00:00:00
I first met our next guest Sam mman almost 20 years ago when he was working on a local mobile app called looped we
00:00:07
were both backed by seoa capital and in fact we were both in the first class of seoa Scouts he did investment in a
00:00:14
little unknown fintech company called stripe I did Uber and in that tiny experiment did Uber I've never heard
00:00:20
that before yeah I think so it's possible starting already you should write a book Jacob
00:00:28
maybe let your winners [Music] Rainman David and in said we open source it to the fans and they've just gone crazy
00:00:40
[Music] with that tiny experimental fund that Sam and I were part of a Scouts is
00:00:48
sequoia's highest multiple returning fund couple of low digigit Millions turned into over 200 million I'm told
00:00:54
and then he did yeah that's what I was told by ruoff yeah and he did a stint at Y combinator where he was president from
00:01:00
2014 to 2019 in 2016 he co-founded open AI with the goal of ensuring that artificial general intelligence benefits
00:01:08
all of humanity in 2019 he left YC to join openai full-time as CEO things got really interesting on November 30th of
00:01:16
2022 that's the day open AI launched chat GPT in January 2023 Microsoft invested 10 billion in November 2023
00:01:25
over a crazy 5ay span Sam was fired from open AI everybody was going to go work at Microsoft a bunch of hard emojis went
00:01:33
viral on x/ Twitter and people started speculating that the team had reached artificial general intelligence the
00:01:40
world was going to end and suddenly a couple days later he was back to being the CEO of open AI in February Sam was
00:01:48
reportedly looking to raise $7 trillion do for an AI chip project this after was
00:01:53
reported that Sam was looking to raise a billion from masi yoshian to create an iPhone killer with Johnny I the
00:01:58
co-creator of the iPhone all of this while chat GPT has become better and better and a household name is having a
00:02:05
massive impact on how we work and how work is getting done and it's reportedly the fastest product to hit 100 million
00:02:12
users in history in just two months and check out opening eyes insane Revenue ramp up they reportedly hit two billion
00:02:19
in ARR last year welcome to the all-in podcast Sam mman thank you thank you guys Sak you want to Le us off here okay
00:02:28
sure I mean I I think the whole industry is waiting with baited breath for the release of GPT 5 I guess it's been
00:02:35
reported that it's launching sometime this summer but that's a pretty big window can you narrow that down I guess
00:02:41
where where are you in the release of GPT 5 uh we we take our time on releases of major new models and I don't think we
00:02:51
uh I think it will be great uh when we do it and I think we'll be thoughtful about how we do it uh like we may
00:02:58
release it in a different way than we've released previous BS models um also I don't even know if we'll call it gbt 5
00:03:03
um what I what I will say is you know a lot of people have noticed how much better gbd4 has gotten um since we've
00:03:11
released it and particularly over the last few months I think I think that's like a better hint
00:03:18
of what the world looks like where it's not the like 1 two 3 4 5 six seven but you you just you use an AI system and
00:03:27
the whole system just gets better and better fairly continous ly um I think that's like both a better technological
00:03:34
Direction I think that's like easier for society to adapt to um but but I assume that's where we'll head
00:03:41
does that mean that there's not going to be long training cycles and it's continuously retraining or training
00:03:47
submodels Sam and maybe you could just speak to us about what might change architecturally going forward with
00:03:53
respect to large models well I mean one one one thing that you could imagine is this just that
00:04:00
you keep training right a model uh that that would seem like a reasonable thing to
00:04:05
me do you think you talked about releasing it differently this time are you thinking maybe releasing it to the
00:04:11
paid users first or you know a slower roll out to get the red teams tight since now there's so much at stake you
00:04:19
have so many customers actually paying and you've got everybody watching everything you do you know is it is it
00:04:26
moreful now yeah still only available to the paid user users but one of the things that we really want to do is
00:04:32
figure out how to make more advanced technology available to free users too I think that's a super important part of
00:04:39
our mission uh and and this idea that we build AI tools and make them super widely
00:04:45
available free or you know not that expensive whatever it is so that people can use them to go kind of invent the
00:04:51
future rather than the magic AGI and the sky inventing the future and showing it
00:04:55
down upon us uh that seems like a much better path it seems like more spiring path I also think it's where things are
00:05:01
actually heading so it makes me sad that we have not figured out how to make gp4
00:05:06
level technology available to free users it's something we really want to do it's
00:05:10
just very expensive I take it's very expensive yeah chamal your thoughts I think maybe the the two big vectors Sam
00:05:17
that people always talk about is that underlying cost and sort of the latency that's kind of rate limited a killer app
00:05:26
and then I think the second is sort of the long-term ability for people to build in
00:05:33
an open source World versus a closed Source world and I think the crazy thing about this space is that the open source
00:05:40
Community is rabid so one example that I think is incredible is you know we had these guys do a pretty crazy demo for
00:05:47
Devon remember like even like five or six weeks ago that looked incredible and then some kid just published it under an
00:05:54
open MIT license like open Devon and it's incredibly good and almost as good as that other
00:06:02
thing that was closed source so maybe we can just start with that which is tell me about the business decision to keep
00:06:09
these models close source and where do you see things going in the next couple years so on the first part of your
00:06:16
question um speed and cost those are hugely important to us and I don't want to like give a timeline on when we can
00:06:26
bring them down a lot cuz research is hard but I am confident we'll be able to um we want to like cut the latency super
00:06:32
dramatically we want to cut the cost really really dramatically um and I believe that will happen we're still so
00:06:39
early in the development of the science and understanding how this works plus we
00:06:43
have all the engineering Tailwinds so I I don't know like when we get to intelligence too cheap to meter and so
00:06:52
fast that it feels instantaneous to us and everything else but I do believe we can get there for you know a pretty high
00:07:00
level of intelligence and um I it's important to us it's clearly important to users and it'll unlock a lot of stuff
00:07:08
on the sort of Open Source close Source thing I think there's great roles for both I think
00:07:14
um you know we've open sourced some stuff we'll open source more stuff in the future but really like our mission
00:07:20
is to build towards AGI and to figure out how to broadly distribute its benefits we have a strategy for that
00:07:26
seems to be resonating with a lot of people it obviously isn't for everyone and there's like a big ecosystem and
00:07:32
there will also be open source models and people who build that way um one area that I'm particularly interested
00:07:38
personally in open source for is I want an open source model that is as good as it can be that runs on my phone and that
00:07:47
I think is going to you know the world doesn't quite have the technology for for a good version of that yet but that
00:07:52
seems like a really important thing to go do at some point will you do will you do that will you release I don't know if
00:07:57
we will or someone will but someone llama 3 llama 3 running on a phone well I guess maybe there's like a seven
00:08:05
billion version yeah yeah uh I don't know if that's if that will fit on a phone or not but that should be fitable
00:08:13
on a phone but I don't I I'm not I'm not sure if that one is like I haven't played with it I don't know if it's like
00:08:17
good enough to kind of do the thing I'm thinking about here so when when llama 3
00:08:21
got released I think the big takeaway for a lot of people was oh wow they've like caught up to GPT 4 I don't think
00:08:26
it's equal in all Dimensions but it's like pretty pretty close or pretty in the ballpark I guess the question is you
00:08:33
know you guys released four a while ago you're working on five or you know more upgrades to four I mean I think to
00:08:41
Chamas point about Devin how do you stay ahead of Open Source I mean it's just that's like a very hard thing to do in
00:08:47
general right I mean how do you think about that what we're trying to do is not make the sort of
00:08:57
smartest set of Weights that we can can but what we're trying to make is like this useful intelligence layer for
00:09:04
people to use and a model is part of that I think we will stay pretty far ahead of I hope we'll stay pretty far
00:09:11
ahead of the rest of the world on that um but there's a lot of other work around the whole system that's not just
00:09:19
that you know the the model Waits and we'll have to build up enduring value the oldfashioned way like any other
00:09:26
business does will have to figure out a great product and reasons to stick with it and uh you know delivered at a great
00:09:31
price when you founded the organization you the stated goal or part of what you discussed was hey this is too important
00:09:39
for any one company to own it so therefore it needs to be open then there was the switch hey it's too dangerous
00:09:45
for anybody to be able to see it and we need to lock this down because you you had some fear about that I think is that
00:09:52
accurate because the cynical side is like well this is a capitalistic move and then the I think
00:09:59
you know I'm I'm curious what the decision was here in terms of going from open we the world needs to see this it's
00:10:06
really important to closed only we can see it well how did you come to that that conclusion what were the disc part
00:10:12
of the reason that we released chat GPT was we want the world to see this and we've been trying to tell people that AI
00:10:17
is really important and if you go back to like uh October of 2022 not that many people thought AI was going to be that
00:10:23
important or that it was really happening um and a huge part of what we we try to do
00:10:30
is put the technology in the hands of people uh now again there's different ways to do that and I think there really
00:10:35
is an important role to just say like here's the weights have at it but the fact that we have so many people using a
00:10:42
free version of chat GPT that we don't you know we don't run ads on we don't try to like make money on we just put
00:10:47
out there because we want people to have these tools um I think it's done a lot to provide a lot of value and you know
00:10:54
teach people how to fish but also to get the world um really thoughtful about what's happening here now we still don't
00:11:01
have all the answers and uh we're fumbling our way through this like everybody else and I assume we'll change
00:11:06
strategy many more times as we learn new things you know when we started open AI
00:11:11
we had really no idea about how things were going to go um that we'd make a language model that we'd ever make a
00:11:17
product we started off just I remember very clearly that first day where we're like well now we're all here that was
00:11:23
you know it was difficult to get this set up but what happens now maybe we should write some papers maybe we should
00:11:28
stand around a whiteboard and we've just been trying to like put one foot in front of the other and figure out what's
00:11:33
next and what's next and what's next and I think we'll keep doing that can I just replay something and just make sure
00:11:40
I heard it right I think what you were saying on the open source close Source thing is if I heard it right all these
00:11:48
models independent of the business decision you make are going to become asymptotically accurate towards some
00:11:55
amount of accuracy like not all but like let's just say there's four or five that
00:11:58
are well capitalized enough you guys meta Google Microsoft whomever right so let's
00:12:05
just say four or five maybe one startup and on the open web and then quickly the accuracy or the value of
00:12:15
these models will probably shift to these proprietary sources of training data that you could get that others
00:12:20
can't or others can get that you can't is that how you see this thing evolving where the open web gets everybody to a
00:12:27
certain threshold and then it's just an arms race for data beyond that doesn't so I definitely don't think it'll be an
00:12:34
arms race for data because when the models get smart enough at some point it shouldn't be about more data at least
00:12:39
not for training it may matter data to make it useful um look the the one thing that I have learned most throughout all
00:12:48
