00:00:00
All right, everybody. Welcome back to
00:00:01
the AllIn podcast. The number one
00:00:03
podcast in the world. You got what you
00:00:05
wanted, folks. The original Quartet is
00:00:08
here live from DC with a great shirt. Is
00:00:12
that is your habitasher making that
00:00:13
shirt or is that a Tom Ford? That white
00:00:16
shirt is so crisp, so perfect. David
00:00:18
Sachs, you're talking about me. You're
00:00:20
Zar. You're Zary. I'll tell you exactly
00:00:22
what it is. I'll tell you what it is.
00:00:24
You can tell me if it's right. Brioni.
00:00:25
Yes, of course it's Brioni. Briion
00:00:28
spread car. Look at that. How many years
00:00:30
have I spent being rich? When a man
00:00:32
turns 50, the only thing he should wear
00:00:34
is broni. The stitching is Looks very
00:00:37
luxurious. That's how Chimath knew,
00:00:38
right, Jimoth? How'd you figure it out?
00:00:40
The stitching? It's just how it lays
00:00:42
with the collar. To be honest with you,
00:00:43
it's the button catch. The Brion has a
00:00:46
very specific style of button catches.
00:00:47
If you don't know what that means, it's
00:00:49
because you're [Â __Â ] ignorant,
00:00:50
malcontent yourself. I'm looking it up
00:00:52
right now. Right. Yeah.
00:00:55
I just asked you to continue.
00:00:59
We'll let your winners ride.
00:01:02
[Music]
00:01:07
We open source it to the fans and
00:01:08
they've just gone crazy with it.
00:01:14
All right, everybody. The All-In Summit
00:01:15
is going into its fourth year September
00:01:17
7th through 9th and the goal is of
00:01:19
course to have the world's most
00:01:20
important conversations. Go to allin.com
00:01:23
yada yada yada to join us at the summit.
00:01:25
All right. It's a lot in the docket, but
00:01:27
there's kind of a very unique thing
00:01:29
going on in the world. David, everybody
00:01:31
knows about AI doomerism. Basically,
00:01:33
people who are concerned uh rightfully
00:01:35
so that AI could have some, you know,
00:01:38
significant impacts on the world. Dario
00:01:41
Amod said he could see employment spike
00:01:44
to 10 to 20% in the next couple years.
00:01:46
They're 4% now as we've always talked
00:01:47
about here. He told Axio that AI
00:01:49
companies and government needs to stop
00:01:51
sugar coating what's coming. He expects
00:01:53
a mass elimination of jobs across tech,
00:01:55
finance, legal, and consulting. Okay,
00:01:57
that's a debate we've had here. And
00:01:59
entrylevel workers will be hit the
00:02:01
hardest. He wants law makers to take
00:02:03
action and more CEOs to speak out. Poly
00:02:06
market thinks regulatory capture via
00:02:08
this AI safety bill is very unlikely. US
00:02:10
enacts AI safety bill in 2025 currently
00:02:13
stands at a 13% chance. But uh Sax, you
00:02:15
wanted to discuss this because it seems
00:02:17
like there is more at work than just a
00:02:20
couple of technologists with I think
00:02:22
we'd all agree there are legitimate
00:02:25
concerns about job destruction or job
00:02:29
and employment displacement that could
00:02:30
occur with AI. We all agree on that
00:02:32
where we're seeing robo taxis start to
00:02:34
hit the streets and I don't think
00:02:35
anybody believes that being a cab driver
00:02:37
is going to exist as a job 10 years from
00:02:38
now. So there seems to be something here
00:02:42
about AI dumerism but it's being taken
00:02:44
to a different level by a group of
00:02:46
people maybe uh with a different agenda.
00:02:49
Yeah. Well, first of all, let's just
00:02:50
acknowledge that there are concerns and
00:02:53
risks associated with AI. It is a
00:02:56
profound and transformative
00:02:58
technology and there are legitimate
00:03:00
concerns about where it might lead. I
00:03:02
mean the future is unknown and that can
00:03:05
be kind of
00:03:06
scary. Now, that being said, I think
00:03:09
that when somebody makes a pronouncement
00:03:12
that says something like 50% of white
00:03:14
collar jobs are going to be lost within
00:03:15
2 years, that's a level of specificity
00:03:19
that I think is just unknowable and is
00:03:22
more associated with an attempt to grab
00:03:24
headlines. And to be frank, if you go
00:03:27
back and look at Anthropic's
00:03:30
announcement or Daario's announcement,
00:03:33
there is a pattern of trying to grab
00:03:35
headlines by making the most
00:03:36
sensationalist version of what could be
00:03:40
a legitimate concern. If you go back
00:03:42
three years ago, they created this
00:03:44
concern that AI models could be used to
00:03:47
create bioweapons.
00:03:49
And they showed what was supposedly a
00:03:52
sample I think of claw generating an
00:03:56
output that can be used by a
00:03:57
bioteterrorist or something like that.
00:03:59
And on the basis of that it actually got
00:04:01
a lot of play and in the UK Rishiun got
00:04:06
very interested in this cause and that
00:04:08
led to the first AI safety summit at
00:04:10
Bletchley Park. So that sort of concern
00:04:13
really drove some of the initial AI
00:04:16
safety concerns. But it turns out that
00:04:17
that particular output was discredited.
00:04:20
It wasn't true. I'm not saying that AI
00:04:23
couldn't be used or misused to maybe
00:04:26
create a boweapon one day, but it was
00:04:28
not an imminent threat in the way that
00:04:30
it was portrayed. There have been other
00:04:31
examples of this. You know, obviously
00:04:33
people are concerned about could the AI
00:04:35
develop into a super intelligence that
00:04:37
grows beyond our control. Could it lead
00:04:39
to widespread job loss? I mean, these
00:04:42
are legitimate things to worry about,
00:04:44
but I think these concerns are being
00:04:46
hyped up to a level that there's simply
00:04:48
no evidence for. And the question is
00:04:50
why? And I think that there is an agenda
00:04:54
here that people should be concerned
00:04:56
about. So, let's start with maybe
00:05:00
Freeberg things that we all agree on
00:05:02
here. There are millions of people who
00:05:05
drive trucks and Ubers and lifts and
00:05:08
door dashes.
00:05:10
You would, I think, agree the majority
00:05:12
of that work in but 5 to 10 years, just
00:05:15
to put a number on it, will be done by
00:05:18
self-driving, robots, cars, etc.,
00:05:21
trucks. Yeah, Dave, I think it's that
00:05:24
might be the wrong way to look at it or
00:05:26
I wouldn't look at it that way. And
00:05:27
maybe I'll just give frame it a
00:05:29
different way, please. If I'm deploying
00:05:32
capital, let's say I'm a CEO of a
00:05:35
company and I can now have
00:05:40
software that's written by AI. Does that
00:05:43
mean that I'm going to fire 80% of my
00:05:45
software engineers? Basically, it means
00:05:48
one software engineer can output, call
00:05:50
it 20, 50 times as much software as they
00:05:53
previously could by using that software
00:05:55
generation tool.
