Kind: captions
Language: en
emergency podcast everybody the markets
went crazy today due to the Deep seek
phenomenon that seems to prove that the
age of AI has shifted from training
models to a model's ability to reason
I'll explain why I think that is very
incomplete and why this is the worst
possible time in human history to be the
average investor drew it was quite the
weekend man deep seek took over the
Internet number one on the App Store
they beat open AI at number two Gemini
at number four I seen people tweeting
about it like crazy so please we had to
break our production schedules we shut
down the the team meeting everything
like please explain to us what deep seek
actually is and why it matters in this
moment to us so what deep seek did was
go well you guys spent the billions of
dollars to train uh chat GPT on what the
world is like that's probably the
easiest way to think about it what is
the world like and what I'm going to do
it's something called synthetic data I'm
just going to prompt chat GPT over and
over and over and over and over like a
lot I'm going to prompt The Living
Daylights out of this thing to describe
to me the model that it has hidden
inside of it that it learned by crawling
all of this data and so I'm not going to
do the work of building that model I'm G
to play a game oh this is probably a
terrible example but like Battleship
where I'm like is there anything here or
maybe more accurately what's here when I
ask this question what do you tell me
about the world when I ask this question
what do you tell me about the world and
now I can reduce educe all of these
numbers again going from training cost
100 million to 5 million the gpus that I
need to do it from 100,000 to 2,000 and
so driving all these costs way down just
by going well you know about the world
let me ask you so deep seeks
relationship to chat GPT is the same as
our relationship to chat GPT so as we
prompt CH chat GPT and feel like we're
getting smarter about the world deep
seek was like well I'm just going to do
the same thing I'm going to ask it all
these questions and I'm going to build
my sense of what the world is by
prompting the life out of it which is
why you could ask deep seek what are you
and it would tell you that it's chat GPT
40 because literally it had prompted it
so many times it was basically like well
this is what I am so utterly fascinating
but it really does beg the question of
okay well then are the days of training
behind us and now this is just about
okay you have a picture of the world in
your mind dear llm and now it's about
who can look at that picture of the
world and come up with the most
compelling answers so when I ask it for
whatever you're asking it to do write a
bit of marketing copy uh have it improve
your screenplay which is something that
I'm was literally doing this weekend as
I'm reading all of this
drama and getting really good at that is
now the name of the game and if that's
true and this is why the markets went
crazy if that's true all of these
companies that we've been rewarding by
buying their stock like Nvidia for
making these chips that allow people to
the train well all of a sudden you don't
need as many gpus because you can do all
that training on much less sophisticated
Hardware uh and you can still get a
really compelling answer so hey amazing
gift to humanity that open AI went and
spent all these billions of dollars but
now there's no moat because that exists
and I'm able to C query the life out of
it to build on a much cheaper basis a
working model and as long as my
reasoning is better or at least is good
um I get you all the answers that you
need so now it's like the market was
like whoa whoa whoa then our whole
investment thesis has changed the reason
I think that this this is incredible I I
do not want to skip past how amazing it
is now that people can shift their focus
to reasoning and getting really good at
extracting the wisdom out of the model's
trained sense of what the world is but
would like to remind everybody that
right now what it's doing is just saying
what is the most likely next letter
running a bunch of experiments about
which of those sequences is the most
useful which is utterly transformative
in and of itself but I really believe
that the next moment is going to be okay
that's awesome for copy uh for a lot of
things like law and all that stuff is
all going to be cognitive horsepower um
but I think the next moment where AI
really does the things we want to do
like when you hear me talking about it's
going to drive energy cost to zero and
all that stuff uh you needed to
understand the physical world meaning
you need that same thing at the level of
physics now if I'm right about that then
you're going to have to train the model
on the world of physics and I think
that's going to take another insane
round of training and that probably
shatters into a thousand different
Avenues so understanding um chemistry
understanding uh the way that it works
inside of a closed biological system
understanding quantum physics uh all
that stuff it things that I can't even
begin to imagine things I don't even