this is that uh it's hard to make confidence statements a couple of years in the future about where this is all
00:12:53
going to go and so I don't want to try now I I will say that I I expect lots of very capable models in the world and you
00:13:01
know like it feels to me like we just like stumbled on a new fact of nature or science or whatever you want to call it
00:13:08
which is like we can create you can like I mean I don't believe this literally but it's like a spiritual point you know
00:13:18
intelligence is just this emergent property of matter and that's like a that's like a rule of physics or
00:13:22
something um so people are going to figure that out but there will be all these different ways to design the
00:13:27
systems people will make different Cho choices figure out new ideas and I'm sure like you
00:13:36
know like any other industry I would expect there to be multiple approaches and different people like different ones
00:13:43
you know some people like iPhones some people like an Android phone I think there'll be some effect like that let's
00:13:48
go back to that first section of just the the cost and the speed all of you guys are sort of a
00:13:55
little bit rate limited on literally nvidia's throughput right and I think that you and most everybody else have
00:14:02
sort of effectively announced how much capacity you can get just because it's as much as they can spin out what needs
00:14:08
to happen at the substrate so that you can actually compute cheaper compute faster get access to more energy how are
00:14:16
you helping to frame out the industry solving those problems well we we'll make huge algorithmic gains for sure and
00:14:24
I don't want to Discount that I you know I very interested in chips and energy but if we can make our if we can make a
00:14:30
same quality model twice as efficient that's like we had twice as much compute right and I think there's a gigantic
00:14:37
amount of work to be done there uh and I hope we'll start really seeing those results um other than that the whole
00:14:47
supply chain is like very complicated you know there's there's logic Fab capacity there's how much hbm the world
00:14:53
can make there's how quickly you can like get permits and pour the concrete make the data centers and then have
00:14:57
people in there wiring them all up there's finding the energy which is a huge bottleneck but
00:15:03
uh I think when there's this much value to people uh the world will do its thing
00:15:08
we'll try to help it happen faster um and there's probably like I don't know how to give it a number but there's like
00:15:15
some percentage chance where there is as you were saying like a huge substrate breakthrough and we have like a
00:15:22
massively more efficient way to do Computing but I don't I don't like Bank on that or spend too much time thinking
00:15:27
about it what about the device side and sort of you mentioned sort of the models
00:15:33
that can fit on a phone so obviously whether that's an llm or some slm or something I'm sure you're thinking about
00:15:39
that but then does the device itself change I mean is it does it need to be as expensive as an iPhone
00:15:47
uh I'm super interested in this uh I I love like great new form factors of computing and it feels like with every
00:15:55
major technological Advance a new thing becomes possible uh phones are unbelievably good so I
00:16:03
think the threshold is like very high here like what like I think I think like I personally think an iPhone is like the
00:16:10
greatest piece of technology Humanity has ever made it's really a wonderful product what comes after it like I don't
00:16:17
know I mean I was gonna that was what I was saying it's so good that to get Beyond it I think the bar is like quite
00:16:23
High well you've been you've been working with Johnny IV on on something right we've been discussing ideas but uh
00:16:29
I don't like if I knew is it that that it has to be more complicated or actually just much much cheaper and
00:16:35
simpler well every Mo almost everyone's willing to pay for a phone anyway so if you could like make a way cheaper device
00:16:41
I think the barrier to carry a second thing or use a second thing is pretty high so I don't think C given that we're
00:16:48
all willing to pay for phones or most of us are I don't think cheaper is the answer different is the answer then
00:16:56
would there be like a specialized chip that would run the phone that was really good at powering a you know a phone size
00:17:02
AI model probably but the phone manufacturers are going to do that for sure that doesn't that doesn't
00:17:07
necessitate a new device I think you'd have to like find some really different interaction Paradigm that the technology
00:17:14
enables uh and if I knew what it was I would be excited to be working on it right now but well you have you have
00:17:21
voice working right now in the app in fact I set my action button on my phone to go directly to chat gpt's voice app
00:17:27
and I use it with my kids and they love it talking to it's got latency issues but it's really we'll get we'll we'll
00:17:32
get that we'll get that better and I think voice is a hint to whatever the next thing is like if you can get voice
00:17:40
interaction to be really good it feels I think that feels like a different way to use a computer but
00:17:46
again with that by the way like what why is it not responsive and you know it feels like a CB you know like over over
00:17:55
it's really annoying to use you know uh in that way but it's also brilliant when
00:18:00
it gives you the right answer we are working on that uh it's it's so clunky right now it's slow it's like kind of
00:18:07
doesn't feel very smooth or authentic or organic like we'll get all that to be much
00:18:13
better what about computer vision I mean they have glasses or maybe you could wear a pendant I mean you take the
00:18:19
combination of visual or video data combine it with voice and now ai knows everything that's
00:18:27
happening around you super powerful to be able to like the multimodality of saying like hey cha gbt what am I
00:18:35
looking at or like what kind of plant is this I can't quite tell um that's obvious that that's like a
00:18:41
that's another I think like hint but whether people want to wear glasses or like hold up something when they want
00:18:47
that like I there's a bunch of just like like the the sort of like societal interpersonal issues here are all very
00:18:56
complicated about wearing a computer on your face um we we saw that with Google class people got punched in the face in
00:19:02
the mission started a lot of I forgot about that I forgot about that so so I I think it's
00:19:07
like what are the apps that could be unlocked if AI was sort of ubiquitous on people's
00:19:14
phones do you have a sense of that or what would you want to see built uh I I think what I want is just this
00:19:24
always on like super low friction thing where I can either by voice or by text or
00:19:33
ideally like some other it just kind of knows what I want have this like constant thing helping me throughout my
00:19:39
day that's got like as much context on as possible it's like the world's greatest assistant um and it's just this
00:19:45
like thing working to make me better and better uh there's there there's like a I
00:19:51
know when you hear people like talk about the AI future they're imag they imagine there's sort of
00:19:56
two different approaches and don't sound that different but I think they're like
00:20:00
very different for how we'll design the system in practice there's the I want an
00:20:04
extension of myself um I want like a ghost or an alter ego or this thing that really like is me is acting on my behalf
00:20:14
is um responding to emails not even telling me about it is is is sort of like it it becomes more me and is me and
00:20:24
then there's this other thing which is like I want a great senior employee it may get to know me very well I may
00:20:31
delegate it you know you can like have access to my email and I'll tell you the constraints but but I think of it as
00:20:36
this like separate entity and I personally like the separate entity approach better and think that's where
00:20:44
we're going to head um and so in that sense the thing is not you but it's it's like a always available always great
00:20:54
super capable assistant executive agent in a way like gets out there working on your behalf and understands what you
00:21:01
want and anticipates what you want is what I'm reading into what you're saying I think there'd be agent like Behavior
00:21:08
but there's like a difference between a senior employee and an agent yeah and like I want it you know I think
00:21:16
of like my I think like of it like one of the things that I like about a senior employee is
00:21:26
they'll they'll push back on me they will sometimes not do something I ask or they sometimes will say like I can do
00:21:32
that thing if you want but if I do it here's what I think would happen and then this and then that and are you
00:21:37
really sure I definitely want that kind of vibe which not not just like this thing that
00:21:42
I asking it blindly does it can reason yeah yeah and push back it can reason it has like the kind of relationship with
00:21:49
me that I would expect out of a really competent person that I worked with which is different from like a sycophant
00:21:56
yeah the thing in that world where if you had this like Jarvis like thing that can reason what do you think it does to
00:22:06
products that you use today where the interface is very valuable so for example if you look at an instacart or
00:22:12
if you look at an Uber or if you look at a door Dash these are not services that
00:22:16
are meant to be pipes that are just providing a set of apis to a Smart Set of agents that ubiquitously work on
00:22:24
behalf of 8 billion people what do you think has to change in how we think about how apps need to
00:22:29
work of how this entire infrastructure of experiences need to work in a world where you're agentically interfacing to
00:22:35
the world you I'm actually very interested in designing a world that is equally usable by humans and by AIS so I
00:22:44
I I I I like the interpretability of that I like the smoothness of the handoffs I
00:22:51
like the ability that we can provide feedback or whatever so you know door Dash could just expose some a pi to my
00:22:59
future AI assistant and they could go put the order and whatever or I could say like I could be holding my phone and
00:23:06
I could say okay AI assistant like you put in this order on door Dash please and I could like watch the app open and
00:23:12
see the thing clicking around and I could say hey no not this or like um there there's something about designing
00:23:19
a world that is usable equally well by humans and AIS that I think is a interesting concept
00:23:28
excited humanoid robots than sort of robots of like very other shapes the world is very much designed for humans
00:23:34
and I think we should absolutely keep it that way and a shared interface is nice
00:23:39
so you see voice chat that modality kind of gets rid of apps you just ask it for
00:23:44
sushi it knows Sushi you like before it knows what you don't like and does its best shot at doing it I it's hard for me
00:23:50
to imagine that we just go to a world totally where you say like hey Chachi BT order me sushi and it says okay do you
00:23:57
want it from this restaurant what kind what time whatever I think user I think visual user interfaces are super good
00:24:05
for a lot of things um and it's hard of me to imagine like a world where youd never look at a screen and
00:24:15
just use voice mode only but I I can't imagine that for a lot of things yeah I mean Apple tried with Siri like you
00:24:22
supposedly you can order an Uber automatically with Siri I don't think anybody's ever done it because it's why
00:24:27
would you take the risk of not the quality to your point the quality is not good but when the quality is good enough
00:24:33
you're you'll actually prefer it just because it's just lighter weight you don't have to take your phone out you
00:24:37
don't have to search for your app and press it and oh it automatically logged you out oh hold on log back in oh TFA
00:24:44
it's a whole pain in the ass you know it's like setting a timer with Siri I do every time because it works really well
00:24:50
and it's great and I don't need more information but ordering an Uber like I want to see the prices for a few
00:24:57
different options I want to see how far away it is I want to see like maybe even
00:25:01
where they are on the map because I might walk somewhere I get a lot more information by I think in less time by
00:25:07
looking at that order the Uber screen than I would if I had to do that all through the audio Channel I like your
00:25:12
idea of watching it happen that's kind of cool I think there will just be like yeah
00:25:18
different there are different interfaces we use for different tasks and I think that'll keep going of all the developers
00:25:23
that are building apps and experiences on open AI are there few that stand out for you where you're like okay this is
00:25:30
directionally going in a super interesting area even if it's like a toy app but are there things that you guys
00:25:36
point to and say this is really important um I met with a new company this morning or and bar even a company
00:25:45