00:05:57
So the return on the invested capital,
00:06:01
the money I'm spending to pay the salary
00:06:03
of that software engineer is now much
00:06:06
much higher. I'm getting much more out
00:06:08
of that person because of the unlocking
00:06:10
of the productivity because of the AI
00:06:12
tool than I previously could. So when
00:06:15
you have a higher ROI on deployed
00:06:18
capital, do you deploy more capital or
00:06:21
less capital? Suddenly you have this
00:06:23
opportunity to make 20 times on your
00:06:25
money versus two times on your money. If
00:06:28
you have a chance to make 20 times on
00:06:29
your money, you're going to deploy a lot
00:06:30
more capital. And this is the story of
00:06:33
technology going back to the first
00:06:36
invention of the first technology of the
00:06:39
caveman. When we have this ability to
00:06:41
create leverage, humans have a tendency
00:06:44
to do more and invest more, not less.
00:06:47
And I think that's what's about to
00:06:49
happen. I think we see this across the
00:06:52
spectrum. People assumed, "Oh my gosh,
00:06:54
software can now be written with one
00:06:55
person. You can create a whole startup.
00:06:57
You don't need to have venture capital
00:06:58
anymore. In fact, what I think we're
00:07:00
going to see is much more venture
00:07:01
capital flowing into new tech startups.
00:07:03
Much more capital being deployed because
00:07:05
the return on the invested capital is so
00:07:07
so so much higher because of AI." So
00:07:10
generally speaking, I think that the the
00:07:12
premise that AI destroys jobs is wrong
00:07:15
because it doesn't take into account the
00:07:17
significantly higher return on invested
00:07:19
capital, which means more capital is
00:07:20
going to be deployed, which means
00:07:21
actually far more jobs are going to be
00:07:23
created, far more work is going to get
00:07:24
done. And so I think that the
00:07:26
counterbalancing effect is really hard
00:07:27
to see without taking that zoomed out
00:07:29
perspective. To to to respond to Sax's
00:07:32
point, I do think anytime you see a
00:07:34
major change socially, society, there's
00:07:37
a vacuum. How's the system going to
00:07:39
operate in the future? And anytime
00:07:41
there's a vacuum in the system, a bunch
00:07:43
of people will rush in and say, I know
00:07:45
how to fill that vacuum. I know what to
00:07:46
do because I am smarter, more educated,
00:07:50
more experienced, more knowledgeable,
00:07:51
more moral. I have some superiority over
00:07:54
everyone else. And therefore, I should
00:07:56
be in a position to define how the new
00:07:58
system should operate. And so, there's a
00:08:00
natural kind of power vacuum that
00:08:02
emerges anytime there's a major
00:08:03
transition like this. and there will be
00:08:05
a scrambling and a fighting and a whole
00:08:07
bunch of different representation.
00:08:08
Typically fear is a great way of getting
00:08:10
into power and people are going to try
00:08:12
and create new control systems because
00:08:13
of the transition that's underway. Okay.
00:08:15
You're going to see this around the
00:08:16
world. Yeah. I mean, uh, so Chimath,
00:08:19
it's pretty clear, you know, Freeberg
00:08:22
didn't answer this question
00:08:23
specifically, so I'm going to give it to
00:08:24
you again. You would agree jobs like
00:08:28
driving things are going to go away. If
00:08:30
we had to pick a number somewhere
00:08:32
between 5 and 10 years, the majority of
00:08:33
those would go away. he's positioning,
00:08:35
hey, a lot more jobs will be created
00:08:36
because there'll be all these extra
00:08:38
venture capital and opportunities, etc.
00:08:41
But job displacement will be very real
00:08:43
and we're seeing, I think, job
00:08:45
displacement. Now, you had a tweet
00:08:46
recently, you know, you were talking
00:08:47
about entry level jobs and how that
00:08:50
seems to be going away in the white
00:08:52
collar space. So, where do you land on
00:08:55
job displacement? Freeberg's already
00:08:57
kind of given the big picture here, but
00:08:58
let's step back to for people who are
00:09:00
listening who have relatives who drive
00:09:02
Uber or a truck or are graduating from
00:09:05
college and want to go work at a, you
00:09:08
know, I don't know, the Magnificent 7 or
00:09:10
in tech and they're not hiring and and
00:09:13
we know the reason they're not hiring
00:09:14
because they're leaning into AI. So,
00:09:17
let's talk about the job displacement in
00:09:18
the medium term. I'm going to ignore
00:09:21
your question and I'm going to Why
00:09:24
should you be any different than the
00:09:25
other? So I now content on this podcast.
00:09:29
There's two people not wanting to answer
00:09:31
the question about job displacement.
00:09:32
Interesting trend. No, no. We'll go back
00:09:34
to that. Let me start by just saying
00:09:35
that it seems that these safety
00:09:38
warnings tend to be pretty coincidental
00:09:41
with key fundraising moments in
00:09:43
anthropic journey. So let's just start
00:09:45
with that. And if you put that into an
00:09:48
LLM and try to figure out if what I just
00:09:50
said was true, it's interesting, but you
00:09:52
find it's relatively accurate. I think
00:09:54
that there is a very smart business
00:09:57
strategy here. And I've said a version
00:09:59
of this about the other companies at the
00:10:01
foundational model layer that aren't
00:10:04
Meta and Google because Meta and Google
00:10:06
frankly sit on these money gushers where
00:10:09
they just generate so much capital that
00:10:11
they can fund these things to infinity.
00:10:13
But if you're not them, so if you're
00:10:15
OpenAI or if you're anthropic, you have
00:10:18
to find an angle. And I think the angles
00:10:20
are slightly different for both. But I
00:10:23
think what this suggests is that there's
00:10:24
a pattern that exists and I think that
00:10:26
that
00:10:27
explains some of the framing of what we
00:10:30
see in the press, Jason, and why we get
00:10:32
these exaggerated claims. Perfect.
00:10:36
So there are people who are doing this
00:10:39
for nefarious reasons is I I guess where
00:10:42
you're sort of getting at here. It's a
00:10:44
way to market. It's smart. It's smart.
00:10:47
If you fall, it's up to you. Yeah. Okay.
00:10:50
Well, there's also an industrial complex
00:10:53
according to some folks that
00:10:56
are backing this. If you've heard of
00:10:59
effective altruism, that was like this
00:11:01
uh movement of a bunch of I don't know,
00:11:04
I guess they consider themselves
00:11:05
intellectual sachs and they uh were kind
00:11:10
of backing a large swath of
00:11:13
organizations that I guess we would call
00:11:16
in the industry astroturfing or what do
00:11:18
they call it when you make so many of
00:11:19
these organizations that they're not
00:11:23
real in politics and flooding the zone
00:11:25
perhaps. So if you were to look at this
00:11:27
article here, Nick, I think you have the
00:11:29
AI existential risk um industrial
00:11:33
complex graphic there. It seems like a
00:11:36
group of people according to this
00:11:37
article have backed to the tune of 1.6
00:11:40
billion, a large number of organizations
00:11:42
to scare the be Jesus out of everybody
00:11:45
and make YouTube videos, Tik
00:11:47
Toks, and they've they've made a map of
00:11:49
it. There's some key takeaways here from
00:11:52
that article where it says here that
00:11:55
it's an inflated ecosystem. There's a
00:11:57
great deal of redundancy. Same names,
00:11:59
acronyms, logos with only minor changes.
00:12:01
Same extreme talking points. Same group
00:12:03
of people just with different titles.