know exist that are lanes that people
need to go down to figure out okay how
does it work in this scenario um so it's
that kind of that shift from I'm
regurgitating internet information to
now I understand these topics I
understand physics I understand so
imagine you're in llm and you're running
those thought experiments but as you go
down right now you're just running
predictive analysis in a bunch of
different ways What If instead you were
validating your prediction again
the known laws of physics so that
doesn't help you in marketing copy than
maybe getting better at understanding
human psychological responses like if
you knew oh God this it ends up getting
so complicated but just to finish the
thought if you knew these words and that
sequence triggers this dopamine response
and this type of person holy hell that
would be very powerful to know so that
you would be going down the path of
saying okay I think this is the right
next letter or the right next word or
the right next sentence
based on what I know about
neurochemistry yeah now that's a lot
more powerful than just saying this is
the most common thing that I see and
this right now is the advantage that
humans still have is we understand like
when I read this it makes me feel some
kind of way and if I feel some kind of
way then other people are going to feel
some kind of way but the only way the
llm has any sense that that is true is
more people have put that sequence of
question and answer uh or you know
whatever strings of sentences right out
into the world so I will assume based on
volume that it's
better okay hopefully that gives people
a complete sense of a very simplified
version of this uh why the markets are
reacting the way that they're reacting
now if I'm right that this is all going
to be upended again by the companies
that have balls to still be a company
that's going to spend the billions of
dollars to train on uh physics or you
know chemistry or whatever the case may
be um but what scares me about that is
this moment may stop all of that
investment and watching the market over
the next weeks months we're really going
to see if people like no no no I do not
want to be holding the stock of
companies that are making huge Capital
expenditures into the physical
infrastructure Stargate anyone uh
because that's not the play anymore the
play for the fast money which make no
mistake is what people are here for the
play for the fast money is is in that
the logic part of it the
inference okay if that happens and I'm
right that the next big break through
will require training these models on
things like physics uh then you may get
the market being shy to do that because
the second somebody spends the billions
of dollars to do all that you're going
to get another deep seek like technology
that goes rad thank you for building
that model of the world and now I'm
going to extract everything that I can
from that drive all the cost down to
next to nothing and so how this will
play out I cannot even begin to tell you
I love to look at things as a Sci-Fi
writer and say that it's entirely
possible that we get into people really
being um protective of that data that
they want to wall things off and you saw
in the conspiracy around deep seek which
is very fascinating to watch it it's all
happening in real time but if you went
if you go as of the time of recording
this on Monday uh if you go to deep
seeks website they have restricted
access to deep seek now the initial take
on this on X was that this is because um
the CCP is behind deep seek and they
launched deep seek R1 on Trump's
inauguration day as a way to say oh you
think you're going to Isis out you guys
tried to make it impossible for us to
access these chips cool so that we
couldn't build the training models and
we found found a way around it using way
inferior technology the ones that we
could get made it way more efficient and
we're still able to match you guys
within months so you thought you had a
year or twoyear head start on this you
have a month two months nobody's going
to care and so we're now at parody now
what has actually happened at least as
it seems right now is that people have
started attacking deep seek they said
malicious attacks of our site and that's
why we're restricting access so yes
right now they're restricting it to
people with Chinese phone numbers but
everybody that already has it and
downloaded already has an account can
still get one now we'll find out what's
true if this is really a CCP play to say
haha [ __ ] you thought you were
going to be able to limit us we dropped
this on inauguration date now we're
limiting access it's only for people in
China but now you know it exists wahaha
uh I don't think that's going to be how
this plays out there's certainly
censorship and all that which is a topic
for another day um but
this is definitely a wakeup call to
America uh so let's talk about what Mark
andreson said that's exactly where I was
going can we pull up that Mark andrion
uh tweet about deep seek R1 is AI
Sputnik M I honestly I don't know what
sputnick is so you might have to break
this down give me the the elementary