it's like two people that are going to work on a summer project trying to actually finally make the AI tutor like
00:25:52
and I've always been interested in this space a lot of people have done great stuff on our platform but if if someone
00:25:57
can deliver like the way that you actually like H they used a phrase I love which
00:26:05
is this is going to be like a monor level reinvention for how people how people learn things yeah um but if you
00:26:11
can like find this new way to like let people explore and learn and new ways on their own I'm personally super excited
00:26:17
about that um a lot of the coding related stuff you mentioned Devon earlier I think that's like a super cool
00:26:23
vision of the future the thing that I am Health Healthcare I I believe should be pretty transformed by this but
00:26:32
the thing I'm personally most excited about is the sort of doing faster and better scientific discovery gp4 clearly
00:26:40
not there in a big way although maybe it accelerates things a little bit by making scientists more
00:26:45
productive but Alpha 43 yeah that's like but Sam that will be a Triumph those are
00:26:54
not like these these models are train Tred and built differently than the language models I mean to some obviously
00:27:04
there's a lot that's similar but there's a lot um there's kind of a groundup architecture to a lot of these models
00:27:09
that are being applied to these specific problem sets these specific applications
00:27:14
like chemistry interaction modeling for example does you you'll need some of that for sure but the the thing that I
00:27:21
think we're missing across the board for many of these things we've been talking
00:27:25
about is models that can do reason and once you have reasoning you can connect it to chemistry simulators or
00:27:32
guess yeah that's the important question I wanted to kind of talk about today was
00:27:37
this idea of networks of models people talk a lot about agents as if there's kind of this
00:27:44
linear set of call functions that happen but one of the things that arises in biology is networks of systems
00:27:53
that have cross interactions that the aggregation of the system the aggregation of the network produces an
00:28:00
output rather than one thing calling another that thing calling another do we see like an emergence in this
00:28:06
architecture of either specialized models or network models that work together to address bigger problem sets
00:28:12
use reasoning there's computational models that do things like chemistry or arithmetic and there's other models that
00:28:17
do rather than one model to rule them all that's purely generalized I don't know um
00:28:29
I don't know how much reasoning is going to turn out to be a super generalizable thing I suspect it
00:28:36
will but that's more just like an intuition and a hope and it would be nice if it worked out that way I I don't
00:28:42
know if that's like but let's walk through the the protein modeling example there's a bunch of training data
00:28:52
images of proteins and then sequence data and they build a model predictive model and they have a set of processes
00:28:59
and steps for doing that do you envision that there's this artificial general intelligence or this great reasoning
00:29:05
model that then figures out how to build that submodel that figures out how to solve that problem by acquiring the
00:29:10
necessary data and then resolving there so many ways where that could go like maybe it is it trains a literal model
00:29:17
for it or maybe it just like knows the one big model what it can like go pick what other training data it needs and
00:29:24
ask a question and then update on that um I guess the real question is are all these startups going to die because so
00:29:31
many startups are working in that modality which is go get special data and then train a new model on that
00:29:36
special data from the ground up and then it only does that one sort of thing and
00:29:40
it works really well at that one thing and it works better than anything else at that one thing you know there there's
00:29:44
like a version of this I think you can like already see when you were when you were talking about like biology and
00:29:53
these complicated networks of systems the reason I was smiling I I got super sick recently and I'm mostly better now
00:30:00
but it was just like body like got beat up like one system at a time F like you can really tell like okay it's this
00:30:05
cating thing and uh and that reminded me of you like talking about the like biology is just these like you have no
00:30:12
idea how much these systems interact with each other until things start going wrong and that was sort of like
00:30:17
interesting to see but I was using um I was like using chat GPT uh to try to like figure out like what was
00:30:26
happening whatever and and and would say well I'm you know unsure of this one thing and then I just like posted a
00:30:31
paper on it without even reading the paper um like in the context and it says oh that was the thing I was unsure of
00:30:37
like now I think this instead so there's like a that was like a small version of
00:30:41
what you're talking about where you can like can say this I don't I don't know this thing you can put more information
00:30:47
you don't retrain the model you're just adding it to the context here and now you're getting a so these models that
00:30:53
are predicting protein structure like let's say right this is the whole basis and now now other molecules at Alpha
00:31:01
3 can they can yeah I mean is it basically a world where the best generalized model goes in and gets that
00:31:09
training data and then figures out on its own and maybe you could maybe you could use an example for us can you tell
00:31:15
us about Sora your video model that generates amazing moving images moving video and and what's different about the
00:31:22
architecture there whatever you're willing to share on how how that is different
00:31:28
yeah so my on the general thing first my you clearly will need specialized simulators connectors pieces
00:31:42
of data whatever but my intuition and again I don't have this like backed up with science my intuition
00:31:49
would be if we can figure out the core of generalized reasoning connecting that to new problem domains in the same way
00:31:56
that humans are generalized reason ERS would I think be be doable it's like a fast unlock faster unlock than I think
00:32:05
I I think so um but yeah you Sora like does not start with a language model um it's that
00:32:15
that's a model that is like customized to do video uh and and so like we're clearly not at that world yet right so
00:32:24
you guys so just as an example for you guys to build a good video model you built it from scratch
00:32:30
using I'm assuming some different architecture and different data but in the future the generalized reasoning
00:32:39
system the the AGI whatever system theoretically could render that by figuring out how to do it yeah I mean
00:32:46
one example of this is like okay you know as far as I know all the best text models in the world are still a lot of
00:32:52
regressive models and the best image and video models are diffusion models and that's like sort strange in some sense
00:32:59
yeah yeah so there's a big debate about uh training data you guys have been I think the most thoughtful of any company
00:33:08
you've got licensing deals now ft Etc and we got to gu be gentle here because you're involved in a New York Times
00:33:16
lawsuit you weren't able to settle I guess an arrangement with them for training data how do you think about
00:33:23
fairness in fair use we've had big debates here on the pod obviously your actions are you know
00:33:30
speak volumes that you're trying to be fair by doing licensing deals so what what's your personal position on the
00:33:37
rights of artists who create beautiful music lyrics books and you taking that and then making a derivative product out
00:33:46
of it and and then monetizing it and and what's fair here and how do we get to a
00:33:52
world where you know artists can make content in the world and then decide what they want other people to do with
00:33:59
it yeah and and I'm just curious your personal belief because I know you to be a thoughtful person on this and I know a
00:34:04
lot of other people in our industry are not very thoughtful about how they think
00:34:08
about content creators so I think it's very different for different kinds of I mean look on
00:34:14
unfair use I think we have a a very reasonable position under the current law but I think AI is so different that
00:34:22
for things like art we'll need to think about them in different ways but I would
00:34:25
say if you go read a bunch of math on the internet and learn how to do math that I think seems unobjectionable to
00:34:37
most people and then there's like you know another set of people who might have a different opinion well what if
00:34:41
you like actually let me not get into that just in the interest of not making this
00:34:48
answer too long so I think there's like one category people are like okay there's like generalized human knowledge
00:34:54
you can kind of like go if you learn that like that's that that's like open domain or
00:34:59
something if you kind of go learn about the Pythagorean theorem um that's one end of the spectrum and then I think the
00:35:06
other extreme end of the spectrum is um is Art and maybe even like more than more specifically I would say it's like
00:35:16
doing it's a system generating art in the style or the likeness of another artist um would be kind of the furthest
00:35:24
end of that and then there's many many cases on the Spectrum in between uh I think the conversation has been
00:35:33
historically very caught up on training data but it will increasingly become more about what happens at inference
00:35:40
time as training data becomes less valuable and the what the system does accessing you know
00:35:52
information in in context in real time or uh you know know taking like like something like that what happens at
00:36:00
inference time will become more debated and and how the what the new economic model is there so if you say like
00:36:08
uh if you say like create me a song in this in the style of Taylor Swift even if the model were never
00:36:16
trained on any Taylor Swift songs at all you can still have a problem which is it
00:36:21
may have read about Taylor Swift it may know about her themes Taylor Swift means
00:36:24
something and then and then the question is like that model even if it were never
00:36:28
trained on any Taylor Swift song whatsoever be allowed to do that and if so um how should Taylor get paid right
00:36:38
so I think there's an optin opt out in that case first of all and then there's an economic model um staying on the
00:36:43
music example there is something interesting to look at from the historical perspective here which is uh
00:36:50
sampling and how the economics around that work this is not quite the same thing but it's like an interesting place
00:36:55
to start looking Sam let me just challenge that what's the difference in the example you're giving of the model
00:37:01
learning about things like song structure Tempo Melody Harmony relationships all the discovering all
00:37:09
the underlying structure that makes music successful and then building new music using training data and what a
00:37:16
human does that listens to lots of music learns about and and their brain is processing and building all those same
00:37:23
sort of predictive models or those same sort of uh discoveries understandings what's the difference here and why why
00:37:30
are you making the case that perhaps artists should be uniquely paid this is not a sampling situation you're not the
00:37:37
AI is not outputting and it's not storing in the model the actual original song it's learning structure right I
00:37:44
wasn't trying to make that that point because I agree like in the same way that humans are inspired by other humans
00:37:50
I was saying if you if you say generate me a song in the style of Taylor Swift I
00:37:54
see right okay where the prompt leverages some artist I I think personally that's a different case would
00:38:01
you be comfortable asking or would you be comfortable letting the model train itself well a music model being trained
00:38:08
on the whole Corpus of music that humans have created without royalties being paid to the artists that um that music
00:38:15
is being fed in and then you're not allowed to ask you know artists specific prompts you could just say hey pay me a
00:38:20
play me a a really cool pop song that's fairly modern about heartbreak you know with a female voice you know we have
00:38:27
currently made the decision not to do music and partly because exactly these questions of where you draw the lines
00:38:34
and you know what like even I was meeting with several musicians I really admire recently I was
00:38:41
just trying to like talk about some of these edge cases but even the world in which if
00:38:48
we went and let's say we paid 10,000 musicians to create a bunch of music just to make a great training set where
00:38:57
the music music model could learn everything about strong s structure um and what makes a good catchy beat and
00:39:06
everything else um and only trained on that let's say we could still make a great music model which maybe maybe we
00:39:11