00:12:04
Same funding source. There's a funding
00:12:06
source called Open Philanthropy which
00:12:08
was funded by Dustin Moskovitz who is
00:12:10
one of the Facebook billionaires. Jim,
00:12:13
you worked with him, right? I mean he
00:12:14
was wasn't he like Zuck's roommate at
00:12:17
Harvard or something and one of the
00:12:18
first engineers made a lot of money so
00:12:21
he funded this he he's he's an EA and he
00:12:24
funded this group called Open
00:12:25
Philanthropy which then has become the
00:12:27
feeder for essentially all these other
00:12:30
organizations which are almost different
00:12:32
fronts to basically the same underlying
00:12:34
EA
00:12:35
ideology and what's interesting is that
00:12:37
the guy who set this up for Dustin
00:12:40
Holden Karnovski who is a major
00:12:43
effective altruist and was doing out all
00:12:44
the money. He's married to Daario's
00:12:47
sister and she she's I guess associated
00:12:49
with EA and she was one of the
00:12:50
co-founders of Anthropic. So these are
00:12:53
not coincidences. I mean the reality is
00:12:55
there's a very specific ideological and
00:12:58
political agenda here. Now what is that
00:13:01
agenda? It's basically global AI
00:13:04
governance if you will. They want AI to
00:13:07
be highly regulated but not just at the
00:13:09
level of the nation state but let's say
00:13:12
internationally supernationally to
00:13:15
well if you just do a quick search on
00:13:18
global compute governance it'll tell you
00:13:21
what the key aspects are so number one
00:13:23
they want regulation of computational
00:13:26
resources this includes access to
00:13:30
GPUs they want AI safety and security
00:13:33
regulation they want international you
00:13:35
call from globalist agreements and they
00:13:38
want ethical and societal considerations
00:13:40
or policy built into this. Now, what
00:13:42
does that sound like? That sounds a lot
00:13:44
to me like what the Biden administration
00:13:46
was pursuing. Specifically, we had that
00:13:49
Biden executive order on AI, which was
00:13:51
100 pages of Bernson regulation that was
00:13:54
designed to promote AI safety, but had
00:13:57
all these DEI requirements. So, you
00:13:59
know, it led to woke AI. You remember
00:14:01
when Google launched Black George
00:14:04
Washington and so forth, they had the
00:14:06
Biden diffusion rule which created this
00:14:08
global licensing framework to sell GPUs
00:14:10
all over the world. So extreme
00:14:12
restrictions on proliferation of servers
00:14:15
of computing power. They created what's
00:14:19
called the AI safety institute and they
00:14:22
again fostered these international AI
00:14:24
summits. So if you actually look at what
00:14:28
the Biden administration was tangibly
00:14:30
doing in terms of policy and you look at
00:14:32
what EA's agenda is with respect to
00:14:35
global compute governance, they were
00:14:37
pushing hard on these fronts. And now if
00:14:40
you look at the level of personnel,
00:14:41
there were very very
00:14:43
powerful Biden staffers who now all work
00:14:46
in anthropic. So probably the most
00:14:49
powerful Biden staffer on AI over the
00:14:52
past four years was a lawyer named Tun
00:14:55
Chabra and he now works at Anthropic for
00:14:59
Daario. Elizabeth Kelly who was the
00:15:01
founding director of the AI safety
00:15:03
institute in the government now works at
00:15:06
Anthropic. Like I mentioned Daario's
00:15:10
sister is married to Holden Karnnowski
00:15:12
who dos out all the money to these EA
00:15:14
organizations. So if you were to do
00:15:16
something like create a network map, you
00:15:18
would see very quickly that there's
00:15:20
three key nodes here. There's the
00:15:22
effective altruist movement of which Sam
00:15:25
Bankman Freed's the most notable member
00:15:26
but which I think Dustin Mos is now the
00:15:29
main funer. There's the Biden
00:15:30
administration and like the key staffers
00:15:32
and then you've got anthropic and it's a
00:15:36
very tightly wound network. Now why does
00:15:39
this matter? Let's get Yeah. Also the
00:15:41
goals I think is Yes. Well, the the
00:15:43
goal, like I said, is is global compute
00:15:45
governance. It's basically establishing
00:15:47
national and then international
00:15:49
regulations of AI. Now, but they would
00:15:52
claim, let's just pause here for a
00:15:54
minute. They would claim the reason
00:15:55
they're doing it. And so, we we'll we'll
00:15:58
save if we believe this or not, but they
00:16:01
are concerned about job destruction in
00:16:03
the short term. They're also concerned
00:16:05
as science fiction as it is that the AI
00:16:08
when we get to like a sort of
00:16:10
generalized super intelligence is going
00:16:12
to kill humanity. That this is a nonzero
00:16:14
chance. Elon has said this before.
00:16:16
They've sort of taken it to a almost
00:16:18
like a certainty. Yes, we're going to
00:16:20
have so many of these general
00:16:22
intelligences. But they only believe
00:16:24
that when they're raising money. Well,
00:16:26
that's what I'm sort of getting at.
00:16:27
Like, so I think they believe it all the
00:16:28
time, but maybe maybe the press releases
00:16:30
are time for for the fun building. But
00:16:32
let me let me answer that. Right. Yeah.
00:16:35
Yeah. Look, I mean, it is a great
00:16:37
product. Claude kicks ass. I'm more
00:16:38
interested in the political dimension of
00:16:39
this. I'm not bashing a specific product
00:16:42
or company. But look, I think that there
00:16:44
is some nonzero risk of AI growing into
00:16:48
a super intelligence that's beyond our
00:16:50
control. They have a name for that. They
00:16:51
call it X risk or existential risk. I
00:16:55
think it's very hard to put a percentage
00:16:56
on that. I'm willing to acknowledge that
00:16:59
is a risk. You know, I think about that
00:17:00
all the time and I do think we should be
00:17:02
concerned about it. But there's two
00:17:04
problems I think with this approach.
00:17:05
Number one is X-Risk is not the only
00:17:08
kind of risk. I would say that China
00:17:10
winning the AI race is a huge risk. I
00:17:12
don't really want to see a CCP AI
00:17:15
running the world. And if you hobble our
00:17:19
own innovation, our own AI efforts in
00:17:21
the name of stomping out every
00:17:22
possibility of X-risk, then you probably
00:17:25
end up losing the AI race to China
00:17:27
because they're not going to abide by
00:17:28
those same regulations. So again, you
00:17:31
can't optimize for solving only one risk
00:17:34
while ignoring all the others. And I
00:17:36
would say the risk of China winning the
00:17:38
AI race is, you know, it might be like
00:17:42
30%. Whereas I think X risk is probably
00:17:44
a much lower percentage. So there are
00:17:47
there are other risks to to worry about.
00:17:49
And I I do think that they are
00:17:50
single-mindedly focused on scaring
00:17:53
people with some of these headlines
00:17:55
around first it was the bioweapons then
00:17:57
it was the super intelligence now it's
00:17:58
the job loss and I think it's a tried
00:18:01
andrue tactic of people who want to give
00:18:05
more power to the government to scare
00:18:07
the population right because if you can
00:18:09
scare the population and make them
00:18:11
fearful then they will cry out for the
00:18:13
government to solve the problem and
00:18:15
that's what I see here is that you've
00:18:16
got this elaborate network of front
00:18:20
organizations which are all motivated by
00:18:22
this EA ideology. They're funded by a
00:18:25
hardcore leftist. And by the way, I
00:18:27
became aware of Dustin's politics
00:18:30
because of the Chase Bodin recall. I
00:18:33
found out that he was a big funer of
00:18:34
Chase Bodin. Remember this? Yeah. Dustin
00:18:36
Mossmus and Carrie Tuna, his wife.
