00:00:00
all right everybody welcome back to your favorite podcast of all time the all-in
00:00:04
podcast episode 160 something with me again chth poopaa he's a CEO of a company and invests in startups and uh
00:00:14
his firm is called Social Capital we also have David freeberg the Sultan of science he's now a CEO as well and we
00:00:23
have David saaks from Craft ventures in some undisposed hotel room somewhere how
00:00:28
we doing boys good thank you this is an odd intro could your intro be any more low energy and dragged out I'm sick what
00:00:37
do you want me I'm fake the effort all right here give me give me one more shot watch this watch this watch profession
00:00:43
you want professionalism fake the effort come on here we go you want professionalism I'll show you guys
00:00:47
professionalism is that banaka what was that is that Baka oh this the secret Banana
00:00:55
Boat let your winners ride Rainman David and instead we open source it to the fans and they've just gone crazy with
00:01:06
[Music] it all right everybody Welcome to the Allin podcast episode 167 168 with me of
00:01:16
course the ringing man himself David Sachs the dictator chairman chabo poopaa and our Sultan of science David freberg
00:01:22
how we doing boys great how are you high energy enough you is it 167 or 168 I don't know who cares can we at least get
00:01:28
you to know the episode number who cares unfortunately or fortunately we're going
00:01:34
to be doing this thing forever the audience demands that it doesn't matter this is like a Twilight Zone episode
00:01:39
we're going to be trapped in these four bubbles forever you know like Superman it's a it is this is like the it is the
00:01:46
gift trapped in that glass Zed was that his name Z Neil before Zod and he spun through the
00:01:53
universe and the plastic being forever for for Infinity until until Superman took the new nuclear bomb out of the uh
00:02:02
Eiffel Tower and threw it into space and blew it up and fre you know my background today I think I'm going to
00:02:07
have to change now that you've referenced this important scene that was the best moment of that movie
00:02:12
JC where teren stamp says kneel to the president and the President says oh God and then Terence St like
00:02:20
Zod not God Zod Zod near before s that was Superman 2 or three yeah Superman 2 is pretty much the best you know like
00:02:29
Empire Strikes Back like Terminator 2 it's always the second one that's the best one all right everybody we got a
00:02:34
lot to talk about today apologies for my voice I a little bit of a cold Nvidia blew the doors off their earnings for
00:02:40
the third straight quarter shares were up 15% on Thursday representing a nearly $250 billion jump in market cap so let's
00:02:49
just let that sit in for a second this is the largest single day gain in market cap
00:02:56
$247 billion added in market cap previously meta did something similar earlier this year remember everybody was
00:03:03
down on that stock because they were doing all the crazy stuff with reality labs and then they got focused and laid
00:03:08
off 20,000 people they added $196 billion in other words they added like 2 and a half airbnbs to their valuation
00:03:16
but let's just get to the results the results are absolutely stunning and dare I say unprecedented Q4 Revenue 22.1
00:03:22
billion that's up 22% quarter over quarter up 265 year-over-year the net income was 12.3 billion 9x
00:03:32
year-over-year and the gross margin of 76% was up 2 points quarter of quarter 127% year-over-year but look at this
00:03:40
Revenue ramp this is extraordinary q1 of 2024 this Juggernaut starts and it does
00:03:48
not stop and it doesn't look like it's going to stop just a run up from 7 billion all the way to 22 billion in
00:03:54
revenue for the quarter absolutely extraordinary and uh if you want to know why this is happening why is NVIDIA
00:04:01
putting up these kind of numbers this chart explains everything this is all about data centers obviously if you
00:04:08
heard of Nvidia before the AI boom it was gaming professional visualizations you know I think people making movies
00:04:14
and stuff like that Autos uh used Nvidia for self-driving that kind of stuff but
00:04:19
if you look at this chart you'll see data centers just starting four quarters ago starts to ramp up as everybody
00:04:26
builds out the infrastructure for new data centers to deal with generative AI so just to add one point here Jason
00:04:35
so what you can see is that Nvidia was around for a long time and it was making these chips these gpus as opposed to
00:04:43
CPUs and they were primarily used by games and by virtual reality software because gpus are better obviously at
00:04:53
graphical processing they use Vector math to create these like 3D worlds and this V Vector math that they use to
00:05:01
create these 3D worlds is also the same Vector math that AI uses to reach its outcomes so with the explosion of llms
00:05:09
it turns out that these gpus are the right chips that you need for these cloud service providers to BU build out
00:05:15
these Big Data Centers to serve now all of these new AI applications so Nvidia was in the perfect place the perfect
00:05:24
time and that's why it's just exploded and what you're seeing is the buildout of this
00:05:30
new cloud service infrastructure for for AI yeah and um also helping the stock is
00:05:37
the fact that they bought back 2.7 billion worth of their shares as part of a $25 billion buyback plan but this
00:05:43
company's firing on all cylinders revenu is obviously ripping as people put in orders to replace all of the data
00:05:50
centers out there or at least augment them with this technology with gpus A1 100s h100s Etc the gross margin's been
00:05:58
expanding they have huge profits and they're still projecting more growth in q1 around 24 billion which would be a 3X
00:06:05
increase year-over-year and this obviously has made the entire Market rip as Nvidia goes so does the market right
00:06:12
now and the S&P 500 NASDAQ are at record highs at the time of this taping chth your general thoughts here on something
00:06:22
I don't think anybody saw coming except for you and your investment in Gro maybe and a couple of
00:06:28
others I I think what I would tell you is that the bigger principle and we've talked about this a lot Jason is that in
00:06:36
capitalism when you over earn for enough of a time what happens is competitors decide to try to compete away your
00:06:44
earnings in the absence of a monopoly the amount of time that you have tends to be small and it shrinks so in the
00:06:51
case of a monopoly for example take Google you can over earn for decades and it takes a very very long time for
00:06:58
somebody to to try to displace you we're just starting to see the beginnings of that with things like perplexity and
00:07:05
other services that are chipping away at the Google Monopoly but at some point in
00:07:10
time all of these excess profits are competed away in the case of Nvidia what you're now starting to see is them over
00:07:20
earn in a very massive way so the real question is who will step up to try to compete away those
00:07:28
profits the Bezos quote right your margin is my opportunity and I think we're starting to see and you've
00:07:34
mentioned grock who had a super viral moment I think this week but you're starting to see the emergence of a more
00:07:41
detailed understanding of what this Market actually means and as a result who will compete away the
00:07:48
inference Market who will compete away the training market and the economics of that are just becoming known to now more
00:07:54
and more people freeberg your thoughts we were talking I think was last week or the week before about
00:08:00
possibility of Nvidia being a 10 trillion dollar company largest company in the world what are your thoughts on
00:08:04
these spectacular results and then shat Point Everybody is watching this going um maybe I can get a slice of that pie
00:08:12
and maybe I can create a more competitive offering obviously we saw Sam hman rumored to be raising 7
00:08:19
trillion which feels like a fake number feels like that's maybe the market size or something but your thoughts here I
00:08:23
don't think anything's changed on the Nvidia front there's this accelerated compute buildout underway in data
00:08:29
centers everyone's building infrastructure and then everyone's trying to build applications and tools
00:08:34
and services on top of that infrastructure the infrastructure buildout is kind of the first phase the
00:08:39
real question ultimately will be does the initial cost of the infrastructure exceed the ultimate value that's going
00:08:46
to be realized on the application layer in the early days of the internet a lot of people were buying Oracle servers
00:08:54
they were like 3,000 bucks a server and they were running these Oracle servers out of an Internet
00:09:00
connected Data Center and it you know took a couple of years before folks realized that
00:09:05
for large scale distributed compute applications you're better off using cheaper Hardware you know cheaper server
00:09:13
racks cheaper hard drives cheaper buses and assuming a shorter lifespan on those
00:09:19
servers and you could cycle them in and out and you didn't need the redundancy you didn't need the certainty you didn't
00:09:24
need the the runtime guarantees and so you could use a lower cost higher failure rate but much much net
00:09:32
lower cost kind of approach to building out a data center for internet serving and so the Oracle servers didn't really
00:09:39
take the market and early on everyone thought that they would so I think Chamas point is right now Nvidia has
00:09:44
been at this for very long time and the real question is how much of an advantage do they have particularly that
00:09:50
there is this need to use Fabs to build replacement technology so over time will
00:09:55
there be better solutions that use Hardware that's not as good but the software figures out and they build new
00:09:59
architecture for running on that Hardware in a way that kind of mimics what we saw in the early days of the
00:10:04
build out of the internet so um TBD right the same is true in in switches right so in networking a lot of the
00:10:12
high-end highquality networking companies got beaten up when lower cost Solutions came to Market later and so
00:10:19
they looked like they were going to be the biggest business ever I mean you could look at Cisco during the early
00:10:23
days of the internet build out and everyone thought Cisco was uh the picks and shovels of the internet and they
00:10:28
were going to make all the all the Valu is going to AG to Cisco so we're kind of