level for me okay so uh I'm sure for a
lot of people so um Sputnik was what
1956 1958 something like that uh so
Russia launches the first um man
satellite into space not satellite they
they sent somebody into space they
didn't get him to the moon but I think
he was the first person to orbit uh and
so it was a big deal and it was like
whoa Russia uh communist country they're
ahead of us oh my God they're going to
win the space race this lit a fire under
America's ass and I would say leads to
um JFK saying we're going to get to the
Moon by the end of this decade not
because it easy but because it is hard
okay so it's this whole Cold War
mentality of we now have an enemy and
they're making a sweat it's like
everybody if you're trying to do
something great in life having a worthy
adversary is incredibly useful but I
have a feeling that a cold war is going
to be good for Humanity for sure uh may
even be good for America because of this
Spirit of we can't lose we absolutely
cannot let China get ahead of us now I
think people are going to forget that as
of right now deep seek is open source so
it's on GitHub anybody can grab it they
can download it um but there is
something unnerving about seeing your
opponent move that fast that you're like
huh I don't know if we're ahead and it's
going to matter a lot who wins the race
on AI so I think that this is really
going to wake us up I have I have felt
like I've been screaming into the void
for a long time about the thing that if
I really want to piss people off I'll
just say you need to work hard you need
to work smart and you need to work long
hours if I say those three things it
triggers the internet and and my X feed
is going to be full of people hey
dumbass uh if I'm working hard and smart
why do I have to work long hours and I
will just say because China's going to
hand you your ass you're always going to
have somebody else is like oh they're
working hard and smart rad then I'm
going to work hard smart and long hours
the somebody is going to do all three
and when you meet them you will lose
that's a riff on a famous quote um
that I think is really true and I really
hope that this does in the way that Mark
andreon is nodding to inspire us to
[ __ ] go hard in the paint my message
to every young person out there genius
is a young man's game I am doing my best
Drew to outrun the truth of that quote
and I don't think I'll be able to do it
and so it really is going to take this
next Generation to be hungry and to say
I don't give a [ __ ] if you torched the
uh housing market I don't care if you're
inflating my money out into nothing I'm
still going to win so bad policy doesn't
block dunks any more than booze don't
block dunks now sadly that isn't true
and bad policy will [ __ ] you up but you
really can at least still in America I
want to believe you can still overcome
if you inspire people to go hard enough
in the paint and I believe that national
pride is a huge huge thing for sure but
individual
Glory individual sovereignty and Freedom
the ability to flex hard to get laid is
still gonna [ __ ] drive this insane
innovation in America uh and this now
reminds me of how I felt growing up in
the 80s where it was like us versus
Russia now look it was all fake and it
had us walking this Razor's Edge of
blowing each other up but it also pushed
us man like it was crazy and it felt
dope it felt dope to be like we're
winning and we've got this adversary and
we've got to outperform them and it
really does rally people it hits them in
this emotional level so anyway I feel
like I've been screaming into the void
about that for a long time because it
doesn't come back to me with people
being like yeah yeah yeah like I want to
do that and when you get people that
really want to win man like they are on
the field and they are playing to
[ __ ]
win extraordinary things come out of
that and I would say China is a very
worthy adversary there's a quote in The
Social Network where um Mark Zuckerberg
as a character says China has more uh
Geniuses than we have anything more
people with genius IQs living in China
than there are people of any kind living
in the United States that can't possibly
be true it is now if that's true who
it's going to it isn't going to be easy
man like you really have a formidable
Foe and people that want to throw racist
racism on the situation you're going to
lose
because smart and hardworking people are
the world over yeah and the second you
try to be like uh like paint Any Nation
with any brush woof mistake mistake and
so yeah taking these guys seriously I
think it be wise there is still cause
for optimism though David saxs put out
this tweet deep seek R1 shows that the
AI race will be very competitive and
that President Trump was right to resin
the Biden EO which hamstrung American AI
companies without asking whether would
do the same obviously not I'm confident
in the us but we can't be complacent
this this is the vibe for me and the
fact that he is the AI and cryptos are
like having that come from him uh I
think sets the right tone and here's the
thing the youth get whatever country
they want and so it's going to be a