00:18:39
Also, Reed Hastings just joined the
00:18:41
board of of Anthropic. remember when he
00:18:45
back in 2016 to tried to drive Peter
00:18:48
Teal off of the board of Facebook for
00:18:51
supporting Trump. So, you know, these
00:18:53
are like committed leftists. They're
00:18:56
Trump haters. But the point is that
00:18:58
these are people who fundamentally
00:18:59
believe in empowering government to the
00:19:03
maximum more government and empowering
00:19:05
government to the maximum extent. Now my
00:19:07
problem with that is I actually think
00:19:09
that probably the single greatest
00:19:11
dystopian risk associated with AI is the
00:19:15
risk that government uses it to control
00:19:18
all of us. To me like you end up in some
00:19:21
sort of Orwellian future where AI is
00:19:25
controlled by the government and out of
00:19:26
all the risks we've talked about that's
00:19:28
the only one for which I've seen
00:19:30
tangible evidence. So in other words, if
00:19:32
you go back to last year when we had the
00:19:35
whole woke AI, there was plenty of
00:19:37
evidence that the people who were
00:19:40
creating these products were infusing
00:19:42
their left-wing or woke values into the
00:19:45
product to the point where it was lying
00:19:47
to all of us and it was rewriting
00:19:49
history. And there was plenty of
00:19:51
evidence that the Biden EO was trying to
00:19:53
enshrine that idea. Was basically trying
00:19:55
to require DEI be infused into AI
00:19:58
models. and it wanted to anoint two or
00:20:02
three winners in this AI race. So, I'm
00:20:04
quite convinced that prior to Donald
00:20:07
Trump winning the election, we were on a
00:20:08
path of global compute governance where
00:20:10
two or three big AI companies are going
00:20:12
to be anointed as the winners. And the
00:20:14
quid proquo is that they were going to
00:20:15
infuse those AI models with woke values.
00:20:18
And there was plenty of evidence for
00:20:20
that. You look at the policies, you look
00:20:21
at the models. This was not a
00:20:23
theoretical concern. This was real. And
00:20:26
I think the only reason why we've moved
00:20:28
off of that trajectory is because of
00:20:30
Trump's election. But we could very
00:20:32
easily be moved back onto that
00:20:33
trajectory. If you were to look at all
00:20:35
three opinions here and and put them
00:20:37
together, they could all be true at the
00:20:38
same time. You've got a a number of
00:20:40
people, some might call useful idiots,
00:20:42
some might call just, you know, people
00:20:43
with god complexes who believe they know
00:20:45
how the world should operate. Effective
00:20:47
altruism kind of falls into that. Oh, we
00:20:49
can make a formula that that's their
00:20:51
kind of idea where we can tell you where
00:20:53
to put your money, rich people, in order
00:20:55
to create the most good and you know,
00:20:57
we're these enlightened individuals with
00:20:58
the best view of the world. They might
00:20:59
be, who knows, maybe they're the
00:21:00
smartest kids in the room, but they're
00:21:02
kind of delusional. The second piece
00:21:03
I'll do here is I think you're
00:21:05
absolutely correct, Chimat, that there
00:21:07
are people who have economic interests
00:21:09
who are then using those useful idiots
00:21:11
andor delusional people with god
00:21:14
complexes to serve their need, which is
00:21:16
to be one of the three winners. And then
00:21:18
sack
00:21:19
Inherent to all of that is they have a
00:21:21
political ideology. So why not use these
00:21:25
people with delusions of grandeur in
00:21:27
order to secure the bag for their
00:21:29
companies for their investments and
00:21:31
secure their candidates into office so
00:21:34
that they can block further people from
00:21:36
getting H100s cuz they literally want
00:21:38
to. By the way, that's the part that's
00:21:39
very smart about what they're doing
00:21:41
because you know it's not like they're
00:21:44
illquid. They're full of liquidity in
00:21:46
the sense that you're bringing in people
00:21:48
that are very technically capable and
00:21:50
you're setting up these funding rounds
00:21:52
where a large portion goes right back
00:21:54
out the door via secondaries and so
00:21:56
there's all these people that are making
00:21:58
money having this worldview and so to
00:22:00
your point Jason it's going to cement
00:22:02
that worldview and then they are going
00:22:04
to propagate it even more aggressively
00:22:06
into the world. So I think the threshold
00:22:08
question is should you fear government
00:22:11
overregulation or should you fear
00:22:13
autocomplete and I would say you should
00:22:16
not be so afraid of the autocomplete
00:22:18
right now it may get so good that it's
00:22:21
an AGI but right now it's an
00:22:23
exceptionally good autocomplete. Yeah.
00:22:25
And I just think that again it's a tried
00:22:27
and trueue tactic of people who want to
00:22:31
give immeasurably more power to the
00:22:33
government to try and make people afraid
00:22:35
and they stampede people into these
00:22:36
policies. Right. And it gives them
00:22:38
power. Exactly. Now, why do I think this
00:22:40
is important to talk about? On last
00:22:42
week's show, I talked about the trip to
00:22:44
the Middle East and how we started doing
00:22:46
these AI acceleration partnerships with
00:22:48
the Gulf States who have a lot of
00:22:49
resources, a lot of money, and they're
00:22:51
intensely interested in AI. and the
00:22:53
Biden administration was pushing them
00:22:54
away. It basically said, "You can't have
00:22:56
the chips. You can't build data
00:22:57
centers." And it was pushing them into
00:22:58
the arms of China. The thing that I
00:23:00
thought was so bizarre is that the
00:23:03
various groups and organizations and
00:23:05
former Biden staffers who wrote this
00:23:07
policy have been agitating in Washington
00:23:09
and they've been trying to portray
00:23:11
themselves as China hawks. And I'm like,
00:23:13
wait, this doesn't make any sense
00:23:15
because this policy again, there's
00:23:17
there's basically two camps in this new
00:23:18
cold war. It's US versus China. you can
00:23:20
pull the Gulf States into our orbit or
00:23:22
you can drive them into China's orbit.
00:23:24
So, this to me just didn't make any
00:23:25
sense. And what's happened is that
00:23:27
frankly, you've got this EA ideology
00:23:30
that's really motivating things, which
00:23:32
is a desire to lock down compute, right?
00:23:35
They're afraid of proliferation. They're
00:23:38
afraid of diffusion. That's really their
00:23:40
motivation. and they're trying to
00:23:42
rebrand themselves as China hawks
00:23:44
because they know that in the Trump
00:23:45
administration that idea is just not
00:23:47
going to get much purchase. Right. And
00:23:50
your position as ZAR is a level playing
00:23:54
field. People compete and the good guys,
00:23:58
you know, the West should be supported
00:24:00
to hit artificial general intelligence
00:24:03
as fast as possible. So the bad guys,
00:24:05
China, don't get it first. That that's a
00:24:08
well open competition. I don't know if I
00:24:11
would frame it around AGI specifically,
00:24:13
but what I would say is that look, I
00:24:14
think our policy should be to win the AI
00:24:16
race because the alternative is that
00:24:18
China wins it. And that would be very
00:24:20
bad for our economy and our military.