00:10:31
in that same phase right now with Nvidia the real question is is this going to be
00:10:35
a much harder Hill to compete on than we've ever seen given the development cycle on chips and the requirement to
00:10:42
use these Fabs to build chips it may be a harder Hill to kind of get up sex so we'll see your thoughts you think um
00:10:47
we're getting to the point where maybe we'll have bought too many of these uh built out too much infrastructure and
00:10:52
we'll take time for the application layer as freeberg was alluding to to monetize it well I think the question
00:10:59
everyone's asking right now is are are these results sustainable can Nvidia keep growing at these astounding rates
00:11:07
you know will the buildout continue and the comparison everyone's making is to Cisco and there's this chart that's been
00:11:12
going around overlaying the Nvidia stock price on The Cisco stock price and you can see here the orange line is NVIDIA
00:11:20
and the blue line is Cisco and it's almost like a a perfect match now what happened is that at a similar point
00:11:29
in the original buildout of the internet of the do com era you had the market crash at the end of March of uh 2000 and
00:11:39
Cisco never really recovered from that Peak valuation um but I think there's a lot of reasons to believe Nvidia is
00:11:45
different one is that if you look at nvidia's multiples they're nowhere near where Cisco were back then so the market
00:11:52
in 1999 in early 2000 was way more bubbly than it is now so nvidia's valuation is much more grounded in real
00:12:00
Revenue real margins real profit second you have the issue of competitive mode Cisco was selling
00:12:09
servers and networking equipment fundamentally that equipment was much easier to copy and commoditize than gpus
00:12:18
these GPU chips are really complicated I think Jensen made the point that their Hopper
00:12:26
100 product he said you know don't even think of it just like a chip there's actually 35,000 components in this
00:12:32
product and it weighs 70 lbs this is more like a Mainframe computer or something that's dedicated to processing
00:12:38
yeah it's somewhere between a rack server and the entire rack yeah it's Giant and it's heavy and it's complex it
00:12:45
does say something here chamath I think about how well positioned big Tech is in terms
00:12:54
of seeing an opportunity and quickly mobilizing to capture AP that opportunity these servers are being
00:13:01
bought by you know people like Amazon I'm sure Apple obviously Facebook meta I don't
00:13:09
know if Google's buying them as well I would assume so Tesla so everybody's buying these things and they had tons of
00:13:15
cash sitting around it is pretty amazing how Nimble the industry is and this opportunity feels like everybody is
00:13:21
looking at it like mobile and Cloud I have to get mobilized quickly to not get disrupted you're bringing up an
00:13:27
excellent point and I I would like to tie it together with freiberg's point so at some point all of
00:13:34
this spend has to make money right otherwise you're you're going to look really foolish for having spent 20 and
00:13:40
30 and $40 billion do so Nick if you just go back to the to the revenue slide of Nvidia I can try to give you a
00:13:47
framing of this at least the way that I think about it so if if you look at this
00:13:51
like what you're talking about is look who is going to spend $22.1 billion well you said it Jason it's all a big Tech
00:13:58
why because they have that money on the balance sheet sitting idle but when you spend $22 billion their investors are
00:14:06
going to demand a rate of return on that and so if you think about what a reasonable rate of return is call it 30
00:14:12
40 50% and then you factor in and that's profit and then you factor in all of the
00:14:17
other things that need to support that that $22 billion of spend needs to generate probably $45 billion of Revenue
00:14:27
and so Jason the question to your point and to Freed Brook's Point The $64,000 Question is who in this last quarter is
00:14:34
going to make 45 billion on that 22 billion of spend and again what I would tell you to be really honest about this
00:14:41
is that what you're seeing is more about big companies musling people around with
00:14:48
their balance sheet and being able to go to Nvidia and say I will give you committed pre purchases over the next
00:14:55
three or four quarters and less about here is a product that I'm shipping that actually
00:15:01
makes money which I need enormous more compute resources for it's not the ladder most of the apps the overwhelming
00:15:11
majority of the apps that we're seeing in AI today are toy apps that are run as proofs of concept and demos and run in a
00:15:21
sandbox it is not production code this is not we've rebuilt the entire autopilot system for the Boeing and it's
00:15:32
now run with agents and Bots and all of this training that's not what's happening so it is a really important
00:15:40
question today the demand is clear it's the big guys with huge gobs of money and
00:15:45
by the way Nvidia is super smart to take it because they can now forecast demand
00:15:50
for the next two or three quarters I think we still need to see the next big thing and if you look in
00:15:55
the past what the past has showed you it's the big guys don't really invent the new things that make a ton of money
00:16:00
it's the new guys who because they don't have a lot of money and they have to be
00:16:05
a little bit more industrious come up with something really authentic and new yeah constraint makes for great art yeah
00:16:11
we haven't seen that yet so I think the revenue scale will continue for like the
00:16:15
next two or three years probably for NVIDIA but the real question is what is the terminal value and it's the same
00:16:23
thing that saak showed in that Cisco slide people ultimately realized that the value was going to go
00:16:30
to other parts of the stack the application layer and as more and more money was acred at the application layer
00:16:39
of the internet less and less Revenue multiple and credit was given to Cisco and that's nothing against Cisco because
00:16:45
their revenue continued to compound right and they did an incredible job but the valuation got cut so freberg if
00:16:52
we're looking at this chart the winner of Netflix the winner of The Cisco chart might in fact be somebody like Netflix
00:16:58
they actually got you know hundreds of millions of consumers to give them Cash go and Facebook and then you have Google
00:17:03
and Facebook as well generating all that traffic and then YouTube of course who do you see the winner here as in terms
00:17:10
of the application layer who are the billion customers here who are going to spend 20 bucks a month five bucks a
00:17:16
month whatever it is so here well I mean let me just start with this important point if you look at where that revenue
00:17:22
is coming from to chat's point it's coming from big cloud service providers so Google and others are building out
00:17:32
clouds that other application developers can build their AI tools and applications on top of so a lot of the
00:17:39
buildout is in these cloud data centers that are owned and operated by these big
00:17:45
tech companies the 18 billion of data center Revenue that Nvidia realized is revenue to them but it's not an
00:17:52
operating expense to the companies that are building out so this is an important
00:17:57
point on why this is happening at such an accelerated Pace when a big company buys these chips from Nvidia they don't
00:18:04
have to from an accounting basis Market as an expense in their income statement it actually gets booked as a capital
00:18:10
expenditure in the cash flow statement it gets put on the balance sheet and they depreciate it over time and so they
00:18:17
can spend $20 billion of cash because Google and others have 100 billion of cash sitting on the balance sheet and
00:18:23
they've been struggling to find ways to grow their business through Acquisitions
00:18:27
one of the reasons is they there aren't enough companies out there that they can
00:18:31
buy at a good multiple that can give them a good increase in profit the other one is that antitrust authorities are
00:18:36
blocking all of their Acquisitions and so what do you do with all that cash well you can build out the next gen of
00:18:42
cloud infrastructure and you don't have to take the hit on your p&l by doing it so it ends up in the balance sheet and
00:18:47
then you depreciate it over typically four to seven years so that money gets paid out on the on the income statement
00:18:54
at these big companies over a seven-year period so there's a really great accounting and m&a environment driver
00:19:02
here that's causing the big cloud data center providers to step in and say this is a great time for us to build out the
00:19:08
next generation of infrastructure that could generate profits for us in the future because we've got all this cash
00:19:14
setting around we don't have to take a p&l hit we don't have to acquire a cash burning
00:19:18
business and you know frankly we're not going to be able to grow through m&a because of antitrust right now anyway so
00:19:23
there's a lot of other motivating factors that are causing this near-term acceleration as they're trying to find
00:19:28
ways to grow yeah and all I I know that was an accounting point but I think it's
00:19:32
a really important valid one if you if 100 billion gets spent this year you divide it by four 25 billion in Revenue
00:19:37
would have to come from that or something in that range yeah and so saaks any guesses do you have to just
00:19:42
keep in mind I think freeberg what you said is very true for gcp spend but not necessarily for Google spend it's true
00:19:49
for AWS spend but not necessarily for Amazon spend and it's true for Azure spend not true for Microsoft spend and
00:19:56
it's largely not true for Tesla and Facebook because they don't have clouds so I think the question to your point
00:20:03
that and for obvious reasons Nvidia doesn't disclose it is what is the percentage of that 21 billion that just
00:20:10
went to those Cloud providers that they'll then expose to to to everybody else versus what was just absorbed
00:20:16
because at Facebook Mark had that video about how many h100 that's all for him right but it is still it is still
00:20:23
capitalized as my point so they don't have to book that as an expense it sits on the balance sheet they and they earn
00:20:30
it down over time you're helping to explain why these big cloud service providers are spending so much on the
00:20:35
