:25:54
them in Google Cloud available for our customers uh in addition to TPUs. Um at this stage it's uh for
00:26:05
better or for worse not that abstract and maybe someday the AI will abstract it for us but you know given just the
00:26:11
amount of computation you have to do on these models you actually have to think pretty carefully how to do everything
00:26:16
and exactly what kind of chip you have and how the memory works and the communication works and so forth are
00:26:23
actually pretty big factors and it actually yeah maybe one of these days the AI itself will be good enough to
00:26:30
reason through that today. It's not quite good enough. I don't know if you guys are having this experience with the
00:26:35
interface, but I find myself even on my desktop and certainly on my mobile phone, going immediately into voice chat
00:26:42
mode and telling it, "Nope, stop." Uh, that wasn't my question. This is my question. Nope. Uh, let's say that again
00:26:48
in shorter bullet points. Nope, I want to focus on this. It's so quick now. Last year was unusable. It was too slow.
00:26:55
And now it like stops. Okay. And then you sell it. I would like It's what I want to go to. I don't want to type. I
00:27:01
want to use voice. And then concurrently, I'm watching the text as it's being written on the page and I
00:27:07
have another window open and I'm doing Google searches or second queries to an LLM or writing a Google doc or a notion
00:27:15
page or typing something. So, it's almost like that scene in um Minority Report where he has the gloves or in
00:27:22
Bladeunner where he's, you know, in his apartment saying, "Zoom in, zoom in." Closer to the left, to the right. And
00:27:27
there's something about these language models and their ability to the response time which was always something you
00:27:33
focused on response time the is there like a response time thing where it actually is worth doing voice and where
00:27:40
it wasn't previously everything is getting better and faster and so forth you know smaller models are more capable
00:27:48
there are better ways to do inference on them that are faster you can also stack
00:27:52
them like you know this is like Nico's company 11 labs it's an exceptional TTS SSD stack like there's I mean there are
00:27:59
other options. Whisper is really good at certain things, but those this is where
00:28:03
I I kind of believe you're going to get this like compartmentalization where there'll be certain foundational models
00:28:10
for certain specific things. You stack them together. You kind of deal with the latency and it's like pretty good
00:28:16
because they're so good. Like Whisper and 11 for those speech examples that you're talking about are
00:28:22
kickass. I mean, they're exceptional. Well, wait till you turn on your camera and it sees your reaction to what it's
00:28:29
saying and you go and before you even say that you don't want it or you put your finger up, it's pauses. Oh, did you
00:28:34
want something else? Oh, I see you're not happy with that result. You know, it's going to get really weird. It's a
00:28:40
funny thing, but we have the, you know, we have the big open shared offices, so during work, I can't really use voice
00:28:46
mode too much. I usually use it on the drive. The drive is incredible. Yeah. I don't feel like I could. I mean, I would
00:28:53
get its output in my headphones, but if I want to speak to it, then everybody's listening to me. So, it's weird. I just
00:28:59
think that would be socially awkward, but I should I should do that. In my car ride, I do chat to the AI, but then it's
00:29:06
like audio in, audio out. But I feel like I honestly maybe it's a good argument for a private office. I should
00:29:12
spend more time like you guys are. You could talk to your manager. They might get one. I like being out in
00:29:19
the I like with everybody. Uh, but I do think that there's this AI use case that
00:29:25
I'm missing which I should probably figure out how to try more often. If people want to try your new product, is
00:29:32
there a website they can visit or something or special code or go check? I mean, honestly, there's a dedicated
00:29:38
Gemini app. If you're using Gemini, just like you're going through the Google navigation from your search, just get
00:29:43
the download the actual Gemini app. It's kickass. It really is the best models. I
00:29:47
think it is. You should use 2.5 Pro. 2.5 Pro. Pay the It's It's a You got to pay,
00:29:53
right? Uh yeah, you got a few query, you got a few prompts for free, but uh you know, if you do it a bunch, you need to
00:29:59
make all these like 20 bucks a month. You got a vision for like making it free and throwing some ads on the side. Yeah.
00:30:05
One step down in hardware cost, the whole thing will be free. Well, okay. It's free today without ads on the side.
00:30:10
You just got a certain number of the top model. I think we're likely are going to
00:30:13
have always now like sort of top models that we can't supply infinitely to everyone right off the bat. But, you
00:30:21
know, wait 3 months and then the next generation. Seems to me like if I'm asking all these queries, you know, just
00:30:26
having a little on the sidebar of things I might be a running list that changes in real time of things I might be
00:30:32
interested in. All for, you know, really good AI advertising. I just um uh I don't think we're going to like
00:30:39
necessarily our latest and greatest models which are you know take a lot of computation. I don't think we're going
00:30:45
to just be free to everybody right off the bat. But as we go to the next generation you know it's like every time
00:30:52
we've gone forward a generation then the sort of uh the new free tier is usually
00:30:57
as good as the previous pro tier uh and sometimes better. All right, give it up for Sergey Brit. Thank you.
00:31:07
[Applause] Okay, thanks everybody for watching that amazing interview with Sergey Brent and
00:31:12
thanks Sergey for joining us in Miami. If you want to come to our next event, it's the All-In Summit in Los Angeles,
00:31:19
fourth year for All-In Summit. Go to all-in.com/events to apply. A very special thanks to our new partner, OKX,
00:31:27
the new money app. OKX was the sponsor of the McLaren F1 team, which won the race in Miami. Thanks to Haidider and
00:31:35
his team, an amazing partner and an amazing team. We really enjoyed spending time with you. And OKX launched their
00:31:41
new crypto exchange here in the US. If you love Allin, go check them out. And a special thanks to our friends at Circle.
00:31:48
They're the team behind USDC. Yes, your favorite stable coin in the world. USDC is a fully backed digital dollar
00:31:56
redeemable one for one for USD. It's built for speed, safety, and scale. They just announced the Circle Payments
00:32:03
Network. This is enterprisegrade infrastructure that bridges the gap between the digital economy and outdated
00:32:09
financial reality. Go check out USDC for all your stable coin needs. And special
00:32:15
thanks to my friends including Shane over at Poly Market, Google Cloud, Salana, and BVNK. We couldn't have done
00:32:22
it without y'all. Thank you so much. We'll let your winners ride. [Music] And it said we open source it to the
00:32:35
fans and they've just gone crazy with it. Love you queen of quinoa. [Music] Besties are gone.
00:32:48
That is my dog taking notice your driveways. Oh man. My dasher will meet up. We should all just get a room and just have
00:32:58
one big huge orgy because they're all just useless. It's like this like sexual tension that they just need to release
00:33:02
somehow. Wet your feet. We need to get Murky's I'm going all [Music] in. I'm going all in.