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