question do they want a country that is
dominant globally or not and if they do
they will get it because we have got we
are still the place that people the
world over want to come we are still the
place as generation after generation
after generation of Hardcore
[ __ ] and so if we can harness
that in policy and in cultural
enthusiasm we've got a shot nice now we
have to Circle back to the investor
markets because markets across the board
were down this morning some people said
Nvidia was done people were swinging I
seen shorts I seen all types of stuff
why did this deep seek announcement have
such a rampant effect on the stock
market okay so this is where uh I'm
going to start at the beginning and I'm
going to speedrun my analysis of uh life
in
20125 your money is being inflated which
either makes you know intellectually or
intuitively that saving money is a
losing strategy which it quite literally
is people that save money are actively
losing money so you have to do something
with it some people just spend it
willy-nilly YOLO their way through life
and they're never going to get ahead uh
some people will Gamble and market and
to get ah and if focus in on the that we
are beinged to gamble at a time where
the world is changing so rapidly that
people are just going to get kicked in
the face and people that win one day are
going to lose the next that that is the
reality when you are trying to use any
gambling mechanism as the mechanism to
get ahead the way I would love to see it
and I am a simple mind fine but the way
I would love to see it is people know
number one I save a certain amount of my
money and then I take a certain portion
of that and I'm going to try either
longterm over the stock market I'm going
to try to get ahead or I'm going to do
day trading just because it's fun it's
exhilarating and it gives me a chance to
be right and beat people so you have a
lot of very smart people trying to put
together a thesis on where the world is
going and then they're going to bet in
the markets that their thesis is correct
now when you have a stable Market where
changes happen over years hey you can do
that you can get a nice stable way of
looking at the world that pays off over
time call it value investing when you
have a world where you get over the
weekend people's entire view of what the
future is changes on a dime and PS The
Magnificent Seven which have carried the
vast majority of the gains in the stock
market are all predicated on technology
technology right now largely being
carried by ai ai suddenly over the
weekend just gets wildly disrupted and
so I'm sitting there going right now I
am glad that I'm so certain I cannot see
the future clearly that I don't make big
bets in the stock market so uh go up go
down I'm so conservative that it it
isn't going to make a big difference to
me but it means that I don't win big it
also means that I don't lose big which
is great but I'm not winning big so
that's where this really gets hard
because if you go in and tell people
stop betting on the market right now the
future is so uncertain they're going to
say well the part of the future that I
can see pretty clearly is that my money
is being inflated to zero and so they
can't stop they literally can't stop
you'll lose your money uh so I hate that
and because the future is changing so
rapidly this really is the worst time in
human history to be trying to make money
off of a thesis about where things are
going for the average investor I get
that your super bright quants will be
able to do it but Jesus now I remember
you said last week we could barely see
Over the Horizon so let alone trying to
have a 30-year portfolio based on Nvidia
or based on Tesla based on some of these
companies where our the entire remote of
AI was it costed too much to build a
data center so we're the best that can
do it Elon famously spent a bunch of
money and built a data center in six
weeks everybody was kind of applauding
how we can invest massive amount of
capital to get this strategic Advantage
just for to your point overnight that
Advantage is now gone so is there any
moat left in technology with AI like
they what is this a day one rebuild if
I'm open AI like how do how does the
company respond to when their remote is
fundamentally removed this is the part
that I can't see I don't know if money
so uh I call it the cover story of the
stock market the cover story of the
stock market which is just gambling but
nobody wants to call it that is that and
this is true so it's not a fake cover
story it just isn't why people put their
money into it but the result of them
putting their money into it is that
these companies are able to aggregate
capital and deploy that against some
thing that they want to bring to Market
that they think will be incredibly
valuable to people people are going to
spend their money on that thing cool
okay we want that to happen it's
wonderful the markets are wonderful I
love it the most I just don't love the
compulsory nature of having to gamble
okay so now the question becomes all the
individuals that are allocating their
Capital trying to get these fast winds