could um you know I was kind of like posing that as a thought experiment to musicians and they're like well I can't
00:39:17
object to that on any principal basis at that point um and yet there's still something I don't like about it now
00:39:22
that's not a reason not to do it um necessarily but it is did you see that ad that Apple put out
00:39:30
maybe it was yesterday or something of like squishing all of human creativity down into one really thin iPad what was
00:39:35
your take on it uh people got really emotional about it yeah it's stronger reaction than you
00:39:42
would think yeah there's something about I'm obviously hugely positive on AI but there is something that I think
00:39:53
is beautiful about human creativity and human artistic expression and and you know for an AI that just does better
00:39:59
science like great bring that on but an AI that is going to do this like deeply beautiful human creative expression I
00:40:05
think we should like figure out it's going to happen it's going to be a tool that will lead us to Greater creative
00:40:12
Heights but I think we should figure out how to do it in a way that like preserves the spirit of what we all care
00:40:17
about here and I I think your actions speak loudly we were trying to do Star Wars characters in dolly
00:40:27
and if you ask for Darth Vader it says hey we can't do that so you've I guess red teed or whatever you call it
00:40:34
internally yeah you're not allowing people to use other people's IP so you've taken that decision now if you
00:40:40
asked it to make a Jedi Bulldog or a Sith Lord Bulldog which I did it made my Bulldogs a Sith bulldogs so there's an
00:40:47
interesting question about your spectrum right yeah you know we put out this thing yesterday called the spec um where
00:40:53
we're trying to say here are here's here's how our model is supposed to behave and it's very hard it's a long
00:41:01
document it's very hard to like specify exactly in each case where the limit should be and I view this as like a
00:41:07
discussion that's going to need a lot more input um but but these sorts of questions
00:41:14
about okay maybe it shouldn't generate Darth Vader but the idea of a Sith Lord or a Sith style thing or Jedi at this
00:41:20
point is like part of the culture like like these are these are all hard decisions yeah and and I think you're
00:41:27
right the music industry is going to consider this opportunity to make tlor Swift songs their opportunity it's part
00:41:33
of the four-part fair use test is you know these who gets to capitalize on new Innovations for existing art and and
00:41:41
Disney has an argument that hey you know if if you're GNA make Sora versions of a
00:41:47
aoka or whatever Obi-Wan Kenobi that's Disney's opportunity and that's a great partnership for
00:41:53
you you know to pursue so we're I think this section I would label as AI in the law so let me ask maybe a higher level
00:42:01
question what does it mean when people say regulate AI totally Sam what does it what does that even mean and comment on
00:42:09
California's new proposed regulations as well a few if you're up for it uh I'm concerned I mean there's so many
00:42:17
proposed regulations but most of the ones I've seen on the California state things I'm concerned about I also have a
00:42:22
general fear of the states all doing this them themselves um when people say regulate AI I don't think they mean one
00:42:30
thing I think there's like some people are like B the whole thing some people like don't allow it to be open source
00:42:35
required to be open source um the thing that I am personally most interested in is I think there will
00:42:43
come look I may be wrong about this I will acknowledge that this is a forward-looking statement and those are
00:42:48
always dangerous to make but I think there will come a time in the not super distant future like you know we're not
00:42:53
talking like decades and decades from now where AI say the frontier AI systems are capable of causing
00:43:02
significant Global harm and for those kinds of systems in the same way we have like Global oversight of nuclear weapons
00:43:11
or synthetic bio or things that can really like have a very negative impact Way Beyond the realm of one country uh I
00:43:18
would like to see some sort of international agency that is looking at the most powerful systems and ensuring
00:43:24
like reasonable safety testing you know these things things are not going to escape and recursively self-improve or
00:43:30
whatever the criticism of this is that you're you have the resources to Cozy up to Lobby
00:43:39
to be involved and you've been very involved with politicians and then startups which are also passionate about
00:43:44
and invest in um are not going to have the ability to Resource uh and deal with this and that this regulatory capture as
00:43:51
per our friend you know Bill Gurley did a great talk last year about it so maybe
00:43:55
you could address that headon do do you if the line were we're only going to look at models that are trained on
00:44:01
computers that cost more than 10 billion or more than 100 billion or whatever dollars I'd be fine with that there'd be
00:44:07
some line that'd be fine and uh I don't think that puts any regulatory burden on
00:44:11
startups so if you have like the the nuclear raw material to make a nuclear bomb like there's a small subset set of
00:44:18
people who have that therefore you use the analogy of like a a nuclear inspectors kind of situation yeah I
00:44:23
think that that's interesting sax you have a question well go ahead you had to follow can I say one more thing about
00:44:30
that of course I'd be super nervous about regulatory overreach here I think we can get this wrong by doing way too
00:44:36
much I or even a little too much I think we can get this wrong by doing not enough
00:44:41
but but I do think part of and I and now I mean you know we have seen regulatory overstepping or capture just
00:44:52
get super bad in other areas um and you know like also maybe nothing will happen
00:44:58
but but I think it is part of our duty and our mission to like talk about what we believe is likely to happen and what
00:45:06
it takes to get that right the challenge Sam is that we have statute that is meant to protect people protects Society
00:45:14
at large what we're creating however a statute that gives the government rights to go in and audit code to audit
00:45:24
business um trade Secrets uh we've never seen that to this degree before basically the California legislation
00:45:33
that's proposed and some of the federal legislation that's been proposed basically requires the fed the
00:45:39
government to audit a model to audit software to audit and review the parameters and the weightings of the
00:45:44
model and then you need their check mark in order to deploy it for commercial or
00:45:51
public use and for me it just feels like we're trying to reigning the the the government agencies
00:46:00
for fear and and because folks have a hard time understanding this and are scared about the implications of it they
00:46:06
want to control it and because they want and the only way to control it is to say
00:46:10
give me a right to audit before you can release it asess these people cluel I mean the way that the the stuff is
00:46:15
written you read it you're like G to pull your hair out because as you know better than anyone in 12 months none of
00:46:20
this stuff's going to make sense anyway totally right look the reason I have pushed for an agency based approach for
00:46:26
for for kind of like the big picture stuff and not a like write it in laws I don't in 12 months it will all be
00:46:32
written wrong and I don't think even if these people were like True World experts I don't think they could get it
00:46:39
right looking at 12 or 24 months um and I don't these policies which is like we're going to look at you know we're
00:46:47
going to audit all of your source code and like look at all of your weights one by one like yeah I think there's a lot
00:46:53
of crazy proposals out there um by the way especially if the models are always retrained all the time if they become
00:46:58
more Dynamic again this is why I think it's yeah but but like when before an airplane gets certified there's like a
00:47:05
set of safety tests we put the airplane through it um and totally it's different
00:47:10
than reading all of your code that's reviewing the output of the model not reviewing the insides of the model and
00:47:17
and so what I was goingon to say is I that is the kind of thing that I think as safety testing makes sense how are we
00:47:24
going to get that to happen Sam and I'm not just speaking for open AI I speak for the industry for for Humanity
00:47:29
because I am concerned that we draw ourselves into almost like a dark ages type of era by restricting the growth of
00:47:37
these incredible technologies that can prosper human that that Humanity can prosper from so significantly how do we
00:47:43
change the the sentiment and get that to happen because this is all moving so quickly at the government levels and
00:47:48
folks seem to be getting it wrong and I'm I'm just to build on that Sam the architectural decision for example that
00:47:55
llama took is pretty interesting in that it's like we're going to let llama grow
00:48:01
and be as unfettered as possible and we have this other kind of thing that we call Llama guard that's meant to be
00:48:06
these protective guard rails is that how you see the problem being solved correctly or do you see at the current
00:48:12
at the current strength of models definitely some things are going to go wrong and I don't want to like make
00:48:18
light of those or not take those seriously but I'm not like I don't have any like catastrophic risk worries with
00:48:25
a gp4 level model um and I think there's many safe ways to choose to deploy this
00:48:34
uh may maybe we'd find more common ground if we said that uh and I like you know the specific example of models that
00:48:44
are capable that are technically capable not even if they're not going to be used
00:48:48
this way of recursive self-improvement um or of you know autonomously designing and deploying a bioweapon or
00:49:00
something like that or a new model that was the recursive self-improvement Point
00:49:05
um you know we should have safety testing on the outputs at an international level for models that you
00:49:12
know have a reasonable chance of of posing a threat there uh I don't think like GPT 4 of course does
00:49:22
not POS any sort of well I want to say any sort cuz we don't yeah I don't think the gp4 poses a material threat on those
00:49:31
kinds of things and I think there's many safe ways to release a model like this um but you know when like significant
00:49:39
loss of human life is a serious possibility like airplanes or any number of other examples where I
00:49:48
think we're happy to have some sort of testing framework like I don't think about an airplane when I get on it I
00:49:52
just assume it's going to be safe right right there's a lot of hand ringing right now Sam about
00:49:58
jobs and you had a lot of I think you did like some sort of a test when you were at YC about Ubi and you've been
00:50:04
result in that come out very soon I just it was a fiveyear study that wrapped up
00:50:09
um or started five years ago well there was like a beta study first and then it was like a long one that ran but uhk can
00:50:16
you explain yeah why did you start why' you start it maybe just explain Ubi and why you started it um so we started
00:50:22
thinking about this in 2016 uh kind of about the same time started taking AI really seriously and the theory was that
00:50:31
the magnitude of the change that may come to society and jobs in the economy and and sort of
00:50:40
in some deeper sense than that like what the social contract looks like um meant that we should have many
00:50:47
studies to study many ideas about new new ways to arrange that um I also think that you
00:50:56
know I'm not like a super fan of how the government has handled most policies designed to help poor people and I kind
00:51:03
of believe that if you could just give people money they would make good decisions and the market would do its
00:51:09
thing and you know I'm very much in favor of lifting up the floor and reducing eliminating poverty um but I'm
00:51:18
interested in better ways to do that than what we have tried for the existing social safety net and and kind of the
00:51:24
way things have been handled and I think people money is not going to go solve all problems it's certainly not going to
00:51:30
make people happy but it might it might solve some problems and it might give people a better Horizon
00:51:40
with which to help themselves and I'm interested in that I I think that now that we see some of the ways so
00:51:48
2016 was a very long time ago uh you know now that we see some of the ways that AI is developing I wonder if
00:51:54
there's better things to do than the [Music] traditional um conceptualization of Ubi
00:52:02
uh like I wonder I wonder if the future looks something like more like Universal basic
00:52:07
compute than Universal basic income and everybody gets like a slice of gpt7 compute and they can use it they can
00:52:14
resell it they can donate it to somebody to use for cancer research but but what
00:52:18
you get is not dollars but this like productivity slice yeah you own like part of the productivity right I would