00:24:22
How do you win the AI race? You got to
00:24:23
out innovate. You got to have
00:24:24
innovation. That means we can't have
00:24:26
overregulation and red tape. We got to
00:24:28
build out the most AI infrastructure,
00:24:30
data centers, energy, which includes our
00:24:32
partners. And then third, I think it
00:24:34
means AI diplomacy because we want to
00:24:35
build out the biggest ecosystem. We know
00:24:37
that biggest app store wins, biggest
00:24:39
ecosystem wins, right? And the policies
00:24:41
under the Biden administration were
00:24:43
doing the opposite of all those things.
00:24:44
But again, you have to go back to what
00:24:46
was driving that. And it was not driven
00:24:48
by this China hawk mentality. That is
00:24:50
now a convenient rebranding. It was
00:24:53
driven by this EA ideology, this
00:24:56
doomerism. And so this is why I'm
00:24:58
talking about it is I want to expose it
00:25:00
because I think a lot of people on the
00:25:02
Republican side don't realize where the
00:25:04
ideology is really coming from and who's
00:25:06
funding it. They're obviously Trump
00:25:08
haters and they need to be lumored quite
00:25:10
frankly
00:25:14
when we look at they do they need to be
00:25:16
lumored. I mean, you know, Freberg, I
00:25:19
want I want to come back around again
00:25:20
cuz I respect your opinion on, you know,
00:25:24
how close we are to turning certain
00:25:26
corners, especially in science. So, I
00:25:29
understand big picture you believe that
00:25:32
the opportunity will be there. Hey, we
00:25:33
got people out of fields, you know, in
00:25:35
the agricultural revolution, we put them
00:25:37
into factories, industrial revolution,
00:25:38
then we went to this information
00:25:39
revolution. So, your position is we will
00:25:41
have a similar transition and it'll be
00:25:44
okay.
00:25:46
But do you not believe that the speed
00:25:50
because we've talked about this
00:25:51
privately and and publicly on the pod
00:25:53
that this speed the velocity at which
00:25:55
these changes are occurring you would
00:25:57
agree are faster than the industrial
00:25:59
revolution much faster than the
00:26:01
information revolution. So let's one
00:26:03
more time talk about job displacement
00:26:05
and I think the real concern here for a
00:26:08
group of people who are buying into this
00:26:09
ideology is specifically unions job
00:26:12
displacement. This is something the EU
00:26:14
cares about. This is something the Biden
00:26:16
administration cares about. If truck
00:26:18
drivers lose their jobs, just like we
00:26:20
went to bat previously for coal miners,
00:26:23
and there were only 75,000 or 150,000 in
00:26:25
the country at the time, but it became
00:26:27
the national dialogue. Oh my god, the
00:26:28
the the coal
00:26:29
miners. How fast is this going to
00:26:32
happen? One more time on drivers
00:26:34
specifically. Okay, coders, you think
00:26:36
there'll be more code to write, but
00:26:38
driving, there's not going to be more
00:26:39
driving to be done. So is this time
00:26:42
different in terms of the velocity of
00:26:44
the change and the job displacement in
00:26:46
your mind? Freedber the velocity is
00:26:49
greater but the benefit will be faster.
00:26:51
So the benefit of the industrial
00:26:53
revolution which ultimately drove lower
00:26:55
price products and broader availability
00:26:57
of products through manufacturing was
00:26:59
one of the key outputs of that
00:27:01
revolution. Meaning that we created a
00:27:03
consumer market that largely didn't
00:27:04
exist prior. Remember prior to the
00:27:07
industrial revolution, if you wanted to
00:27:08
buy a table or some clothes, they were
00:27:10
handmade. They were kind of artisal.
00:27:13
Suddenly, the industrial revolution
00:27:14
unlocked the ability to massproduce
00:27:16
things in factories. And that dropped
00:27:18
the cost and the availability and the
00:27:20
abundance of things that everyone wanted
00:27:22
to have access to, but they otherwise
00:27:24
wouldn't have been able to afford. So
00:27:26
suddenly everyone could go and buy
00:27:27
blankets and clothes and canned food and
00:27:31
all of these incredible things that
00:27:32
started to come out of this industrial
00:27:34
revolution that happened at the time.
00:27:36
And I think that folks are
00:27:37
underestimating and underrealizing the
00:27:39
benefits at this stage of what's going
00:27:41
to come out of the AI revolution and how
00:27:44
it's ultimately going to benefit
00:27:45
people's um availability of products,
00:27:48
cost of goods, access to things. So the
00:27:50
counterbalancing force Jcal is
00:27:52
deflationary which is um let's assume
00:27:55
that the cost of everything comes down
00:27:57
by half. That's a huge relief on
00:28:01
people's need to work 60 hours a week.
00:28:04
Suddenly you only need to work 30 hours
00:28:06
a week and you can have the same
00:28:09
lifestyle or perhaps even a better
00:28:10
lifestyle than you have today. So the
00:28:13
counterargument to your point, and I'll
00:28:15
talk about the pace of change in
00:28:16
specific jobs in a moment, but the
00:28:18
counterargument to your point is that
00:28:20
there's going to be this cost reduction
00:28:23
and abundance that doesn't exist today.
00:28:26
Give an example. Let's give like some
00:28:27
examples that we could see automation
00:28:30
and food prep. So we're seeing a lot of
00:28:31
restaurants install robotic systems to
00:28:34
make food and people are like, "Oh, job
00:28:36
loss, job loss." But let me just give
00:28:38
you the counter side. The counter side
00:28:39
is that the cost of your food drops in
00:28:41
half. So suddenly, you know, all the
00:28:43
labor cost that's built into making the
00:28:45
stuff you want to pick up, everyone's
00:28:46
freaking out right now about inflation.
00:28:47
Oh my god, it's $8 for a cup of coffee.
00:28:49
It's $8 for a latte. This is crazy,
00:28:51
crazy, crazy. What if that dropped down
00:28:52
to two bucks? You're going to be like,
00:28:54
man, this is pretty awesome with good
00:28:55
service and good experience. And don't
00:28:57
make it all dystopian, but suddenly
00:28:59
there's going to be this like incredible
00:29:01
reduction or deflationary effect in the
00:29:02
cost of food. And we're already starting
00:29:04
to see automation play it way in the
00:29:06
food system to bring inflation down. And
00:29:07
that's going to be very powerful for
00:29:09
people. Shout out to uh ita cloud
00:29:11
cushions and cafe X. We all took swings
00:29:13
at the bat at that exact concept is that
00:29:15
it could be done better, cheaper,
00:29:17
faster. One of the amazing things of
00:29:18
these vision action models that are now
00:29:21
being employed is you can rapidly learn
00:29:24
using vision systems and then deploy
00:29:26
automation systems in those sorts of
00:29:28
environments where you have a lot of
00:29:29
kind of repetitive tasks that the system
00:29:32
can be trained and installed in a matter
00:29:34
of weeks. And historically that would
00:29:35
have been a whole startup that it would
00:29:36
have taken years to figure out how to
00:29:38
get all these things together and custom
00:29:39
program it, custom code it. So the flip
00:29:41
side is like when Uber hit those people
00:29:43
were not drivers. Think about the jobs
00:29:45
that all those people had prior to Uber
00:29:47
coming to market. And then the reason
00:29:49
they drove for Uber is they could make
00:29:50
more money driving for Uber or now
00:29:52
driving or Door Dash and the
00:29:54
flexibility. So their lifestyle got
00:29:56
better. They had all of this more
00:29:57
control in their life. Their incomes
00:29:59
went up. And so there's a series of
00:30:01
things that you are correct won't make
00:30:04
sense in the future from a kind of
00:30:07
standard of work perspective. But the
00:30:08
right way to think about it is
00:30:10
opportunity gets created. New jobs
00:30:12
emerge, new industry, new income, costs
00:30:15
go down. And so I keep harping on this
00:30:17
that it's really hard today to be very
00:30:19
prescriptive to Sax's point about what
00:30:21
exactly is around the corner. But it is
00:30:24
an almost certainty that what is around
00:30:26
the corner is more capital will be
00:30:28
deployed. That means the economy grows.