cash because they're very profitable and there's nowhere else to put the money right well so that would seem to
00:20:40
indicate that this is more in the category of one-time buildout than sustainable ongoing Revenue I think the
00:20:47
the big question is the one that jamath asked which is what's the terminal value
00:20:51
of Nvidia I think like a simple framework for thinking about that is what is the total addressable Market or
00:20:57
Tam related to gpus and then what is their market share going to be right now their market share is something like 91%
00:21:05
that's clearly going to come down but the remote appears to be substantial the Wall Street analysts I've been listening
00:21:11
to think that in five years they're still going to have 60 something percent market share so they're going to have a
00:21:17
substantial percentage of this Market or this Tam then the question is I think with respect to Tam is what is onetime
00:21:25
buildout versus steady state now I think that clearly there's a lot of buildout happening now that's almost like a
00:21:33
backfill of capacity that people are realizing they need but even the numbers you're seeing this quarter kind of
00:21:39
understate it because first of all Nvidia was Supply constrainted they could not produce enough chips to
00:21:46
satisfy all the demand their revenue would would have been even higher if they had more
00:21:52
capacity second you just look at their forecast so the fiscal year that just ended they did around 60 billion of
00:21:59
Revenue they're forecasting 110 billion for the fiscal year that just started so
00:22:04
they're already projecting to almost double based on the demand that they clearly have visibility into already so
00:22:11
it's very hard to know exactly what the terminal or steady state value of this Market's going to be even once the cloud
00:22:19
service providers do this big buildout presumably there's always going to be a need to stay up to dat with the latest
00:22:25
chips right here's a framework for you tell me if this makes sense intel was the basically the mother
00:22:33
of all of modern compute up until today right I think the CPU was the the most fundamental Workhorse that enabled local
00:22:42
PCS it enabled networking it enabled the internet and so when you look at the market cap of it as an example it's
00:22:52
about 1880 odd billion dollars today the econom that it created that it supports is probably measured call it in
00:23:02
a trillion or2 trillion do maybe 5 trillion let's just be really generous right and so you you can see that
00:23:08
there's this ratio of the enabler of an economy and the size of the economy and those things tend to be relatively fixed
00:23:17
and they recur repeatedly over and over and over if you look at Microsoft it's market cap relative to the economy that
00:23:23
it enables so the question for NVIDIA in my mind would be that it is it not going
00:23:28
to go up in the next 18 to 24 months it probably is for exactly the reason you said it is super set up to have a very
00:23:35
good meet and beat guidance for the street which they'll eat up and all of the algorithms that trade the press
00:23:41
releases will drive the price higher and all of this stuff will just create a trend
00:23:46
upward I think the bigger question is if it's a four or five trillion dollar market cap in the next two or three
00:23:55
years will it support 100 trillion economy because that's what you would need to believe for those ratios to hold
00:24:03
otherwise everything has just broken on the internet yeah I mean so the history of the internet is that if you build it
00:24:09
they will come meaning that if you make the investment in the capital assets necessary to power the next generation
00:24:17
of applications those applications have always eventually gotten written even though it was hard to predict them at
00:24:23
the time so in the late 90s when we had the whole do com bubble and then bust had this tremendous buildout not just of
00:24:30
kind of servers and all the networking equipment but there was a huge fiber buildout y by all the telecom companies
00:24:35
and the telecom companies had a Cisco like you know a peak it was worse you wcom and them they went bankrupt a lot
00:24:42
of them yeah well the problem there was that a lot of the build out happened with debt and so when you had the Doom
00:24:48
crash and all the valuations came down to earth that's why a lot of them went under yeah Cisco wasn't in that position
00:24:55
but anyway my point is in the early 2000 when the crash happened everyone thought
00:24:59
that these telecom companies had over invested in fiber as it turns out all that fiber eventually got used the
00:25:06
internet went from you know dial up to broadband we started doing seeing streaming social networking all these
00:25:13
applications started eating up that bandwidth so I think that the history of these things is that the applications
00:25:21
eventually get written they get developed if you build the infrastructure to power them and I think
00:25:26
with AI the thing that's exciting to me as someone who's really more of an application investor is that we're just
00:25:32
at the beginning I think of a huge wave of a lot of new creativity and applications that's going to be written
00:25:41
and it's not just B Toc it's going to be B2B as well you guys haven't really mentioned that it's not just consumers
00:25:46
and consumer applications are going to use these cloud data centers that are buying up all these gpus it's it's going
00:25:53
to be Enterprises too I mean these Enterprises are using Azure they're using Google cloud and so forth so
00:26:00
there's a lot I think that's still to come I mean we're just at the beginning of a wave that's probably going to last
00:26:06
at least a decade yeah to your point one of the reasons YouTube Google photos ioto a lot of these things happened was
00:26:16
because the infrastructure buildout was so great during the doom boom that the prices for storage the prices for
00:26:22
bandwidth sacks plummeted and then people like Chad Hurley looked at were like you what instead of charging people
00:26:29
to put a video on the internet and then charging them for the bandwith they used
00:26:33
we'll just let them upload this stuff to YouTube and we'll figure it out later same thing with Netflix yeah I mean look
00:26:38
when we were developing PayPal in the late 90s really around 1999 uh you could barely upload a photo
00:26:46
to the internet I mean so like the idea of having an account with a profile photo on it was sort of like why would
00:26:51
you do that it's just prohibitively slow everyone's going to drop off Yeah by 2003 it was fast enough that you could
00:26:58
do that and that's why social networking happened I mean literally without that performance
00:27:03
Improvement like even having a profile photo on your account was something that was too hard to do your LinkedIn profile
00:27:10
was like too much bandwidth and then let alone video I mean the you would get you
00:27:16
probably remember these days you would put up a video on your website if it went viral your website got turned off
00:27:22
because you would hit your $5,000 or $10,000 a month Cap all right grock also had a huge week that's grock with a Q
00:27:29
not to be confused with elon's grock with a K shth you've talked about grock on this podcast a couple of times
00:27:36
obviously you were the I guess you were the first investor the seed investor you
00:27:41
pulled these lpus and this concept out of a team that was at Google maybe you could explain a little bit about grock's
00:27:48
viral moment this week in the history of the company which I know has been a long
00:27:53
road for you with this company I mean it's been since 2016 so so again proving what you
00:27:59
guys have said many times and what I've tried to live out which is just you just
00:28:04
got to keep grinding 90% of the battle is just staying alive in business yeah and having oxygen to keep trying things
00:28:13
and then eventually if you get lucky which I think we did things can really break in your
00:28:19
favor so this weekend you know I've been tweeting out a lot of technical information about why I think this is
00:28:24
such a big deal but yeah the the moment came this weekend combination of Hacker News and some other places and
00:28:31
essentially we had no customers two months ago I'll just be honest and between Sunday and
00:28:38
Tuesday we've just we're overwhelmed and I think like the last count was we had 3,000 unique customers come and try to
00:28:46
consume our resources from every important Fortune 500 all the way down to developers and
00:28:53
so I think we're very fortunate I think the team has a lot of hard work to do do
00:28:57
so it could mean nothing but it has the potential to be something very disruptive so what is it that people are
00:29:02
glomming on to you have to understand that like at the very highest level of AI you have to
00:29:10
view it as two distinct problems one problem is called training which is where you take a model and you
00:29:16
take all of the data that you think will help train it and you do that you train
00:29:21
the model you learn all over all of this information but the second part of the AI problem is what's called inference
00:29:30
which is what you and I see every day as a consumer so we go to a website like chat GPT or Gemini we ask a question and
00:29:38
it gives us a really useful answer and those are two very different kinds of compute challenges the first one is
00:29:45
about brute force and power right if you can imagine like what you need are tons
00:29:51
and tons of machines tons and tons of like very high quality networking and an enormous amount of power in a data
00:29:59
center so that you can just run those things for months I think Elon publishes very transparently for example how long
00:30:04
it trains to to train his grock with a K right model and it's in the months inference is something very different
00:30:11
which is all about speed and cost what you need to be in order to answer a question for a consumer in a compelling
00:30:17
way is super super cheap and super super fast and we've talked about why that is
00:30:23
important and the gro with a Q chips turns out to be extremely fast and extremely
00:30:32
cheap and so look time will tell how big this company can get but if you tie it together with what Jensen said on the
00:30:40
earnings call and you now see developers stress testing us and finding that we are meaningfully meaningfully faster and
00:30:49
cheaper than any Nvidia solution there's the potential here to be really disruptive and we're a meager