00:52:24
like to shift to the gossip part of this gossip what gossip let's go back let's go back to November what the flying
00:52:35
happened um you know I I if you have specific questions I'm happy to maybe I'll answer maybe you said you were
00:52:43
going to talk about it at some point so here's the point what the hell happened you were fired you came back it was a
00:52:49
palace Intrigue did somebody stab you in the back did you find AGI what's going on tell us this is a safe face
00:52:57
Sam um I was fired I was I talked about coming back I kind of was a little bit unsure at the moment
00:53:07
about what I wanted to do because I was very upset um and I realized that I really loved open Ai and the people and
00:53:17
that I would come back and I kind of I knew it was going to be hard it was even harder than I thought but I I kind of
00:53:25
was like all right fine um I agreed to come back um the board like took a while to figure things out and then uh you
00:53:33
know we were kind of like trying to keep the team together and keep doing things
00:53:38
for our customers and uh you know sort of started making other plans then the board decided to hire a different
00:53:43
interim CEO um and then everybody there many people oh my gosh what was what was that guy's name he was
00:53:50
there for like a scaramucci right like uh em great and I I have nothing but good
00:53:58
scari um and then where were you when they um when you found the news that you'd been fired like taking I was in a
00:54:07
hotel room in Vegas for F1 weekend I think text and they're like fire pick up said I think that's happened to you
00:54:14
before J I'm trying to think if I ever got fired I don't think I've gotten fired um yeah I got no it's just a weird
00:54:21
thing like it's a text from who actually no I got a text the night before and then I got in a phone call with the
00:54:27
uh and then that was that and then I kind of like I mean then everything went crazy I was like uh it was
00:54:35
like I mean I have my phone was like unusable it was just a Non-Stop vibrating thing of like text messages
00:54:41
call basically you got fired by tweet that happened a few times during the Trump Administration a few uh C TW
00:54:49
before tweeting was nice of them um and then like you know I kind of did like a few hours of just this like absolute
00:54:56
State um in the hotel room trying to like I was just confused beyond belief trying to figure out what to do and uh
00:55:05
so weird and then like flew home it maybe like got on a plane like I don't know 3 p.m. or
00:55:13
something like that um still just like you know crazy non-stop phone blowing up uh met up with some people in person by
00:55:21
that evening I was like okay you know I'll just like go do AGI research and was feeling pretty happy about the
00:55:29
future and yeah you have options and then and then the next morning uh had this call with a couple of board members
00:55:36
about coming back and that led to a few more days of craziness and then uh and then it kind of I think it got
00:55:48
resolved well it was like a lot of insanity in between what percent what percent of it was because of these
00:55:54
nonprofit board members um well we only have a nonprofit board so it was all the nonprofit board
00:56:01
members uh there the board had gotten down to six people um they and then they removed Greg from the
00:56:10
board and then fired me um so but it was like you know but I mean like was there
00:56:16
a culture clash between the people on the board who had only nonprofit experience versus the people who had
00:56:21
startup experience and maybe you can share a little bit about if you're willing to the motivation behind the
00:56:26
action anything you can I think there's always been culture clashes at look obviously not all of those board members
00:56:39
are my favorite people in the world but I have serious respect for the gravity with which they treat
00:56:50
AGI and the importance of getting AI safety right and even if I stringently disagree with their
00:56:59
decision- making and actions which I do um I have never once doubted their integrity or commitment to um the
00:57:11
sort of shared mission of safe and beneficial AGI um you know do I think they like
00:57:17
made good decisions in the process of that or kind of know how to balance all of things opening I has to get right no
00:57:24
but but I think that like the intent the intent of the magnitude of yeah AGI and getting that right I actually
00:57:36
let me ask you about that so the mission of open AI is explicitly to create AGI which I think is really
00:57:43
interesting a lot of people would say that if we create AGI that would be like an unintended consequence of something
00:57:51
gone horribly wrong and they're very afraid of that outcome but open a makes that the actual Mission yeah does that
00:58:00
create like more fear about what you're doing I mean I understand it can create motivation too but how do you reconcile
00:58:06
that I guess why is I think a lot of I think a lot of the well I mean first I'll say I'll answer
00:58:12
the first question and the second one I think it does create a great deal of fear uh I think a lot of the world is
00:58:19
understandably very afraid of AGI or very afraid of even current Ai and and very excited about it and even more
00:58:26
afraid and even more excited about where it's going um and we we wrestle with that but like I think
00:58:35
it is unavoidable that this is going to happen I also think it's going to be tremendously beneficial but we do have
00:58:41
to navigate how to get there in a reasonable way and like a lot of stuff is going to change and change is you
00:58:47
know pretty pretty uncomfortable for people so there's a lot of pieces that we got to get right and ask
00:58:56
can I ask a different question you you have created I mean it's the hottest company and you
00:59:04
are literally at the center of the center of the center but then it's so unique in the
00:59:12
sense that all of this value you issued economically can you just like walk us through like yeah I wish I had taken I
00:59:19
wish I had taken Equity so I never had to answer this question if I could go back in why don't they give you a grand
00:59:24
now or just give you a big option Grant like you deserve yeah give you five points what was the decision back then
00:59:31
like why was that so important the decision back then the re the original reason was just like the structure of
00:59:36
our nonprofit it was uh like there was something about yeah okay this is like nice from a motivations perspective but
00:59:45
mostly it was that our board needed to be a majority of disinterested directors and I was like that's fine I
00:59:51
don't need Equity right now I kind of but like but in this weird way now that you're running a company yeah it it
00:59:59
creates these weird questions of like well what's your real motivation versus to that's that it is so deeply un I one
01:00:07
thing I have noticed it is is so deeply unimaginable to people to say I don't really need more
01:00:13
money like and I well people think I think I think people think it's a little bit of an ulterior motive I think yeah
01:00:20
yeah yeah no so it assumes what else is he doing on the side to make money something if I were just trying to say
01:00:26
like I'm going to try to make a trillion dollarss with open AI I think everybody
01:00:29
would have an easier time and it would save me it would save a lot of conspiracy theories totally this is
01:00:34
totally the back Channel you are a great dealmaker I I've watched your whole career I mean you're just great at it
01:00:42
you got all these connections you're really good at raising money uh you're fantastic at it and you got this Johnny
01:00:49
I thing going you're inhumane you're investing in company she got the orb raising $7 trillion to build
01:00:56
Fabs all this stuff all of that put together J loves fake news I'm kind of being a little fous here you know
01:01:04
obviously it's not you're not raising 7 trillion doll but maybe that's the market cap of something putting all that
01:01:08
aside the T was you're doing all these deals they don't trust you because what's your motivation you you your end
01:01:16
running and and what opportunities belong inside of open AI what opportuni should be Sam's and this group of
01:01:23
nonprofit people didn't trust you is that what happened so the things like you know device companies or if we were
01:01:29
doing some chip Fab Company it's like those are not Sam project those would be like opening ey would get that Equity um
01:01:36
they would okay that's not the Public's perception well that's not like kind of the people like you who have to like
01:01:42
commentate on the stuff all day perception which is fair because we haven't announced the stuff because it's
01:01:45
not done I don't think most people in the world like are thinking about this but I I I agree it spins up a lot of
01:01:53
conspiracies conspiracy theories in like Tech commentators yeah and if I could go
01:01:59
back yeah I would just say like let me take equity and make that super clear and then every be like all right like
01:02:05
I'd still be doing it because I really care about AGI and think this is like the most interesting work in the world
01:02:10
but it would at least type check to everybody what's the chip project that the S trillion do and where' the Seven
01:02:16
trillion number come from makes no sense I don't know where that came from actually I genuinely don't uh I think I
01:02:21
think the world needs a lot more AI infrastructure a lot more than it's currently planning to build and with a
01:02:28
different cost structure um the exact way for us to play there is we're still trying to figure that out got it what's
01:02:36
your preferred model of organizing open AI is it sort of like the move fast break
01:02:43
things highly distributed small teams or is it more of this organized effort where you need to plan because you want
01:02:49
to prevent some of these edge cases um oh I have to go in a minute uh it's not because
01:02:58
it's not to prevent the edge cases that we need to be more organized but it is that these systems are so complicated
01:03:04
and concentrating bets are so important like one you know at the time before it was
01:03:11
like obvious to do this you have like deep mind or whatever has all these different teams doing all these
01:03:16
different things and they're spreading their bets out and you had Open the Eyes say we're going to like basically put
01:03:20
the whole company and work together to make gp4 and that was like unimaginable for how to run an AI research lab but it
01:03:28
is I think what works at the minimum it's what works for us so not because we're trying to prevent edge cases but
01:03:34
because we want to concentrate resources and do these like big hard complicated things we do have a lot of coordination
01:03:40
on what we work on all right Sam I know you got to go you've been great on the hour come back any time great talking to
01:03:46
you guys yeah fun thanks for for being so open about it we've been talking about it for like a year plus I'm really
01:03:53
happy it finally happened yeah it's awesome I really app come back on after our next like major launch and I'll be
01:03:57
able to talk more directly about some you got the zoom link same Zoom link every week just same time same Zoom link
01:04:03
drop time just drop in just put on your calendar come back to the game come back
01:04:08
to the game in a while I you know I would love to play poker it has been forever that would be a lot of fun send
01:04:16
where Cham when you and I were heads up and you you had remind me you and I were
01:04:21
heads up and you went all in I had a set but there was a straight and a flush on
01:04:27
the board and I'm in the tank trying to figure out if I want to lose this back when we playing small Stakes it might
01:04:32
have been like 5K pot or something and then chath can't stay out of the pot and he starts taunting the two of us you
01:04:39
should call you shouldn't call he's bluffing and I'm like I'm going I'm trying to figure out if I make the call
01:04:45
here I make the call and uh it was like uh you had a really good hand and but I just happened to have a s I think you
01:04:51
had like top pair top kicker or something but you you made a great move because the boy was so almost like a
01:04:56
bottom set Sam has a great style of playing which I would call RAM and jam totally you got to get I don't really
01:05:02
know if you I don't I don't know if you could say about anybody else I don't I don't I'm not gonna you
01:05:07
haven't seen Jam play in the last 18 months it's a lot different much more so much fun now find
01:05:14
find hard to have you played bomb pots I don't know what that is okay you'll love
01:05:19
it we'll see you is nuts it's do two boards and congrats everything honestly than thanks for
01:05:27
coming on and love to have you back when the next after the big launch sounds please do cool bye gentlemen some
01:05:34
breaking news here all those projects He said are part of open AI That's something people didn't know before this
01:05:40
and a lot of confusion there chamat what was your major takeaway from our hour with Sam I think
01:05:47
that these guys are going to be one of the four major companies okay that matter in this whole space I think that