00:30:30
That means there's a faster deployment
00:30:31
of growth of new jobs, new opportunities
00:30:34
for people to make more money, to be
00:30:35
happier in the work that they do. And
00:30:37
the flip side being things are going to
00:30:38
get cheaper. So, I mean, I know we're
00:30:40
waxing philosophical here, but I think
00:30:42
it's really key because you can focus on
00:30:44
the one side of the coin and miss the
00:30:46
whole other. And that's what a lot of
00:30:48
journalist commentators and fearongerers
00:30:51
do is they miss that other side. Got it.
00:30:53
Well said, Freeberg. Well said. I think
00:30:55
I've heard Satcha turn this question
00:30:58
around about job loss saying well do you
00:31:00
believe that GDP is going to grow by 10%
00:31:03
a year because what are we talking about
00:31:05
here I in order to have the kind of
00:31:07
disruption that you're talking about
00:31:09
where I don't know 10 to 20% of
00:31:12
knowledge workers end up losing their
00:31:13
jobs AI is going to have to be such a
00:31:15
profound force that it's going to have
00:31:17
to create GDP growth like we've never
00:31:19
seen before. That's right. So, it's
00:31:21
easier for people to say, "Oh, well, 20%
00:31:23
of people are going to lose their jobs."
00:31:24
But wait, are we we're talking about a
00:31:25
world in where the economy is growing
00:31:26
10% every year. Like, do do you actually
00:31:28
believe that's more income? That's more
00:31:30
income for everyone. That's new jobs
00:31:31
being created. It's an inevitability.
00:31:33
We've seen this in every revolution. You
00:31:36
know, prior to the industrial
00:31:37
revolution, 60% of Americans worked in
00:31:39
agriculture. And when the tractor came
00:31:41
around and factories came around, those
00:31:43
folks got to get out of doing manual
00:31:45
labor in the fields where they were
00:31:46
literally, you know, tilling the fields
00:31:49
by hand. and they got to go work in a
00:31:51
factory where they didn't have to do
00:31:52
manual labor to move things. Yeah, they
00:31:54
did things in the factory with their
00:31:55
hands, but it wasn't about grunt work in
00:31:58
the field all day in the sun. And it
00:32:00
became a better standard of living. It
00:32:01
became new jobs. And today, it became a
00:32:03
5day work week. It went from a 7day 7day
00:32:06
work week to five, 100 hours a week to
00:32:08
45, 50 hours a week. And now I think the
00:32:11
next phase is we're going to end up in
00:32:12
less than 30 hours a week with people
00:32:14
making more money and having more
00:32:16
abundance for every dollar that they
00:32:17
earn with respect to what they can
00:32:19
purchase and the lives they can live.
00:32:21
That means more time with your family,
00:32:22
more time with your friends, more time
00:32:24
to explore interesting opportunities.
00:32:26
So, you know, we've been through this
00:32:27
conversation a number of times. I I know
00:32:29
I'm not No, it's important to bring it
00:32:30
up, I think, and and really unpack it
00:32:33
because the fear is peing now, Sachs.
00:32:35
People are using this moment in time to
00:32:37
scare people that hey the jobs are going
00:32:39
to go away and they won't come back. But
00:32:41
what we're seeing on the ground saxs is
00:32:44
I'm seeing many more startups getting
00:32:46
created and able to accomplish more
00:32:48
tasks and hit a higher revenue per
00:32:50
employee than they did in the last two
00:32:52
cycles. So it used to be you know you
00:32:54
try to get to a quarter million in
00:32:55
revenue per employee than 500. Now we're
00:32:57
regularly seeing startups hit a million
00:32:58
dollars in revenue per employee,
00:33:00
something that was rarified air
00:33:01
previously, which then speaks to your
00:33:03
point, Freeberg, that there'll be more
00:33:05
abundance. There'll be more capital
00:33:07
generated, more more capital deployed,
00:33:10
with more capital deployed for more
00:33:12
opportunities, but you're going to need
00:33:13
to be more resilient. I think yeah, I
00:33:15
think it's actually very hard to
00:33:18
completely eliminate a human job. the
00:33:21
the ones that you cited and JK you keep
00:33:23
citing the same ones because I actually
00:33:24
don't think there's that many that fit
00:33:25
in this category the drivers and maybe
00:33:28
level one customer support because those
00:33:30
jobs are so monolithic but when you
00:33:32
think about even like what a salesperson
00:33:34
does right it's like yes they spend a
00:33:37
lot of time with prospects but they also
00:33:38
spend time negotiating contracts and
00:33:40
they spend time doing postale
00:33:42
implementation and follow-up and they
00:33:44
spend time learning the product and
00:33:46
giving feedback I mean it's a
00:33:48
multifaceted job and you can use AI I to
00:33:51
automate pieces of it, but to eliminate
00:33:53
the whole job is actually very hard. And
00:33:56
so I just think this idea that boom, 20%
00:33:58
of the workforce is going to be
00:33:59
unemployed in two years. I just don't
00:34:01
think that it's going to work that way.
00:34:03
But look, if there is widespread job
00:34:05
disruption, then obviously the
00:34:06
government's going to have to react and
00:34:08
we're going to be in a very different
00:34:09
societal order. But my point is, you
00:34:12
want the government to start reacting
00:34:14
now before this actually happens. We
00:34:16
don't need to be precogs and predict it.
00:34:17
Yeah. It's a total power grab. It's a
00:34:20
total power grab to give the government
00:34:22
and these organizations more power
00:34:24
before the risk is even manifested. And
00:34:27
let me say this as well with respect to
00:34:28
all these regulations that were created,
00:34:30
the 100page by NEO and the 200page
00:34:32
diffusion rule, none of these
00:34:34
regulations solve the excess problem.
00:34:37
None of these things actually would
00:34:39
prevent the most existential risk that
00:34:41
we're talking about. They don't solve
00:34:43
for alignment. They don't sign for the
00:34:45
kill switch. None of that. Yeah. If
00:34:46
someone actually figures out how to
00:34:47
solve that problem, I'm all ears. You
00:34:49
know, look, I'm not cavalier about these
00:34:52
risks. I understand that they exist, but
00:34:54
I'm not in favor of the fear-mongering.
00:34:57
I'm not in favor of giving all this
00:34:58
power to the government before you even
00:34:59
know how to solve these problems.