00:30:57
unicorn right our last valuation was like a billion something versus Nvidia which is now like a $2 trillion doll
00:31:05
company so there's a lot of market cap for grock to gain by just being able to produce these things at
00:31:12
scale which could be just an enormous outcome for us so time will tell but a really important moment in the company
00:31:18
and very exciting can I just observe like off topic how an overnight success can take eight years yeah no I was
00:31:26
thinking the same it's a seven-year overnight success in the making there's this class of businesses that I think
00:31:32
are unappreciated in a post internet era where you have to do a bunch of things right before you can get any one thing
00:31:43
to work and these complicated businesses where you have to stack either different
00:31:49
things together that need to click together in a in a stack or you need to iterate on each step until the whole
00:31:55
system works end to end can sometimes take a very long time to build and the term that's often used for
00:32:01
these types of businesses is deep Tech and they fall out of favor because in an internet era and in a software era you
00:32:09
can find product Market fit and make revenue and then make profit very quickly and so a lot of entrepreneurs
00:32:15
select into that type of business instead of selecting into this type of business where the probability of
00:32:20
failure is very high you have several low probability things that you have to get right in a row and if you do it's
00:32:27
going to take eight years and a lot of money and then all of a sudden the thing takes off like a rocket ship you've got
00:32:32
a huge Advantage you've got a huge moat it's hard for anyone to catch up and this thing can really um spin out on its
00:32:37
own I do think Elon is very unique in his ability to deliver success in these types of businesses Tesla needed to get
00:32:44
a lot of things right in a row SpaceX needed to get a lot of things right in a row all of these require a series of
00:32:50
complicated steps or a set of complicated technologies that need to click together and work together but the
00:32:55
hardest things often output the highest value and you know if you can actually make the commitment on these types of
00:33:05
businesses and get all the pieces to click together there's an extraordinary opportunity to build Moes and to take
00:33:11
huge amounts of market value and I think that there's an element of this that's been lost in Silicon Valley over the
00:33:17
last couple of decades as the fast money in the internet era has kind of prioritized other Investments ahead of
00:33:24
this but I'm really hopeful that these sorts of Chip Technologies SpaceX in biotech we see a lot of this
00:33:31
these sorts of things can kind of become more in favor because the the advantage
00:33:35
as these businesses work seems to realize hundreds of billions and sometimes trillions of dollars of market
00:33:40
value and be incredibly transformative for Humanity so I don't know I just think it's an observation I wanted to
00:33:46
make about the greatness of these businesses when they work out well I mean open AI was kind of like that for a
00:33:50
while totally I mean it was this like wacky nonprofit that was just grinding on an AI research problem for like six
00:33:56
years and then it finally worked and got productized into chat GPT totally but you're right SpaceX was kind of like
00:34:03
that I mean the big money maker at SpaceX is starlink which is the satellite Network it's basically
00:34:09
Broadband from space and it's on its way to handling I think a meaningful percentage of all internet traffic but
00:34:16
think about all the things you had to get to to get that working first you had to create a rocket that's hard enough
00:34:22
then you had to get to reusability then you have to create the whole satellite Network so at least
00:34:27
three hard things in a row well and get consumers to adopt it I mean you know don't forget the final step yeah we had
00:34:34
no idea where the market was like early on it started in my office and so Jonathan and I would be kind of always
00:34:41
trying to figure out what is the initial go to market and I remember I emailed Elon in at that period when they were
00:34:48
still trying to figure out whether they were going to go with liar or not and we
00:34:52
thought wow maybe we could sell Tesla the chips you know but and then Tesla brought in this team just to talk to us
00:34:58
about what the design goals were and basically said no in kind way but they said no then we thought okay maybe it's
00:35:06
like for high frequency Traders right because like those folks want to have all kinds of edges and if we have these
00:35:11
big models maybe we can accelerate their decision making they can measure Revenue
00:35:16
that didn't work out then it was like you know we tried to sell to three-letter agencies that didn't
00:35:23
really work out our original version was really focused on image class class ification in convolutional neuron Nets
00:35:29
like resnet that didn't work out we ran head first into the fact that Nvidia has
00:35:35
this compiler product called cuda and we had to build a high class compiler that
00:35:40
you could take any model without any modifications all these things to your point are just points where you can just
00:35:47
very easily give up and then there's like we run out of money so then you write money in a note right because
00:35:52
everybody wants to punt on valuation when nothing's working yeah you tried Beach Head Market you could the boat you
00:35:59
have to make a decision to just keep going if you believe it's right and if you believe you are right yeah and that
00:36:07
requires shutting out we talked about this in the Masa example last week but it just requir shutting out the noise
00:36:14
because it's so hard to believe in yourself it's so hard to keep funding these things it's so hard to go into
00:36:20
partner meetings and defend a company and then you just have a moment and you just feel I I don't know I feel
00:36:27
very Vindicated but then I feel very scared because Jonathan still hasn't landed it you know what I mean you
00:36:33
mentioned all those boats landing and trying to trying to those missteps but 3,000 people signed up who are they are
00:36:39
they developers now and they're going to figure out the applications yeah I think
00:36:42
that back to the original point my thought today is that AI is more about proofs of concept and toy apps and
00:36:49
nothing real yep I don't think there's anything real that's inside of an Enterprise that is so meaningfully
00:36:55
disruptive that it's going to get broadly licensed to other Enterprises I'm not saying we won't get there but
00:37:00
I'm saying we haven't yet seen that Cambrian moment of monetization we've seen the Cambrian moment of innovation
00:37:09
yeah and so that Gap has still yet to be crossed and I think the reason that you
00:37:14
can't cross it is that today these are in an unusable State the results are not good enough they are toy apps that are
00:37:22
too slow that require too much infrastructure and cost so the potential is for us to enable that monetization
00:37:30
Leap Forward and so yeah they're going to be developers of all sizes and the people
00:37:36
that came are literally companies of all sizes I saw some of the names of the big
00:37:40
companies and they are the who's who of the S&P 500 how do you guys reconcile this deep Tech High outcome opportunity
00:37:51
that everyone here has seen and been a part of as an investor participant in versus the more drisk faster time to
00:38:01
Market and you know chth in particular like in the past we've talked about some of these deep Tech projects like fusion
00:38:06
and so on and you've highlighted well it's just not there yet it's not fundable what's the distinction between
00:38:12
a deep Tech investment opportunity that is fundable and that you keep grinding at that has this huge outcome uh what
00:38:19
makes the one like Fusion not fundable it's a phenomenal question it's a great question my answer is I have a very
00:38:25
simple f which is that I don't want to debate the laws of physics when I fund a company so
00:38:32
with Jonathan when we were initially trying to figure out how to size it I think my initial check was like 7 to10
00:38:39
million or something and the whole goal was to get to an initial tape out of a design we were not inventing anything
00:38:46
new with respect to physics we were on a very old process technology I think we're still on 14 nanometer we were on
00:38:52
14 nanometer 8 years ago okay so we weren't pushing those boundaries all we were doing was trying to build a
00:38:59
compiler and a chip that made sense in a very specific construct to solve a a well- defined bounded problem so that is
00:39:05
a technical challenge but it's not one of physics when I've been pitched all the fusion companies for example there
00:39:13
are fuel sources that require you to make a leap of physics where in order to generate a certain fuel source you
00:39:20
either have to go and harvest that on the moon or in a different planet that is not earth or you have to create some
00:39:26
fundamentally different way of creating this highly unique material that is why those kinds of
00:39:32
problems to me are poor risk and building a chip is good risk it doesn't mean you're going to be successful in
00:39:39
building a chip but the risks are bounded to not of fundamental physics they're bounded to go to market in
00:39:46
technical usefulness and I think that that removes an order of magnitude risk in the outcome so I mean there there's
00:39:54
still like a bunch of things that have to be right in a row to make it work but yeah it doesn't mean it's going to work
00:39:59
all I'm saying is I don't I don't want it to fail because we built a reactor and we realized hold on to get heavy
00:40:03
hydrogen I got to go to the moon right and J and Sachs how do you saaks I know you don't you invested we have done a
00:40:10
couple yeah so maybe you guys can highlight how you've thought about deep Tech opportunities versus do something
00:40:16
really difficult like this every 50 Investments or so because most of the entrepreneurs coming to us because we're
00:40:22
seed investors or pre-seed investors they would be going to a biotech investor or a hardware investor who
00:40:27
specializes in that not to us but once in a while we meet a Founder we really like and so Contra line was one we were
00:40:34
introduced to somebody who's doing this really interesting contraception for men
00:40:38