off the stock market are they going to
want to invest in companies that are
making huge cap X outlays or not and if
they're not then now you get back into
and right now I am in Wild
prognostication phase I don't know if
this is going to play out but this is
how it hits me right now that we're back
into the early days where like we were
promised flying cars and we got 140
characters of if everybody is just going
to just reasoning just reasoning just
reasoning meaning inference that no
nobody cares about training new models
anymore it's just making do with the
models that we already have that's
saying that the models are now as good
as they're ever going to be and we're
just going to try to extract more out of
them my intuition which everybody should
discard but my intuition is that's not
going to be the play that until you have
a working model that can make sense of
quantum fluctuations or whatever the
[ __ ] that is until you can map what dark
energy is um until you can predict
exactly how the particles collided in a
um like the Large Hadron Collider a
particle accelerator until you can
predict all of that with 100% accuracy
you have not yet mapped out the actual
real world uh and I think that's
ultimately where we want to end up but I
can't imagine anything more Capital
intensive than trying to solve those
problems now could just be my ignorance
but I would be sad intuitively to see us
collapse back to just make use of what
we have now versus training the life out
of these but Capital allocators are
going to do what they think is going to
yield a return to them period full stop
end of story uh and so this is where you
get into well that's what the government
is for and they're supposed to invest in
things like this and I don't know man
we'll see if you are um pushing your
government to do entitlements over
everything you just don't have the
capital left for this uh it's going to
be weird I I I have a vision of the
future that I know will have a wrecking
ball taken to it like every 72 hours
it's it's crazy like that's not literal
but that's directionally correct and so
it just makes it hard to build a stable
thesis yeah and then for this one I want
you to kind of take off your AI hat put
your entrepreneur hat on like how do you
outrun your competition like to me like
if my moot was gone over the weekend
like is all right let's pack it up guys
like you spit you put your heart into
this into this company your
entrepreneurship journey is so important
to you and then somebody on the other
side of the world to your point is
working faster longer and harder and
then now your strategic Advantage is
gone like that has to be all I can do is
tell you how you think about it as an
entrepreneur so as an entrepreneur you
say Where am I making Revenue uh in fact
here's a quick 101 for our entrepreneurs
out there uh man this is so simple to do
it's very hard it's simple to do in
theory and very hard to do in practice
you stop telling yourself a story CU
right now open AI all of them they're
telling themselves a story stop telling
yourself a story look at the reality of
how you generate money the stock price
for a lot of these guys is so important
and so now our access to Capital is
predicated on what's going on in the
market okay well now then we have to do
the things that the people in the market
want us to do I don't know what the
people in the market are going to want
to do for the medium to long term the
short term is all eyes on deep seek and
things like that so now we're moving
over to inference or at least the market
is like I think people are going to do
that so we'll see so you look at that
you look at what you sell to the
customer what is the customer actually
willing to pay for you look at Deep seek
you see that o it's almost as good as um
open AI 40 and if they're doing it for
uh an absolute fraction of the cost
making this open source then that's not
going to survive for very long so I know
I'm going to have to find something else
so they are going to ask themselves one
question and one question only where can
I put the smallest investment to get the
largest return and if deep seek proves
day after day that that's in being
really good at extracting better
Knowledge from the models we already
have all the companies that are way too
far over their skis and building out
these large physical things they're
going to be in trouble and they may
literally get relegated to the dust bin
and then other people come in and go all
right this is the new thing this is what
people care about and so we're going to
be able to give them that
and that's it like that is how the world
will play out what are people willing to
pay for period full stop end of story
man we shall see uh that's all I got all
right everybody if you haven't already
be sure to subscribe and until next time
my friends be legendary take care peace
if you like this conversation check out
this episode to learn more this just in
an asteroid is hurdling towards Earth
and it's called AI billionaire Chris
Saka and the dinosaurs tell us to brace
for impact all while X reminds us that
Sam Alman himself