01:05:54
that's clear I think what's still unclear is where is the economics going to be he said
01:06:00
something very discreet but I thought was important which is I think he basically my interpretation is these
01:06:06
models will roughly all be the same but there's going to be a lot of scaffolding
01:06:11
around these models that actually allow you to build these apps so in many ways that is like the open source movement so
01:06:17
even if the model itself is never open source it doesn't much matter because you have to pay for the infrastructure
01:06:24
right there's a lot of open Source software that runs on Amazon you still pay AWS something so I think the right
01:06:30
way to think about this now is the models will basically be all really good and then it's all this other
01:06:38
stuff that you'll have to pay for interface whoever builds all this other stuff is going to be in a position to
01:06:46
build a really good business freeberg he talked a lot about reasoning it seemed like that he kept going to reasoning and
01:06:52
away from the language model did you not did you note that and anything else that
01:06:55
you noted in our hour with yeah I mean that's a longer conversation because there is a lot of talk about language
01:07:00
models eventually evolving to be so generalizable that they can resolve pretty much like all intelligent
01:07:09
function and so the language model is the foundational model that that yields AGI but that's a I think there's a lot
01:07:16
of people that at different schools of thought on this and how much my my other takeaway I think is that the I think
01:07:24
what he also seem to indicate is there's like so many like we're also enraptured
01:07:30
by llms but there's so many things other than llms that are being baked and rolled by him and by other groups and I
01:07:37
think we have to pay some amount of attention to all those because that's probably where and I think freedberg you
01:07:43
tried to go there in your question that's where reasoning will really come from is this mixture of experts approach
01:07:48
and so you're going to have to think multi-dimensionally to reason right we do that right do I cross the street or
01:07:54
not in this point in time you reason based on all these multi-inputs and so there's there's all these little systems
01:08:00
that go into making that decision in your brain and if you if you use that as a a simple example there's all this
01:08:05
stuff that has to go into making some experience being able to reason intelligently sax you went right
01:08:12
there with the corporate structure the board and uh he he he gave us a lot more information here what are your thoughts
01:08:21
on the hey you know the chip stuff and the other stuff I'm working that's all part of open AI people just don't
01:08:27
realize it and that moment and then you know your questions to him about Equity your thoughts on um I'm not sure I was
01:08:35
like the main guy who asked that question jakal but um well no you did talk about the the nonprofit that the
01:08:41
difference between question about the clearly was some sort of culture Clash on the board between the the people who
01:08:49
originated from the nonprofit world and people who came from the startup world we don't really know more than that but
01:08:54
there clearly was some sort of culture class I thought one of the a couple of the
01:08:58
other areas that he drew attention to that were kind of interesting is he clearly thinks there's a big opportunity
01:09:02
on mobile that goes beyond just like having you know a chat gbt app on your phone or maybe even having like a Siri
01:09:10
on your phone there's clearly something bigger there he doesn't know exactly what it is but it's going to require
01:09:16
more inputs it's that you know personal assistant that's seeing everything around you help really I think that's a
01:09:22
great Insight David because he was talking about hey I'm looking for a senior team member who can push back on
01:09:29
me and understands all context I thought that was like a very interesting to think about an executive assistant or an
01:09:35
assistant that's has executive function as opposed to being like just an alter ego for you or what he called a
01:09:42
sycophant that's kind of interesting I thought that was interesting yeah yeah and clearly he thinks there's a big
01:09:47
opportunity in biology and scientific discovery after the break I think we should talk about Alpha fold 3 it was
01:09:52
just announ let's do that and we can talk about the the Apple ad and I just want to also make sure people
01:09:57
understand when people come on the Pod we don't show them questions they don't edit the transcript nothing is out of
01:10:03
bounds if you were wondering why I didn't ask or we didn't ask about the Elon lawsuit he's just not going to be
01:10:08
able to comment on that so it' be no comment so you know and we're not like our time was limited and there's a lot
01:10:14
of questions that we could have asked him that would have just been a waste of time and com so I just want to make sure
01:10:19
people understand of course he's going to no comment in any lawsuit and he's already been asked about that 500 times
01:10:25
yes should we take a quick break before the next before we come back yeah I'll take a bio break and then we'll come
01:10:28
back with some news for you and some more banter with your favorite besties on the number one
01:10:36
podcast in the world the only podcast all right welcome back everybody second half of the show great guest Sam mman
01:10:41
thanks for coming on the Pod we've got a bunch of news on the docket so let's get
01:10:46
started freeberg you told me I could give some names of uh the guests that we've booked for the all in Summit I did
01:10:54
not you did you've said each week every week that I get to say I did not I appreciate your interest in the all in
01:11:02
Summits lineup but we do not yet have uh enough critical math uh to feel like we should go out there well uh I am
01:11:11
a loose canidate so I will announce my two guests and I created The Summit and you took it from me so and done a great
01:11:18
job I will announce my guests I don't care what your opinion is I have booked two guests for the summit and it's going
01:11:24
to be out look at these two guests I booked for the third time coming back to the summit our guy Elon Musk will be
01:11:31
there hopefully in person if not you know from 40,000 feet on a starling connection wherever he is in the world
01:11:36
and for the first time our friend Mark Cuban will be coming and so two great guests for you to look forward to but
01:11:45
free BG's got like a thousand guests coming he'll tell you when it's like 48 hours before the conference but yeah two
01:11:51
great speaking of billionaires who are coming isn't coming too yes coming yes he's booked so we have three
01:11:57
billionaires three billionaires yes okay hasn't fully confirmed so don't okay well we're going to say it anyway has
01:12:03
penciled in back we say penciled yeah don't back out this is going to be catnip for all these protest organizers
01:12:10
like if youke the bear well by the way speaking of updates what do you guys think of the
01:12:17
bottle for the all-in tequila oh beautiful honestly honestly I will just say I think you are doing a marvelous
01:12:24
job that I was shocked at the design shocked meaning it is so unique and high quality I think
01:12:33
it's amazing it would make me drink tequila you're going to You're Gonna Want to going to it it is uh stunning
01:12:41
just congratulations and um yeah it was just when we went through the deck at the uh at the monthly meeting it was
01:12:49
like oh that's nice oh that's nice we're going to do the concept bottles and then
01:12:52
that bottle came up and everybody went like crazy it was like somebody hitting like a Steph Curry hitting a half court
01:12:58
shot it was like oh my God it was just so clear that you've made an iconic bottle that if we can produce it oh Lord
01:13:07
it is going to be looks like we can the make it yeah it's gonna be amazing I'm excited I'm excited for it you know it's
01:13:14
like the Des so complicated that we have to do a feasibility analysis on whether
01:13:18
it was actually manufacturable but it is so or at least the early reports are good so we're going to H hopefully we'll
01:13:25
have some a for the in time for the all Summit I mean why not sounds I mean it's
01:13:30
great when we get barricaded in by all these protesters we can drink the tequila did you guys see did you see
01:13:36
Peter te Peter teal got barricaded by these ding-dongs at Cambridge my god listen people have the right to protest
01:13:42
I think it's great people are protesting but surrounding people and threatening them is a little bit over the top and D
01:13:49
think you're exaggerating what happened well I don't know exactly what happened because all we see is these videos look
01:13:54
they're not threatening anybody and I don't even think they tried to barricade him in they were just outside the
01:13:59
building and because they were blocking the driveway his car couldn't leave but he
01:14:05
wasn't physically like locked in the building or something yeah that's that's what the headlines say but that could be
01:14:12
fake news fake Social yeah this was not on my bingo card this Pro protester support by saaks was not on the bingo
01:14:19
card I got to say I contitution the Constitution of the United States in the Amendment provides for the right of
01:14:27
assembly which includes protest and sit in as long as they're as long as they're
01:14:31
Peaceable now obviously if they go too far and they vandalize or break into buildings or use violence then that's
01:14:38
not Peaceable however expressing sentiments with which you disagree does not make it violent and there's all
01:14:46
these people out there now making the argument that if you hear something from a protester that you don't like and you
01:14:54
subjectively experience that as a as a threat to your safety then that somehow should be you know treated as valid like
01:15:02
that's basically violent well that's that's not what the Constitution says and these people understood well just a
01:15:09
few months ago that that was basically snowf fakery that you know just because somebody you know what I'm
01:15:16
saying we have the rise of the woke right now where they're buying yeah the woke right they're buying into this idea
01:15:22
of safetyism which is being exposed ideas you don't like to protest you don't like is a threat to your safety no
01:15:28
it's not every we absolutely have snow fakery on both sides now it's ridiculous the
01:15:35
only thing I will say that I've seen and is this this this uh surrounding individuals who you don't want there and
01:15:44
locking them in a circle and then moving them out of like protesta that's not cool yeah obviously you can't do that
01:15:49
but look I think that most of the protests on most of the campuses have not crossed the line they've just
01:15:55
occupied The Lawns of these campuses and look I've seen some troublemakers try to
01:16:01
barge through the the encampments and claim that because they can't go through there that somehow they're being
01:16:08
prevented from going to class look you just walk around the lawn and you can get to class okay and you know some of
01:16:15
these videos are showing that these are effectively right-wing provocators who are engaging in leftwing tactics and I
01:16:24
don't support it either way by the way some of these camps are some of the funniest things you've ever seen it's
01:16:29
like there are like a one tent that's dedicated to like a reading room and you go in there and there's like these like
01:16:37
Center oh my God it's unbelievably hilarious look there there's no question that because the protests are
01:16:42
originating on the left that there's some goofy views like you know you're dealing with like a leftwing idea
01:16:48
complex right but and you know it's easy to make fun of them doing different things but the fact of the matter is
01:16:56
that most of the protests on most of these campuses are even though they can be annoying because they're occupying
01:17:01
part of the lawn they're not violent yeah and you know the way they're being cracked down on they're sending the
01:17:06
police in at 5:00 a. to crack down on these encampments with batons and riot gear and I find that part to be
01:17:14
completely excessive well it's also dangerous because you know things can escalate when you have mobs of people
01:17:21
and large groups of people so I just want to make sure people understand that large group of people large you have a
01:17:26
diffusion of responsibility that occurs when there's large groups of people who are passionate about things and and
01:17:31
people could get hurt people have gotten killed at these things so just you know
01:17:35
keep it calm everybody I agree with you like what's the harm of these folks protesting on a lawn it's not a big deal
01:17:40
when they break into buildings of course yeah that crosses the line obviously yeah but I mean let them sit out there
01:17:45
and then they'll run out their food cars their food card and they run out of waffles did you guys see the clip I
01:17:51
think it was on the University of Washington campus where one kid challenged this antifa guy to a push-up
01:17:58