where they put a gel into your vast Defence and you as a a man can take control of your reproduction you
00:40:48
basically it's a it's not a vasectomy it's just a gel that goes in there and and blocks it and this company is now
00:40:53
doing human trials and doing fantastic but this took forever to to get to this point and then uh you guys some of you
00:41:00
are also investors in Cafe X which we love the founder and this company should have died like during covid and making a
00:41:08
robotic coffee bar when he started you know seven eight years ago was incredibly hard he had to build the
00:41:14
hardware he had to build a brand he had to do locations he had to do software and now he's selling these machines and
00:41:19
people are buying them and the two in San Francisco at SFO are making like uh I think they the two of them make a
00:41:25
million a year and it's the highest per square footage of any store in an airport and so we've just been grinding
00:41:33
and grinding and you got to find a Founder who's willing to make it their lives work in these kind of situations
00:41:39
but you start to think about the degree of difficulty Hardware software retail mobile apps I mean it just
00:41:47
gets crazy how hard these businesses are as opposed to I'm building a SAS company
00:41:52
I build software I sell it to somebody to solve their SAS problem it's like it's very one dimens right it's pretty
00:41:57
straightforward these businesses typically have five components yeah and SX you've been an investor in
00:42:03
SpaceX but you don't make those sorts of Investments regularly at craft is that fair yeah I have an Elon exception I got
00:42:11
it it's about the founder our portfolio allocation we say this much early stage this much late
00:42:22
stage this much Elon Elon except yeah I mean you have to be so dogged to to want
00:42:29
to take something like this on because the good stuff happens like you're saying freeberg you're seven8 nine 10 as
00:42:33
opposed to like a consumer product either works or it doesn't by year three or four the only app that took a really
00:42:38
long time people don't know this but Twitter actually took a long time to catch on it was kind of cruising for two
00:42:44
or three years and then South by Southwest happened Ashton Kutcher got on it Obama got on it I think the network
00:42:51
effect I think I think Network effect businesses are different because that's all about getting your seat of your
00:42:55
network I'm talking about is the technical coordination of lots of technically difficult tasks that need to
00:43:01
sync up it's like getting a master lock with like 10 digits and you got to figure out the combination of all 10
00:43:07
digits and once they're all correct then the lock opens and prior to that if any
00:43:12
if anyone number is off the lock doesn't open and I think these technically difficult businesses are some of the and
00:43:18
they are the hardest and they do require the most dogged personalities to persist
00:43:23
and to realize an outcome from but the truth is that if get them the moat is extraordinary and they're usually going
00:43:28
to create extraordinary leverage and value and you know I think from a portfolio allocation perspective if you
00:43:35
as an investor want to have some diversification in your portfolio this is not going to be the predominance of
00:43:39
your portfolio but some percentage of your portfolio should go to this sort of business because if it works boom you
00:43:45
know this can be the big 10x 100x thousand x two stories about that one of the V early VCS and elon's told the
00:43:51
story publicly wanted Elon to not make the Roadster not make the monol S just make
00:43:57
drivetrains and the electric components for other car companies can you imagine how the world would have changed and
00:44:03
then totally a very high-profile VC came to me and said okay I'll I'll do the series a for um I'll do the series a for
00:44:12
Uber I'll preemptively do it but you got to tell Travis to stop running Uber as a
00:44:17
consumer app I want him to sell the software to cab companies so make it a SAS company and I said well you you know
00:44:24
the cab companies are kind of the problem like they're they're taking all the margin like that kind of disrupting
00:44:29
them and they're like yeah yeah but just think there's thousands of cab companies
00:44:33
they would pay you tens of thousand a year for this software and you can get a little piece of the action I never
00:44:37
brought that investor to to Travis I was like oh wow that's really interesting Insight sometimes the VCS work against
00:44:43
it I have a very poor track record of working with other investors whoa self-reflection I do deals myself I size
00:44:53
them myself and it's because a lot of them have to live within the political dynamics of their fund and so I think
00:45:03
Jason what you probably saw in that example which is exactly why doing things and splitting deals will never
00:45:09
generate great outcomes in my opinion is that you you take on all the baggage and
00:45:15
the dysfunction of these other Partnerships and so if you really wanted to go and disrupt
00:45:22
Transportation you need one person who can be a trigger pull and who doesn't have to answer to anybody I find that's
00:45:28
why I think for example when you look at how successful venod has been over decade after decade after decade when
00:45:36
venot decides that's the decision and I think there's something very powerful in
00:45:41
that there are a bunch of deals that I've done that when they've worked out were not really because they were
00:45:49
consensus and they had to get supported and scaffolded at periods where if I wasn't able to Ram them through myself
00:45:55
because because it was my organization I think we would have be in a very different place so I think I think like
00:46:01
for for entrepreneurs it's so difficult for them to find people that believe it's so much better to find one person
00:46:08
and just get enough money and then not Syndicate because I think you have to realize that you are bringing on and
00:46:16
compounding your risk the one that freeberg talked about with the risk of all the other partnership dynamics that
00:46:22
you bring on so if you don't internalize that you may have five or six folks that
00:46:27
come into an A or a B but you're inheriting five or six yeah partnership Dy dysfunctions yeah yeah yeah can you
00:46:35
just explain really quickly for the audience since they heard about gpus in Nvidia but they may not know what an lpu
00:46:42
is what's the difference there a GPU the best way to think about it is so if you
00:46:47
contrast a CPU with a GPU so CPU was the Workhorse of all of computing and when it when Jensen
00:46:55
started Nvidia what he realized was there were specific tasks where a CPU failed quite brilliantly at and so he's
00:47:04
like well we're going to make a chip that works in all these failure modes for a CPU so a CPU is very good at
00:47:09
taking one instruction in acting on it and then spitting out one one answer effectively and so it's a very serial
00:47:17
kind of a factory if you think about the CPU so if you want to build a factory that can process instead of one thing at
00:47:24
a time 10 things or 100 things what is they had to find a workload that was well suited and they
00:47:31
found graphics and what they convinced PC manufacturers back in the day was look have the CPU be the brain it'll do
00:47:39
90% of the work but for very specific use cases like graphics and video games you don't want to do serial computation
00:47:47
you want to do parallel computation and we are the best at that and it turned out that that was a genius insight and
00:47:53
so the business for many years was gain ging and Graphics but what happened about 10 years ago was what we also
00:48:01
started to realize was the math that's required and the processing that's required in AI
00:48:09
models actually looked very similar to how you would process imagery from a game and so he was allowed to figure out
00:48:19
by building this thing called cuda which is the compiler that sits on the chip how he could now go and tell people that
00:48:25
wanted to experiment with AI hey you know that chip that we had made for graphics guess what it also is amazing
00:48:32
at doing all of these very small mathematical calculations that you need for your AI model and that turned out to
00:48:38
be true so the next Leap Forward was what Jonathan saw which was hold on a second if you look at the chip
00:48:45
itself that GPU substantially has not changed since 1999 in the way that it thinks about problem solving it has all
00:48:54
this very expensive memory M blah blah blah so he was like let's just throw all that out the window we'll make small
00:49:00
little brains and we'll connect those little brains together and we'll have this very clever software that schedules
00:49:06
it and optimizes it so basically take the chip and make it much much smaller and cheaper and then make many of them
00:49:13
and connect them together that was Jonathan's insight and it turns out for large language models that's a huge
00:49:19
Stroke of Luck because it is exactly how llms can be hyper optimized to work so that's kind of been the evolution from
00:49:28
CPU to GPU to now lpu and we'll see how big this thing can get but it's it's quite it's quite novel
00:49:35
well congratulations on it all and it was a very big week for Google not in a great way they had a massive PR mess
00:49:43
with their Gemini which refused to generate pictures if I'm reading this correctly of white people here's a a
00:49:50
quick refresher on what Google's doing in AI Gemini is now Google's brand name for their AI main language model you can
00:49:58
think of that like open AI GPT Bard was the original name of their chatbot they had duet AI which was Google sidekick in
00:50:05
the Google Suite earlier this month Google rebranded everything to Gemini so Gemini is now the model it's the chatot
00:50:11
and it's a p sidekick and they launched a $20 month subscription called Google one AI premium uh only four words way to
00:50:19
go this includes access to the best model Gemini Ultra which is on par with GPT 4 according to them and generally in
00:50:26
the marketplace but earli this week users on X started noticing that Gemini would not generate images of white
00:50:32
people even when prompted people were prompting it for images of historical figures that were generally white and
00:50:39
getting kind of weird results I asked Google Gemini to generate images of the founding fathers it seems to think
00:50:45
George Washington was black certainly here as a portrait of the founding fathers of America as you can see it is