00:35:02
Shimath, you did a tweet about
00:35:04
entry-level jobs being toast. So, I
00:35:06
think there is a nuance here. Uh, and
00:35:08
both parties could be correct. I think
00:35:10
the job destruction is happening as we
00:35:12
speak. I'll just give what one example
00:35:14
and then drop to you, Chimath. One job
00:35:16
in startups that's not driving a car or
00:35:19
you know super entry level was people
00:35:20
would hire consultants to do recruitment
00:35:22
and to write job descriptions. Now I was
00:35:24
at a journal last night talking to a
00:35:26
bunch of founders here in Singapore and
00:35:27
I said how many people have used AI to
00:35:29
write a job description? Everybody's
00:35:30
hand went up. I said how many of you
00:35:32
with that job description was that job
00:35:34
description better than you could have
00:35:36
written or any consultant? And they they
00:35:37
all said yes 100% AI is better at that
00:35:39
job. That was a job a highlevel HR
00:35:42
recruitment job or an aspect of it sack.
00:35:45
So that was half the job, a third of the
00:35:47
job. To your point, the chores are being
00:35:49
automated. So I do think we're going to
00:35:50
see entrylevel jobs. Shimoth, the ones
00:35:53
that get people into an organization,
00:35:56
maybe they're going away. And that was
00:35:58
that your point of your tweet, which
00:35:59
we'll pull up right here. If a GPT is a
00:36:01
glorified
00:36:03
autocomplete, how did we used to do
00:36:05
glorified autocomplete in the past? It
00:36:08
was with new grads. New grads were our
00:36:12
autocomplete. And to your point, the
00:36:15
models are good enough that it
00:36:17
effectively allows a person to rise in
00:36:20
their career without the need of new
00:36:23
grad grist for the mill, so to speak.
00:36:27
So, I think the reason why companies
00:36:28
aren't hiring nearly as many new grads
00:36:30
is that the the folks that are already
00:36:32
in a company can do more work with these
00:36:35
tools. And and I think that that's a
00:36:36
very good thing. So you're generally
00:36:38
going to see OPEX as a percentage of
00:36:41
revenue shrink naturally and you're
00:36:44
going to generally see revenue per
00:36:47
employee go up naturally but it's going
00:36:51
to create a tough job market for new
00:36:53
grads in the established organizations.
00:36:56
And so what should new grads do? They
00:36:58
should probably steep themselves in the
00:36:59
tools and go to younger companies or
00:37:01
start a company. I think that's the only
00:37:03
solution for them. Bingo. The most
00:37:05
important thing for whether there are
00:37:08
jobs available for new grads or not is
00:37:10
whether the economy is booming. So
00:37:12
obviously in the wake of a financial
00:37:14
crisis, the jobs dry up because
00:37:16
everyone's cost cutting and those jobs
00:37:18
are the first ones to get
00:37:20
cut. But if the economy is booming, then
00:37:23
there's going to be a lot more job
00:37:25
creation. And so again, if AI is this
00:37:28
driver and enabler of tremendous
00:37:30
productivity, that's going to be good
00:37:31
for economic growth. And I think that
00:37:33
that will lead to more company
00:37:34
formation, more company expansion at the
00:37:36
same time that you're getting more
00:37:38
productivity. Now, to give an example,
00:37:40
one of the things I see a lot discussed
00:37:43
online about these coding assistants is
00:37:45
that they make junior programmers much
00:37:48
better because, you know, if you're
00:37:50
already like a 10x programmer, very
00:37:52
experienced, you already knew how to do
00:37:54
everything. And you could argue that the
00:37:56
people who benefit the most are the
00:37:59
entry-level coders who are willing to
00:38:01
now embrace the new technology and it
00:38:03
makes them much more productive. So in
00:38:05
other words, it's a huge leveler and it
00:38:08
takes an entry-level coder and makes
00:38:09
them 5x or 10x better. So look, this is
00:38:13
an argument I see online. The point is
00:38:15
just I don't think we know how this cuts
00:38:16
yet. I agree. And I just think there's
00:38:18
like this this dumerism is premature and
00:38:21
it's not a coincidence that it's being
00:38:24
funded and motivated by this hardcore
00:38:29
ideological element. I'll tell you my
00:38:30
hiring experience. We have about 30
00:38:33
people at 8090 and the way that I have
00:38:36
found it to work the best is you have
00:38:38
senior people act as mentors and then
00:38:40
you have an overwhelming corpus of young
00:38:42
very talented people who are AI native.
00:38:45
And if you don't find that mix, what you
00:38:48
have instead are L7s from Google and
00:38:52
Amazon and Meta who come to you with
00:38:55
extremely high salary demands and stock
00:38:58
demands and they just don't thrive. And
00:39:01
part of why they don't thrive is that
00:39:03
they push back on the tools and how you
00:39:05
use them. They push back on all these
00:39:07
things that the tools help you get to it
00:39:08
faster. M this is why I think it's so
00:39:10
important for the young folks to just
00:39:12
jump in with two feet and be AI native
00:39:14
from the jump because you're much more
00:39:16
hirable frankly to the to the emergent
00:39:19
company and the bigger companies you'll
00:39:22
have a lot of these folks that see the
00:39:24
writing on the wall may not want to
00:39:26
adapt as fast as otherwise. Another way
00:39:29
for example that you can measure this is
00:39:30
if you look inside your company on the
00:39:33
productivity lift of some of these
00:39:35
coding assistants for people as a
00:39:37
distribution of age. What you'll see is
00:39:39
the younger people leverage it way more
00:39:41
and have way more productivity than
00:39:43
older folks. And I'm not saying that as
00:39:44
an aegis comment. I'm saying that it's
00:39:46
an actual reflection of how people are
00:39:48
reacting to these tools. What you're
00:39:50
describing is a paradigm shift. It is a
00:39:52
big leap. Is you know it's like when I
00:39:54
went to college, when I took computer
00:39:56
science, it was object-oriented
00:39:58
programming. It was like C++. It was
00:40:00
compiled languages. It was gnarly. It
00:40:02
was nasty work. And then you had these
00:40:05
highle abstracted languages. And I used
00:40:07
to remember at Facebook, I would just
00:40:09
get so annoyed because I was like, why
00:40:10
is everybody using PHP and Python? This
00:40:12
is like not even real. But I was one of
00:40:14
these old lights who didn't understand
00:40:16
that I just had to take the leap. And
00:40:18
what it did was it grew the top of the
00:40:20
funnel of the number of developers by
00:40:22
10x. And as a result, what you had were
00:40:25
all of these advancements for the
00:40:26
internet. And I think what's happening
00:40:28
right now is akin to the same thing
00:40:30
where you're going to grow the number of
00:40:31
developers upstream by 10x. But in order
00:40:34
to embrace that, you just have to jump
00:40:36
in with two feet. And if you're very
00:40:37
rigid in how you think the job should be
00:40:40
done technically, I think you're just
00:40:41
going to get left behind. Just a little
00:40:44
interesting statistic there. Microsoft
00:40:46
announced 6,000 job layoffs, about 3% of
00:40:48
their workforce, while putting up record
00:40:50
profits while being in incredible cash
00:40:52
position. Total confirmation bias. It's
00:40:55
like now every time there's a layoff
00:40:57
announcement, people try to tie it to AI
00:40:58
to feed this doomer story. I don't think
00:41:01
that's an AI story. I well I actually
00:41:03
think it I don't think it's an AI story.
00:41:04
I think it is because the people they're
00:41:06
eliminating are management and I think
00:41:08
the the management layer becomes less
00:41:10
you're saying it was it was entry level
00:41:12
employees. Now you're saying it's
00:41:13
management. This is total confirmation.