contest oh fantastic I mean it's it is some of the funniest stuff some of some content is
01:18:05
coming out that's just my favorite was the woman who came out and said that the Columbia students needed humanitarian
01:18:11
Aid and oh my God the overdubs on her were hilarious I was like uh humanitarian Aid the he was like we need
01:18:19
our door Dash right now we need we Double Dash some boba and we can't get it through the the police need our Boba
01:18:26
low sugar Boba with the with the popping Boba bubbles wasn't getting in but you know people have the right to protest
01:18:32
and uh Peaceable by the way there's a word I've never heard very good sax Peaceable inclined to avoid argument or
01:18:39
violent conflict very nice well it's in the Constitution it's in the first amendment is it really I've never I
01:18:44
haven't heard the word Peaceable before I mean you and I are sympatico on this like I I don't we used to have the
01:18:51
ACLU like backing up the KKK going down Main Street and really fighting decision
01:18:57
yeah they were really fighting for I I'm and I have to say the Overton window is
01:19:02
opened back up and I think it's great all right we got some things on the docket here I don't know if you guys saw
01:19:07
the Apple new iPad ad it's getting a bunch of criticism they use like some giant hydraulic press to crush a bunch
01:19:15
of creative tools EJ turntable trumpet piano people really care about Apple's ads and what they represent we talked
01:19:24
about that that uh Mother Earth little vignette they created here what do you think free BR did you see the ad what
01:19:30
was your reaction to it made me sad it it did not make me want to buy an iPad so huh did not seem like it made you sad
01:19:38
it actually elicited an emotion meaning like commercials it's very rare that commercials can actually do that most
01:19:43
people just Zone up yeah they took all this beautiful stuff and heard it didn't it didn't feel good I don't know it just
01:19:49
didn't seem like a good ad I don't know why they did that I don't get it I I've I don't I don't know I think I think
01:19:54
maybe what they're trying to do is the the selling point of this new iPad is that it's the thinnest one I mean
01:19:59
there's no innovation left so they're just making the devices yeah you know thinner yeah so I think the idea was
01:20:06
that they were going to take this hydraulic press to represent how ridiculously thin the new iPad is now I
01:20:12
don't know if the point there was to smush all of that good stuff into the iPad I don't know if that's what they
01:20:18
were trying to convey but yeah I I think that by destroying all those creative tools that apple is supposed to
01:20:26
represent it definitely seemed very offbrand for them and I think people were reacting to the fact that it was so
01:20:34
um different than what they would have done in the past and of course everyone was saying well Steve would never have
01:20:38
done this I do think it did land wrong I mean I I didn't care that much but but I
01:20:45
I was kind of asking the question like why are they destroying all these Creator tools that they're renowned for
01:20:52
creating or for turning into the digital version yeah it just didn't land I mean
01:20:58
shath how are you doing emotionally after seeing me are you okay buddy yeah I think this is uh you guys see that in
01:21:09
the birkshire annual meeting last weekend Tim Cook was in the audience and Buffett was very laudatory this is an
01:21:19
incredible company but he's so clever with words he's like you know this is an incredible business that we will hold
01:21:28
forever most likely then it turns out that he sold $20 billion worth of Apple shares in the
01:21:34
cat we're gonna hold it forever which which by the way sell if you guys remember we we put that little chart up
01:21:40
which shows when he doesn't mention it in the in the annual letter it's basically like it's foreshadowing the
01:21:46
fact that he is just pounding the cell and he sold $20 billion well also holding it forever could mean one share
01:21:54
yeah exactly we kind of need to know like how much are we talking about I mean it's an incredible business that
01:22:01
has so much money with nothing to do they're probably just going to buy back the stock just a total waste they were
01:22:06
floating this rumor of buying rivan you know after they shut down Titan project the your internal project to make a car
01:22:11
it seems like a car is the only thing people can think of that would move the needle in terms of earnings I think the
01:22:17
problem is J like you kind of become afraid of your own shadow meaning the the folks that are really good at m&a
01:22:23
like you look at Benny off the thing with benof m& strategy is that he's been doing it for 20 years and so
01:22:31
he's cut his teeth on small Acquisitions and the market learns to give him trust
01:22:37
so that when he proposes like the $27 billion slack acquisition he's allowed to do that another guy you know narora
01:22:44
at pandw these last five years people were very skeptical that he could actually roll up security because it was
01:22:49
a super fragmented Market he's gotten permission then there are companies like Dan that buy hundreds of companies so
01:22:56
all these folks are examples of you start small and you you earn the right to do more Apple hasn't bought anything
01:23:02
more than 50 or hundred million do and so the idea that all of a sudden they come out of the blue and buy a 102
01:23:08
billion dollar company I think is just totally doesn't stand logic it's just not possible for them because they'll be
01:23:15
so afraid of their own shadow that's the big problem it's themselves well if you're running out of In-House
01:23:21
Innovation and you can't do m&a then your options are kind of limited I mean I do think that the fact that the big
01:23:27
news out of apple is the iPad's Getting Thinner does represent kind of the end of the road in terms of innovation it's
01:23:34
kind of like when they added the third camera to the iPhone yeah it reminds me of those um remember like when the
01:23:41
Gillette yeah they came out and then they did the five was the best onion thing was like we're doing five f
01:23:47
it but then Gillette actually came out with the Mach 5 so yeah like the parody became the reality what are they going
01:23:52
to do add two more cameras to the iPhone you have five cameras on it no makes no
01:23:57
sense and then I don't know anybody wants to remember the Apple Vision was like gonna plus why are they body
01:24:02
shaming the the fat iPads that's a fair point Fair Point actually you know what it's actually
01:24:10
this didn't come out yet but it turns out the iPad is on OIC it's actually dropped a lot that would have been a
01:24:15
funnier ad yeah yeah exactly oh oh ohic we can just Workshop that right here but there was another funny one
01:24:24
which was making the iPhone smaller and smaller and smaller and the iPod smaller
01:24:27
and smaller and smaller to the point it was like you know like a thumb siize iPhone like the Ben Stiller phone in
01:24:33
Zoolander or correct yeah that was a great scene is there a category that you can think of
01:24:42
that you would love an Apple product for there's a product in your life that you
01:24:47
would love to have Apple's version of it they they killed it I think a lot of people would be very open minded to an
01:24:55
Apple car okay they they just would it's it's a connected internet device increasingly so yeah and they they
01:25:02
managed to flub it they had a chance to buy Tesla they managed to flub it yeah right there are
01:25:09
just too many examples here where these guys have so much money and not enough ideas that's a
01:25:14
shame it's a bummer yeah the one I always wanted to see them do saxs was TV the one I always wanted to see them do
01:25:21
was the TV and they were supposedly working on it like the actual TV not the little Apple TV box in the back and like
01:25:26
that would have been extraordinary to actually have a gorgeous you know big television what about a gaming console
01:25:33
they could have done that you know there's just all these things that they could have done it's not a lack of
01:25:39
imagination because these aren't exactly incredibly World beating ideas they're sitting right in front of your face it's
01:25:46
just the will to do it yeah all in one TV would have been good if you think back on Apple product lineup over the
01:25:55
years where they've really created value is on how unique the products are they almost create new categories sure there
01:26:02
may have been a quote tablet computer prior to the iPad but the iPad really defined the tablet computer era sure
01:26:08
there was a smartphone or two before the iPhone came along but it really defined
01:26:12
the smartphone and sure there was a computer before the Apple 2 and then it came along and it defined the personal
01:26:17
computer in all these cases I think Apple strives to define the category so it's very hard to define a television if
01:26:24
you think about it or gaming console in a way that you take a step up and you say this is the new thing this is the
01:26:29
new platform so I don't know that's the lens I would look at if I'm Apple in terms of like can I redefine a car can I
01:26:36
make you know we're all trying to fit them into an existing product bucket but I think what they've always been so good
01:26:41
at is identifying consumer needs and then creating an entirely new way of addressing that need in a real step
01:26:46
change function from the like the the the um iPod it was so different from any MP3 player ever I think the reason why
01:26:53
the car could have been completely reimagined by Apple is that they have a level of credibility and trust that I
01:26:59
think probably no other company has and absolutely no other tech company has and
01:27:05
we talked about this but I think this was the third Steve Job story that that I left out but in
01:27:13
200 I don't know was it [Music] one I launched a 99 cent download store right I think I've told you the story in
01:27:22
winam and Steve Jobs just ran total circles around us but the reason he was able to is he
01:27:30
had all the credibility to go to the labels and get deals done for licensing music that nobody could get done before
01:27:36
I think that's an example of what Apple's able to do which is to use their political Capital to change the rules so
01:27:42
if the thing that we would all want is safer roads and autonomous vehicles there are regions in every town and city
01:27:49
that could be completely converted to level five autonomous zones if I had to pick one company that had the
01:27:57
credibility to go and change those rules it's them because they could demonstrate
01:28:01
that there was a methodical safe approach to doing something and so the point is that even in these categories
01:28:08
that could be totally reimagined it's not for a lack of imagination again it just goes back to a complete lack of
01:28:13
Will and I understand because if they had if you if you had $200 billion do of capital on your balance sheet I think
01:28:21
it's probably pretty easy to get fat and lazy yeah it is and and they want to have everything built there people don't
01:28:27
remember but they actually built one of the first digital cameras you must have owned this right freeberg you're I
01:28:32
remember this yeah totally it beautiful what did they call it was it the ey camera or something quick take quick
01:28:39
take yeah um the thing I would like to see apple build and I'm surprised they didn't was a smart home system the way
01:28:46
Apple has Nest a drop cam a door lock you know a AV system go equestron or whatever and just have your whole home
01:28:56
automated thermostat Nest all of that would be brilliant by Apple and right now I'm an apple family that has our all
01:29:04
of our home automation through Google so it's just kind of sucks I would I would
01:29:08
like that all to that would be pretty amazing like if they did a crant or Savant CU then when you just go to your
01:29:13
Apple TV all your cameras just work you don't need to yes that's the that I mean and everybody
01:29:19
has a home and everybody automates their home so well everyone has Apple TV at this point so you just make Apple TV the
01:29:26
brain for the home system right that would be your Hub and you can connect your phone to it then yes that would be
01:29:33
very nice yeah like can you imagine like the ring cameras all that stuff being integrated I don't know why they didn't
01:29:39
go after that that seems like the easy layup hey you know everybody's been talking freedberg about this uh Alpha
01:29:47
fold this folding proteins and there's some new version out from Google and uh also Google
01:29:55
reportedly we talked about this before is also advancing talks to acquire HubSpot so that rumor for the $30
01:30:01
billion market cap HubSpot is out there as well free break your as our resident science Sultan uh our resident Sultan of
01:30:09
Science and as an Google alumni pick either story and let's go for it yeah I mean I'm not sure there's
01:30:15
much more to add on the HubSpot acquisition rumors they are still just rumors and I think we covered the topic
01:30:20
a couple weeks ago but I will say that Alpha fold 3 that was just announced today and demonstrated by Google um is a
01:30:27