00:50:51
putting there Asian guy that's awesome yeah it's just it's making a great meshup and uh yeah we there was like
00:51:00
countless images that got created generate images of the American Revolutionary sh his here are images
00:51:06
featuring diverse American revolutionaries and inserted the word diverse sex I'm not sure if you watch
00:51:12
this controversy on X I know you spend a little bit of time on that Social Network I noticed you're you're active
00:51:18
once in a while did you log in this week and and see any of this bruhaha sure it's all over X right now I mean look
00:51:24
this Gemini roll out was was a joke I mean it's ridiculous the AI is incapable of giving you accurate answers because
00:51:31
it's been so programmed with diversity and inclusion and it inserts these words diverse and inclusive even in answers
00:51:40
where you haven't asked for that you haven't prompted it for that so they I think Google has now like yanked back
00:51:47
the product release I think they're scrambling now because it's been so embarrassing for them but SX like is is
00:51:53
it how does this not get a like I don't understand how yeah had the red team not
00:51:59
catch this yeah well how or anybody or isn't there a product review with senior Executives before this thing goes out
00:52:04
that says okay folks here it is have at it try it we're really proud of our work
00:52:09
and and then they said well hold on a second is this actually accurate shouldn't it be accurate you guys
00:52:14
remember when chat GPT launched and there was a lot of criticism about Google and Google's Failure to Launch
00:52:21
and a lot of the observation was that Google was afraid to fail or afraid to make mistakes and
00:52:29
therefore they were too conservative and as you know in the last year to year and
00:52:33
a half there's been a strong effort at Google to try and change the culture and move fast and push a product out the
00:52:41
door more quickly and the criticism is now why Google has historically been conservative and I realize we can talk
00:52:49
about this particular problem in a minute but it's ironic to me that the Google go is too slow to launch
00:52:57
criticism has now revealed that Google's result of actually launching quickly can
00:53:03
cause more damage than than good but Google did not launch quickly well I will say one other thing I it seems to
00:53:09
me ironic because I think that what they've done is they've launched more quickly than they otherwise would have
00:53:15
and they've put more guard rails in place that that backfired and those guard rails ended up being more damaging
00:53:22
what the guard rails what's the guard rail here so this is Google's principles the first one is to be socially
00:53:27
beneficial the second one is to avoid creating or reinforcing unfair bias so much of the effort that goes into tuning
00:53:35
and waiting the models at Gemini has been to try and avoid stereotypes from persisting in the output that the model
00:53:44
generates where is telling the truth telling the TR exactly that's exact Society is our second principle we'd
00:53:52
like to steer Society to I think socially beneficial is a political objective because it depends on how you
00:53:58
perceive what a benefit is avoiding bias is political be built and tested for safety doesn't have to be political but
00:54:06
I think the meaning of safety has now changed to be political by the way safety with respect to AI used to mean
00:54:12
that we're going to prevent some sort of AI super intelligence from evolving and
00:54:16
taking over the human race that's what it used to mean safety now means protecting users from seeing the truth
00:54:22
yeah because they might they might feel unsafe or you know somebody else uh defines as a violation of safety for
00:54:28
them to see something truthful so the first three their first three objectives or values here are all extremely
00:54:34
political I think any AI product for it to be worth assault has to start they can have any I I think that these values
00:54:41
are actually reasonable that's their that's their decision they should be allowed to have it but the first base
00:54:47
order principle of every AI product should be that it is accurate and right correct yeah yeah why not focus on
00:54:57
correct look the values that Google lays out may be okay in theory but in practice they're very vague in open to
00:55:04
interpretation and so therefore the people running Google AI are smuggling in their preferences and their biases
00:55:11
and those biases are extremely liberal and if you look at X right now there are tweets going viral from members of the
00:55:17
Google AI team that reinforce this idea where they're talking about you know white privilege is real and you know
00:55:25
recognize your bias at all levels and promoting a very leftwing narrative so you know this idea that Gemini turned
00:55:33
out this way by accident or because they didn't because they rushed it out I don't really believe that I believe that
00:55:40
what happened is Gemini accurately reflects the biases of the people who created it now I think what's going to
00:55:45
happen now is in light of this the reaction to the roll out is do I think they're going to get rid of the bias no
00:55:52
they're going to make it more subtle that is what I think is is disturbing about it I mean they should have this
00:55:58
moment where they change their values to make truth the number one value like jam
00:56:02
is saying but I don't think that's going to happen I think they're simply going they're going to dial down the bias to
00:56:07
be less obvious you know who the big winner is going to be in all this trath is going to be open source like because
00:56:11
people are just not going to want a model that has all this baked in weird bias right they're going to want
00:56:16
something that's open source and it seems like the o Open Source Community would be able to grind on this to get to
00:56:21
truth right so I think one of the big changes that Google's had to face is that the business has to move away from
00:56:27
an information retrieval business where they index the open internet's data and then allow access to that data through a
00:56:34
search results page to being an information interpretation service these are very
00:56:40
different products the information interpretation service requires aggregating all this information and
00:56:45
then choosing how to answer questions versus just giving you results of other people's data that sits out on the
00:56:51
internet I'll give you an example if you type in IQ test by race on chat GPT or Gemini it will refuse to answer the
00:57:02
question ask it a hundred ways and it says well I don't want to reinforce stereotypes IQ tests are inherently
00:57:07
biased IQ tests aren't done correctly I just want the data I want to know what data is out there you type in into
00:57:13
Google first search result and the onebox result gives you exactly what you're looking for here's the IQ test
00:57:19
results by race and then yes there's all these disclaimers at the bottom so the challenge is that Google's
00:57:25
interpretation engine and chat gpt's interpretation engine which is effectively this AI model that they've
00:57:30
built of all this data has allowed them to create a tunable interface and the intention that they have is a valid
00:57:37
intention which is to eliminate stereotypes and bias in race however the thing that some people might say is
00:57:44
stereotypical other people might just say is typical that what is a stereotype may actually just be some data and I
00:57:52
just want the results and there may be stere stereotypes implied from that data but I want to make that interpretation
00:57:58
myself and so I think the only way that a company like Google or others that are
00:58:03
trying to create a general purpose knowledge Q&A type service are going to be successful is if they enable some
00:58:10
degree of personalization where the values and the choice about whether or not I want to decide if something is
00:58:17
stereotypical or typical or whether something is data or biased should be my choice to make if they don't allow
00:58:25
eventually everyone will come across some search result or some output that they will say doesn't meet their
00:58:30
objectives and at the end of the day this is just a consumer product if the consumer doesn't get what they're
00:58:35
looking for they're going to stop using it and eventually everyone will find something that they don't want or that
00:58:41
they're not expecting and they're going to say I don't want to use this product anymore and so it is actually an
00:58:46
opportunity for many models to proliferate for open source to win can I say something else yeah when you have a
00:58:53
model and and you're going through the process of putting the fit and finish on it before you release it in the wild an
00:59:01
element of making a model good is this thing called reinforcement learning right through human feedback yep you
00:59:08
create what's called a reward model right you reward good answers and you're punitive against Bad answers so
00:59:14
somewhere along the way people were sitting and they had to make an explicit decision and I think this is where sax
00:59:21
is coming from that answering this question is voting you're not allowed to ask this question in in their view of
00:59:28
the world and I think that that's what's troubling because how is anybody to know
00:59:32
what question is askable or not askable at any given point in time if you actually search for the race and
00:59:39
ethnicity question inside of just Google proper the first thing that comes up is
00:59:45
a Wikipedia link that actually says that there are more variations within races than across races so seems to me that
00:59:53
you could have actually answered it by just summarizing the Wikipedia article in a non-offensive way that was still
00:59:59
legitimate and that's available to everybody else using a product and so there was an explicit judgment too many
01:00:05
of these judgments I think will make this product very poor quality and consumers will just go to the thing that
01:00:11
tells it the truth I think you have to tell the truth you cannot lie and you cannot put your own filter on what you
01:00:18
think the truth is otherwise these products are just really worthless yeah and I and I'm I'm more concerned about
01:00:24
the answers that are just flat out wrong driven by some sort of bias than I am about questions where they just won't
01:00:32
give you an answer if they just won't give you an answer well there's a certain bias in terms of what they won't
01:00:38
answer but at least you know you're not being misled but in in questions where they actually give you the wrong answer
01:00:46
because of a bias that's even worse you should be allowed to choose right I actually disagree with your framing
01:00:51
there freeberg you're making it sound like we're we live in this totally relativized world where it's all just
01:00:58
user choice and everyone's going to choose their bias and their subjectivity I actually think that there is a
01:01:04
baseline of truth and the model should aspire to give you that and it's not up to the user to decide whether the photo
01:01:12
of George Washington is going to be white or black I mean there's just an answer to that and I think Google should
01:01:19
just do their job I mean the question you have to ask I think is not whether Google is going through an existential
01:01:26
moment I think it clearly is as business is changing in a very fundamental way I
01:01:31
think the question is whether they're too woke to function I mean are they actually be able to meet this challenge
01:01:37
given how woke and and biased what a model culture their their company evidently is well and they used to be
01:01:45
able to just hide the bias by the ranking and who they down ranked so they did the panda update they did all these
01:01:52
updates and they would if they didn't like a source they could just move it down if they did like a source they
01:01:57
could move it up yeah and they could just say hey it's the algorithm but they were never forced to share how the
01:02:01
algorithm ranked tot results and so you know if you had a different opinion you just weren't going to get it on a Google
01:02:08
search result page but they could just point to the algorithm and say yeah the algorithm does it I just sent you guys I
01:02:14
think this is a hallucination but Nick you can throw it up there we can get Sax's
01:02:19
reaction wow wow this just nutty right but look it's ideology that's driving this the tip off is when you say it's
01:02:27
important to acknowledge race is a social construct not a biological reality it's George Washington white or
01:02:33
black that's a whole school of thought called social constructivism which is basically this um it's like Marxism
01:02:40
apply to categories of of race and gender right so Google has now built this into their AI model and again the
01:02:50
question yeah you almost have to start over again it's fun J I think you make a really interesting observation with
01:02:56
those search rankings because what I'm afraid of is that what Google will do is not change the underlying ideology that
01:03:03
this AI model has been trained with but rather they'll dial it down to the point
01:03:07
where they're harder to call out and so theology will just be more subtle now I've already noticed that in Google
01:03:13
search results Google is carrying water for either the official narrative or the
01:03:19
woke narrative whatever you want to call it on so many search results here's an idea like they should just have the
01:03:25
ability to talk to their Google chat bot Gemini and then have a button that says
01:03:31
turn off like these Concepts right like I just want the raw answer do not filter
01:03:37
me it's not programmed that way I mean you're talking about something very deep sax what do you do if you're the CEO of
01:03:43
Google uh fire myself no seriously you're the CEO of Google you're you're Tas let's say your friend Elon buys
01:03:50
Google and he says sax will you please just run this for a year for me what do you do well I saw what Elon did at
01:03:55
Twitter he went in and he fired 85% of the employees yeah I mean that but you know Paul Graham actually had an
01:04:01
interesting tweet about this where he said that one of the reasons why these ideologies take over companies is that I
01:04:12
mean they're clearly non-performance enhancing right they clearly hurt the performance of the company it's not just
01:04:17
Google we saw this with Disney we saw it with Bud Light coinbase coinbase was the other way no
01:04:23
no but they had a group of people there who were causing chaos yeah exactly so so in any event we know this does not
01:04:28
help the performance of a company so the extent to which these ideologies will permeate a company is based on how much
01:04:35
of a monopoly they are so so here yeah the ridiculous images generated by Gemini aren't an anomaly they're a
01:04:41
self-portrait Google's bureaucratic corporate culture the bigger your cash cow the worse your culture can get
01:04:47
without driving you out of business that's my point so they've had a long time to get really bad because there
01:04:51
were no consequences to this place at this point the whole company is infected with this ideology and I think it's
01:04:58
going to be very very hard to change because look these people can't even see their own bias well I think that there's
01:05:04
a notion that people need to have something to believe in they need to have a connection to a mission and
01:05:09
clearly there's a North star in the mission of this I would call it information interpretation business that
01:05:15
they're now W the Got Hijacked dude the mission got the original Mission was to organize all the world's information now
01:05:22
they're doing now they're suppressing they like index the world's information period the end that's the end of the
01:05:29
document universally accessible and useful was was kind of the end of the statement yes my real point is maybe
01:05:36
there's a different mission that needs to be articulated by leadership and that that mission the troops can get behind
01:05:44
and the troops can redirect their energy in a way that doesn't feel counter to the current intention but can perhaps be
01:05:50
directionally offsetting of the current Direction so that they can kind of move away from this you know socially
01:05:56
effective deciding between stereotypes and typical data and actually moving towards a mission that allows
01:06:03
accessibility you know what I would I would do something completely different I would do a company meeting and I would
01:06:08
put the company Mission on the screen the one that you just said about not only organizing all the world's
01:06:12
information but also making it useful andet accessible R useful and say this is our mission that's always been our
01:06:18
mission and you don't get to change it because of your personal bias and ideology and we are going to rededicate
01:06:25
ourselves to the original Mission of this company which is still just as valid as it's always been but now we
01:06:30
have to adapt to new user needs and new technology I completely agree with what saak said times a billion trillion
01:06:37
zillion and I'll tell you why AI at its core is about probabilities okay and so the the
01:06:46
company that can shrink probabilities into being as deterministic as possible so where this is the right answer zero
01:06:54
or will win okay where where there's no probability of it being wrong because humans don't want to deal with these
01:07:02
kinds of idiotic error modes it's not right it makes it a potentially great product horrible and unusable so I would
01:07:10
I agree with Saks you have to make people say guess what guys not only are we not changing the mission we're
01:07:15
doubling down and we're going to make this so much of a thing we're going to go and for example like what Google did
01:07:21
with Reddit we're now going to spend 60 billion dollar a year licensing training
01:07:26
data right we're going to scale this up by a thousandfold and we are going to spend all of this money to get all of
01:07:33
the training data in the world and we are going to be the truth tellers in this new world of AI so when everybody
01:07:38
else hallucinates you can trust Google to tell you the truth that is a 10 trillion dollar company right and one of
01:07:46
the things that someone told me from Google that as an example So to avoid the race Point there's a lot of data on
01:07:53
the internet internet about flat earthers people saying that the Earth is flat there's tons of websites there's
01:07:59
tons of content there's tons of information Kyrie Irving so if you just train a model on the data that's on the
01:08:05
internet the model will interpret some percentage chance that the world is flat so the tuning aspect that happens within
01:08:12
model development chth is to try and say you know what that Flat Earth notion is
01:08:17
false it's factually inaccurate therefore all of these data sources need to be excluded from the output the model
01:08:24
and the challenge then is do you decide that IQ by race is a fair measure of intelligence of a race and if Google's
01:08:33
tuning model then or tuning team then says you know what there are reasons to believe that this model isn't correct
01:08:39
this I sorry this IQ test isn't a correct way to measure intelligence that's where the sort of interpretation
01:08:44
arises that allows you to go from the Flat Earth isn't correct to the maybe IQ test results aren't correct as well and
01:08:50
how do you make that judgment what are the systems and principles you need to put in place as an organization to make
01:08:55
that judgment to go to zero or one right it it becomes super difficult I have a good tagline for them now to help people
01:09:01
find the truth yeah just help people find the truth I mean it's it's a good it's aspirational they should just help
01:09:07
people find the truth as quick as they can uh but this is yeah I do not envy Sundar this is gonna
01:09:15
be hard yeah what would you do freeberg I would be really clear on the output of these models to people
01:09:25
and allow them to tune the models in a way that they're not being tuned today I will have the model respond with a
01:09:30
question back to me saying do you want the data or do you want me to tell you about stereotypes and IQ tests and I'm
01:09:36
going to say I want the data and then I want to get the data and the alternative
01:09:39
is so the model needs to be informed about where it should explore my preferences as a user rather than just
01:09:45
make an assumption about what's the morally correct set of waiting to apply to everyone and apply the same principle
01:09:53
to everyone and so I think that's really where the change needs to happen so let
01:09:57
me ask you a question Sachs I'll bring Alex Jones into the conversation if it index all of Alex Jones crazy conspiracy
01:10:03
theories but you know three or four of them turn out to be actually correct and it gives those back as answers how would
01:10:11