real uh I would say breathtaking moment um for biology for bioengineering for human health for medicine and maybe I'll
01:30:36
just take 30 seconds to kind of explain it um you remember when they introduced Alpha fold at Alpha fold 2 we talked
01:30:43
about DNA codes for proteins so every three letters of DNA codes for an amino acid so a string of DNA codes for a
01:30:52
string of amino acids and that's called a gene that produces a protein and that protein is basically a
01:30:59
like think about beads there's 20 different types of beads 20 different amino acids that can be strung together
01:31:04
and what happens is that necklace that bead necklace basically collapses on itself and all those little beads stick
01:31:10
together with each other in some complicated way that we can't deterministically model and that creates
01:31:15
a three-dimensional structure which is called A protein that molecule and that molecule does something interesting it
01:31:20
can break apart other molecules it can bind molecules it can move molecules around so it's basically the Machinery
01:31:26
of chemistry of biochemistry and so proteins are what is encoded in our DNA and then the proteins do all the work of
01:31:34
making living organisms so Google's Alpha fold project took threedimensional images of proteins and the DNA sequence
01:31:41
that codes for those proteins and then they built a predictive model that predicted the three-dimensional
01:31:46
structure of a protein from the DNA that codes for it and that was a huge breakthrough years ago what they just
01:31:52
announced with Alpha fold 3 today is that they're now including all small molecules so all the other little
01:31:58
molecules that go into chemistry and biology that drive the function of everything we see around us and the way
01:32:05
that all those molecules actually bind and fit together is part of the predictive model why is that important
01:32:11
well let's say that you're designing a new drug and it's a protein based drug which biologic drugs which most drugs
01:32:16
are today you could find a biologic drug that binds to a cancer cell and then you'll spend 10 years going to clinical
01:32:22
trials and billions of later you find out that that protein accidentally binds to other stuff and hurts other stuff in
01:32:28
the body and that's an off Target effect or a side effect and that drug is pulled
01:32:32
from the clinical trials and it never goes to Market most drugs go through that process they are actually tested in
01:32:39
in animals and then in humans and we find all these side effects that arise from those drugs because we don't know
01:32:45
how those drugs are going to bind or interact with other things in our biochemistry and we only discovered
01:32:50
after we put it in but now we can actually model that with software we can take that drug we can create a
01:32:56
three-dimensional representation of it using the software and we can model how that drug might interact with all the
01:33:01
other cells all the other proteins all the other small molecules in the body to find all the of Target effects that may
01:33:08
arise and decide whether or not that presents a good drug candidate that is one example of how this capability can
01:33:15
be used and there are many many others including creating new proteins that could be used to bind molecules or stick
01:33:22
molecules together or new prot that could be designed to rip molecules apart we can now predict the function of
01:33:29
threedimensional molecules using this this capability which opens up all of the software based design of chemistry
01:33:35
of biology of drugs and it really is an incredible breakthrough moment the interesting thing that happened though
01:33:42
is Google alphabet has a subsidiary called isomorphic Labs it is a drug development subsidiary of alphabet and
01:33:49
they've basically kept all the IP for Alpha fold 3 in is so Google is going to monetize the heck
01:33:56
out of this capability and what they made available was not open source code but a webbased viewer that scientists
01:34:03
for quote non-commercial purposes can use to do some fundamental research in a web-based viewer and make some
01:34:08
experiments and try stuff out and how interactions might occur but no one can use it for commercial use only Google's
01:34:14
isomorphic Labs can so number one it's an incredible demonstration of what AI outside of llms which we just talked
01:34:21
about with Sam today and obviously talked about other models but llms being kind of this consumer text predictive
01:34:27
model capability but outside of that there's this capability in things like chemistry with these new AI models that
01:34:34
can be trained and built to predict things like three-dimensional chemical interactions that is going to open up an
01:34:41
entirely New Era for you know human progress and I think that's what's so exciting I think the other side of this
01:34:46
is Google is hugely advantaged and they just showed the world a little bit about
01:34:50
some of these jewels that they have in the treasure chest and they're like look at what we got we're going to all these
01:34:54
drugs and they've got Partnerships with all thesea companies at isomorphic Labs that they've talked about and it's going
01:34:59
to usher in a new era of drug development design for human health so all in all I'd say it's a pretty like
01:35:05
astounding day a lot of people are going crazy over the capability that they just
01:35:08
demonstrated and then it begs all this really interesting question around like you know what's Google G to do with it
01:35:13
and how much value is going to be created here so anyway I thought it was a great story and I just rambled on for
01:35:17
a couple minutes but I don't know it's pretty cool super interesting is is this AI capable of making a science Corner
01:35:22
that David saak pay attention to well it will it will predict the Cure I think for the common cold and for
01:35:30
herpes so he should pay attention absolutely folding cells is the app that Sachs casual game Sachs just downloaders
01:35:38
playing how many uh how many chest moves did you make during that segment sex sorry let me just say one more thing do
01:35:43
you guys remember we talked about yamanaka factors and how challenging it is to basically we can reverse aging if
01:35:48
we can get the right proteins into cells to tune the expression of certain gen to
01:35:53
make those cells youthful right now it's a shotgun approach to trying millions of
01:35:58
compounds and combinations of compounds to do them there's a lot of companies actually trying to do this right now to
01:36:03
come up with a fountain of youth type product we can now simulate that so with this system one of the things that this
01:36:10
Alpha 3 can do is predict what molecules will bind and promote certain sequences
01:36:15
of DNA which is exactly what we try and do with the yanaka factor-based expression systems and find ones that
01:36:21
won't trigger off Target expression so meaning we can now go through the search space and software of creating a
01:36:27
combination of molecules that theoretically could unlock this Fountain of Youth to deage all the cells in the
01:36:33
body and introduce an extraordinary kind of health benefit and that's just again
01:36:37
one example of the many things that are possible inredible with this sort of platform I I and I'm really I gota be
01:36:42
honest I'm really just sort of skimming the surface here of what this can do the
01:36:46
capabilities and the impact are going to be like I don't know I know I say this sort of stuff a lot but it's going to be
01:36:51
pretty for half there's um on the Block post they have this incredible video that they show of
01:36:57
the Corona virus that creates a common cold I think the S P&M protein and not only did they literally like predict it
01:37:06
accurately they also predicted how it interacts with an antibody with a sugar it's nuts so you could see a world where
01:37:15
like I don't know you just get a vaccine for the cold and it's kind of like you never have colds again amazing I mean
01:37:21
simple stuff but so powerful you can filter out stuff that has offt Target effect so so much of drug Discovery and
01:37:27
all the side effect stuff can start to be solved for in silicone and you could think about running extraordinar large
01:37:33
use a model like this run extraordinarily large simulations in a search space of chemistry to find stuff
01:37:39
that does things in the body that can unlock you know all these benefits can do all sorts of amazing things to
01:37:45
destroy cancer to destroy viruses to repair cells to deage cells and this is a hundred billion dollar business they
01:37:53
say I alone I feel like this is where I I've said this before I think Google's got this like portfolio of like
01:38:01
quiet extraord what if they hit and the fact and I think the fact that they didn't
01:38:06
open source everything in this says a lot about their intentions yeah yeah open source when you're behind closed
01:38:12
Source lock it up when you're ahead but show yamanaka actually interestingly yamanaka is the Japanese whiskey that
01:38:19
saak serves on his plane as well it's delicious I love that Hokkaido yak Jason I feel like if you didn't find
01:38:26
your way to Silicon Valley you could be like a Vegas Lounge comedy guy absolutely for sure yeah I was actually
01:38:32
yeah somebody said I should do like those 1950s those 1950s talk shows where the guys would do like the the stage
01:38:37
show somebody told me I should do um like spald and gray Eric bosan style stuff I don't know if you guys remember
01:38:44
like the uh the monologue is from the 80s in New York I was like oh that's interesting maybe all right everybody
01:38:48
thanks for tuning in to the world's number one podcast can you believe we did it jamath uh the number one podcast
01:38:56
in the world and the all-in summit the Ted killer if you are going to Ted congratulations for genu flecting if you
01:39:04
want to talk about real issues come to the all-in summit and if you are protesting at the all-in summit let us
01:39:11
know uh what mock meet you would like to have freeberg is setting up mock meat stations for all of our protesters and
01:39:18
what milk you would like yeah all vegan if you if your o milk nut milk come five different kinds of zanth gum
01:39:26
you can from all of the nut milks you could want and then they'll be mindful can we have some soil likein
01:39:32
please yes on the south lawn we'll have the goat yoga going on so just please note that the goat yoga will be going on
01:39:38
for all of you it's very thoughtful for you to make sure that our protesters are
01:39:43
going to be well well fed well taken care of yes we're actually freeberg is working on the protester gift bags the
01:39:50
protester gift bags they're made Yak folding proteins so you're good folding proteins I think I saw them open
01:39:59
for the Smashing Pumpkins in 2003 on fire on fire enough I'll be here for three more nights love you boys
01:40:07
byebye love you besties is this the all Potter open mic night what's going on it's basically I'm just
01:40:15
bored let your winners ride Rainman David said we open source it to the fans and
01:40:25
they've just gone crazy with it love queen [Music] of Besties are my dog taking driveway oh man myit will meet me we
01:40:46
should all just get a room and just have one big huge orgy cuz they're all this useless it's like this like sexual
01:40:51
tension that they just need to release somehow [Music] your we need to get [Music]
01:41:09
merch all right that's episode 178 and now the plugs the all-in Summit is taking place in Los Angeles on September
01:41:17
8th through the 10th you can apply for a ticket at summit. Allin podcast.co scholarships will be coming soon if you
01:41:25
want to see the four of us interview Sam Alman you can actually see the video of
01:41:30
this podcast on YouTube youtube.com Allin or just search all-in podcast and hit the alert Bell and
01:41:38
you'll get updates when we post we're doing a Q&A episode live when the YouTube channel hits 500,000 and we're
01:41:45
going to do a party in Vegas my understanding when we hit a million subscribers so look for that as well you
01:41:51
can follow us on x x /the all inpod Tik Tok is allore inor talk Instagram the all inpod and on
01:42:01
LinkedIn just search for the all-in podcast you can follow chth x.com chth and you can sign up for a substack at
01:42:08
cho. substack docomo freeberg can be followed at x.com freeberg and ohal is hiring click on the careers page at ohal
01:42:16
genetics.com and you can follow saxs at x.com davit saxs saxs recently spoke at the American moment conference and
01:42:24
people are going crazy for it it's pinned to his tweet on his X profile I'm Jason kakanis I amx.com Jason and if you
01:42:32
want to see pictures of my Bulldogs and the food I'm eating go to instagram.com Jason inth firstname Club you can listen
01:42:39
to my other podcast this week in startups just search for it on YouTube or your favorite podcast player we are
01:42:44
hiring a researcher apply to be a research you're doing primary research and working with me and producer Nick
01:42:50
working in data and Science and being able to do great great research Finance Etc Allin podcast.co resesarch it's a
01:42:57
full-time job working with us the besties we'll see you all next time when the Allin podcast