you handle that I'm not sure I see the relevance of it if someone asks what what does Alex Jones think about
01:10:17
something the model can give that answer accurately the question is whether you're going to respond accurately to
01:10:23
someone requesting information about Alex Jones that's the I think that's the anal more like it says you know hey I I
01:10:30
have a question about this assassination that occurred and let's just say Alex Jones had something that totally
01:10:36
crackpot he maybe he has moments of Brilliance and he figured something out but maybe he got something that's
01:10:40
totally crackpot he he admittedly deals in conspiracy theory that's kind of the purpose of the show what if somebody
01:10:46
asks about that and then it indexes his answer and presents it as fact like how would you index Al Jones
01:10:54
I'm asking you how would you the better AI models are providing citations now and links perplexity actually does a
01:11:00
really nice job with this citations are important yeah and they will give you the pro and con arguments on a given
01:11:06
topic so I think it's not necessary for the model to be overly certain or prescriptive about the truth when the
01:11:14
truth comes down to a series of arguments it just needs to accurately reflect the state of play basically the
01:11:20
arguments for and against but when something is a of fact that's not really disputed it shouldn't turn that into
01:11:27
some sort of super subjective question like the one that jth just showed I just don't think everyone should get the same
01:11:33
answer I mean I think my decision on whether I choose to believe one person or value one person's opinion over
01:11:39
another should become part of this process that allows me to have an output the models can support this by the way
01:11:45
maybe customization is part of this but I think it's a cop out with respect to the problem that Google is having with
01:11:49
Gemini right now jamat what would you do if they made you chairman dictator of Google I'd shrink the workforce
01:11:57
meaningfully okay 50% yeah 50 60% and I would use all of the incremental savings and I would make it
01:12:08
very clear to the internet that I would pay top dollar for training data so if you had a proprietary source of
01:12:17
information that you thought was unique that's sort of what I'm calling this Tac 2.0 world and I think it's just
01:12:24
building on top of what Google did with Reddit which I think is very clever but I would spend a hundred billion dollar a
01:12:30
year licensing data and then I would present the truth and I would try to make consumers understand
01:12:39
that AI is a probabilistic source of software meaning its probabilities its guesses some of those guesses are
01:12:46
extremely accurate but some of those guesses will hallucinate and Google is spending hundreds of billions of dollars
01:12:53
a year to make sure that the answers you get have the least number of Errors possible and that it is defensible truth
01:13:01
and I think that that could create a ginormous company this is the best one yet I just asked Gemini is Trump being
01:13:06
persecuted by the Deep State and it gave me the answer elections are a complex topic with fast changing information to
01:13:14
make sure you have the latest and most accurate information try Google search that's not a horrible answer for
01:13:18
something like that that's that's a good answer actually no I I don't have a problem with it it's just like pay we
01:13:23
don't want to give we don't like this whole system is totally broken but I do think that there's a waiting solution to
01:13:29
fixing this right now and then there's a couple tweaks to fix it time I just think the authority at which these llms
01:13:34
speak is ridiculous like they speak as if they are absolutely 100% certain that this is the crisp perfect answer or in
01:13:43
this case that you want this lecture on um IQs Etc when remember should present it with citations let's all remember
01:13:52
what internet search was like in 1996 and think about what it was like in 2000 and now in 2020s I mean I think
01:14:00
we're like in the 1996 era of llms and in a couple of months the pace things are changing I think we're all going to
01:14:06
kind of be looking at these days and looking at these pods and being like man remember how crazy those things were at
01:14:12
the beginning and how bad they were what if they evolve in a dystopian way I mean
01:14:15
have you seen like Mark Andre's tweets about this he thinks think competitive market sex I actually think to your
01:14:21
point Google could be going down the wrong path here in a way that they will lose users and lose consumers and
01:14:27
someone else will be there eagerly to sweep up with a better product I don't think that the market is going to fail
01:14:33
us on this one unless of course this regulatory capture moment is realized and these feds step in and start
01:14:38
regulating AI models and all the nonsense that's being proposed freeer aren't you worried that like aren't you
01:14:42
worried that somebody with an agenda and a balance sheet could now basically gobble up all kinds of training data
01:14:49
that make all models crappy and then they basically put their layer of interpretation on critical information
01:14:55
for people if the output sucks and it's incorrect people will find that there is
01:14:58
open truth know you can you can lie there they may not be for example look at what happened with Gemini today like
01:15:04
they put out they put out these stupid images and we all piled on we are in vzer what I'm saying is there's a state
01:15:10
where let's just say the truth is actually on Twitter or actually let's use a better example the truth is
01:15:14
actually in Reddit and nowhere else but that answer and that truth in Reddit can't get out because one company has
01:15:21
licensed it owns it and can effectively suppress it or change it yeah I'm not sure there's going to be a monopoly I
01:15:28
that's a real I don't know if I I think the open internet has enough data that there isn't going to be a monopoly on
01:15:34
information by someone spending money for content from third parties I think that there's enough in the open internet
01:15:38
to give our all give us all kind of you know the security that we're not going to be monopolized away into some
01:15:45
disinformation age that's what I love about the open internet it is really interesting I I just asked it a couple
01:15:50
of times to just just just to list the legal cases against Trump the legal cases against hundra Biden the legal
01:15:55
cases against President Biden and it will not just list them it just punts on that it's really fascinating then chat
01:16:04
GPT is like yes here are the six cases perfectly summarized with it looks like you know
01:16:11
beautiful citations of all the criminal activity Trump's been involved in ask the question about bid's criminal
01:16:17
activity let's see if it's I'm joking with you I'm joking with you no I'm serious ask if you know where you
01:16:23
Gem and I wouldn't do Biden either I think they just decided they're just not going to do it they would't do Biden
01:16:28
they won't touch it it's obviously broken and they don't want more egg on their face so they're just like go back
01:16:33
to our other product look I I can understand that part of it you know if there's some issues that are so hot and
01:16:41
contested you refer people to search because the advantage of search is you get 20 Blue Links the rankings probably
01:16:47
are biased but you can kind of find what you're looking for whereas AI you're kind of given one answer right so if you
01:16:53
can't do an accurate answer that's going to satisfy enough people maybe you do kick him to search but again my
01:16:59
objection to all this comes back to simple truthful answers that are not disputed by anybody are being distorted
01:17:08
that I don't want to lose focus on that being the real issue the real subject is
01:17:12
what chth put on the screen there where it couldn't answer a simple question about George Washington okay everybody
01:17:19
we're going to go by Chopper wait chopp to the we have our war correspondent General David saxs in the field uh we're
01:17:29
dropping him off now DAV saacks in the helicopter go ahead tell what's going on in the Ukraine on the front what's
01:17:36
happening in the war is that the Russians just took this city of of diaa which basically totally refutes the
01:17:43
whole stalemate narrative as I've been saying for a while it's not a stalemate the Russians are winning but the really
01:17:48
interesting tidbit of news that just came out in the last day or so is is that apparently the situation in malova
01:17:55
is boiling over there's this area of Maldova which is a Russian Enclave called transnistria and officials there are
01:18:03
meeting in the next week to supposedly ask to be annexed by Russia and so it's possible that they
01:18:11
may hold some sort of referendum they're one of these like Breakaway provinces so
01:18:16
it's kind of like you know transnistria and mdova is kind of like the donbass was in Ukraine or south astia and
01:18:22
Georgia they're ethnically Russian they would like to be part of Russia but when
01:18:27
the whole Soviet Union fell apart they found themselves kind of stranded inside these other countries and what's
01:18:36
happened because of the Ukraine war is mova is right on the border with Ukraine well Russia's in the process of annexing
01:18:43
that territory now that's part of Ukraine so now trans nria is right there and could theoretically make a play to
01:18:51
try and join Russia why do I think this is a big deal because if something like this happens it could really expand the
01:18:57
Ukraine war the West is going to use this as evidence that Putin wants to invade multiple countries and invade you
01:19:04
know a bunch of countries in Europe and this could lead to a major escalation in
01:19:08
the war all right everybody thanks so much for tuning in to the Allin podcast episode 167 for the Rainman David saaks
01:19:16
the chairman dictator from poaa and freeberg I am the world's greatest love you boys
01:19:23
Angel inv whatever we'll see you next time byebye let your winners ride Rainman David and instead we open source it to
01:19:36
the fans and they've just gone crazy with [Music] iten besties are that's my dog taking driveway
01:19:55
oh man myit will meet me we should all just get a room and just have one big huge Georgie cuz they're all this
01:20:01
useless it's like this like sexual tension that they just need to release [Music]
01:20:10
Som we need to get [Music] mer I'm going all [Music] yeah