00:41:14
I think those are no no I think those
00:41:16
are two areas that specifically get
00:41:18
eliminated. Entry level it's too hard.
00:41:20
It it's too hard to give them the grunt
00:41:22
work. And then for the managers who are
00:41:24
old and I've been there for 20 years.
00:41:26
Hold on. Let me finish. Th for those
00:41:27
people I think they are unnecessary in
00:41:31
this new AI management. What are you
00:41:33
talking about? What what is the AI agent
00:41:35
that's doing management right now in
00:41:37
companies? Oh theory doesn't even make
00:41:39
sense. Oh no it it totally does. There
00:41:41
are tools now that are telling you this
00:41:44
is these are the most productive people
00:41:45
in the organization. Shath just outlined
00:41:47
who's shipping the most etc. who's using
00:41:49
the tools. And then people are saying,
00:41:51
well, why do we have all these highly
00:41:53
priced people who are not actually
00:41:54
shipping code who are LSAs? You're
00:41:56
totally falling for some sort of
00:41:57
narrative here. This makes no sense. I
00:41:59
don't think I am. Yeah, let me be very
00:42:02
clear what I'm saying. What I am saying
00:42:04
is AI natives are extremely productive.
00:42:07
They use these tools. They're very fil
00:42:09
with them. I think it's very reductive,
00:42:12
but what you see is the older or more
00:42:15
established in your career you are in
00:42:16
technical roles, what I see is that it's
00:42:19
harder and harder for folks like that to
00:42:21
embrace these tools in the same way.
00:42:23
Now, how does it play out in terms
00:42:25
of jobs? I think that just these tools
00:42:29
are good enough where the net new
00:42:32
incremental taskoriented role that would
00:42:34
typically go to a new grad, a lot of
00:42:37
that can be defayed by these models.
00:42:38
That's what I'm saying very clear
00:42:40
specifically and I don't think that
00:42:41
speaks to management. I agree with Sax.
00:42:43
It doesn't do Sergey said Freeberg when
00:42:46
he came to uh our F1 that management
00:42:49
would be the first thing to go. I was
00:42:50
talking to some entrepreneurs last night
00:42:52
again here in Singapore and they are
00:42:54
taking all the GitHub and and Jira cards
00:42:57
and and things that have been submitted
00:42:58
plus all the Slack messages in their
00:43:00
organization and they're putting them
00:43:02
into an LLM and having it write
00:43:04
management reports of who is the most
00:43:06
productive in the organization. And in
00:43:07
the new version of Windows, it's
00:43:09
monitoring your entire desktop. Freeberg
00:43:11
management is going to know who in the
00:43:13
organization is actually doing work,
00:43:15
what work they're doing, and what the uh
00:43:18
result of that work is through AI. That
00:43:20
is the future of management. And you
00:43:22
take out all bias, all you know,
00:43:25
loyalty, and the AI is going to do that.
00:43:27
Couldn't disagree with you more saxs on
00:43:29
that, but Freeberg, you wanted to wrap
00:43:31
somewhere on this point. My point is
00:43:32
that AI managers are not losing their
00:43:35
job because AI is replacing them. I
00:43:37
didn't say that AI wouldn't be a
00:43:39
valuable tool for managers to use. Sure,
00:43:42
AI would be a great tool for managers,
00:43:44
but we're not anywhere near the point
00:43:47
where managerial jobs are being
00:43:48
eliminated because they're getting
00:43:50
replaced by AI agents. We're still at
00:43:52
the chatbot stage of this. Literally,
00:43:54
Sergey said he took their internal
00:43:56
Slack, went into like a dev
00:43:58
conversation, and said, "Who are the
00:43:59
underrated people in this organization
00:44:00
who deserve a raise?" and it gave him
00:44:02
the right answer. So, wait, that doesn't
00:44:03
allow you to cut 6,000 people. I think
00:44:05
it's happening as we speak. It's just
00:44:07
not over. You You fell for this
00:44:09
narrative. You grasped onto this
00:44:11
Microsoft restructuring where they
00:44:14
eliminated 6,000 roles and you're trying
00:44:15
to attribute that to AI now. I think it
00:44:18
has to do with AI. I think management is
00:44:19
looking at it saying, "We are going to
00:44:21
replace these positions with AI. We
00:44:23
might as well get rid of them now." It
00:44:24
is in flux. We'll see who's right in the
00:44:26
coming months. Can I make another
00:44:28
comment, Freick? wrap this up here so we
00:44:29
can get on to the next topic. This is a
00:44:32
great topic. This is I want to make one
00:44:34
last point which I think and Sax you may
00:44:37
not appreciate this so we can have a a
00:44:39
healthy argument about this. I think in
00:44:41
the same way that all of this jobs are
00:44:44
going to get lost to AI fear-mongering.
00:44:46
There's a similar narrative that I think
00:44:48
is a false narrative around there's a
00:44:50
race in AI that's underway between
00:44:53
nation states. And the reason I think
00:44:56
it's false is if I asked you guys the
00:44:57
question, who won the industrial
00:44:59
revolution? The industrial revolution
00:45:01
benefited everyone around the world.
00:45:03
There are factories and there's a
00:45:04
continuous effort and continuous
00:45:06
improvements in manufacturing processes
00:45:08
worldwide. That is a continuation of
00:45:10
that revolution. Similar if I asked who
00:45:12
won the internet race, there were
00:45:14
businesses built out of the US,
00:45:16
businesses built out of China,
00:45:17
businesses built out of India and Europe
00:45:19
that have all created value for
00:45:21
shareholders, created value for
00:45:22
consumers, changed the world, etc. And I
00:45:25
think the same is going to happen in AI.
00:45:26
I don't think that there's a finish line
00:45:28
in AI. I think AI is a new paradigm of
00:45:32
work, a new paradigm of productivity, a
00:45:34
new paradigm of business, of the
00:45:36
economy, of livelihoods, of pretty much
00:45:39
everything uh every interaction humans
00:45:41
have with ourselves and the world around
00:45:43
us will have in its substrate AI and as
00:45:47
a result, I think it's going to be this
00:45:48
continuous process of improvement. So,
00:45:49
I'm not sure. Look, there there are
00:45:51
different models and you can look at the
00:45:52
performance metrics of models, but you
00:45:54
can get yourself spun up into a tizzy
00:45:56
over which model is ahead of the others.
00:45:58
Which one's going to quote get to the
00:45:59
finish line first? But I think at the
00:46:01
end of the day, the abundance and the
00:46:03
economic prosperity that will arise from
00:46:05
the continuous performance improvements
00:46:07
that come out of AI and AI development
00:46:09
will benefit all nation states and
00:46:11
actually could lead to a little bit more
00:46:13
of a less resource constrained world
00:46:16
where we're all fighting over limited
00:46:17
resources and there's nation state
00:46:19
definitions around who has access to
00:46:21
what and perhaps more abundance which
00:46:23
means more peace and uh less of this
00:46:25
kind of resourced world. your thought on
00:46:28
the kumbaya theory exposed by Freeberg.
00:46:31
Yeah, exactly. Um I I'll partially agree
00:46:34
in the sense that I don't think the AI
00:46:37
race is a finite game. It's an infinite
00:46:39
game. I I agree that there's no finish
00:46:42
line, but that doesn't mean there's not
00:46:44
a race going on. So for example, an arms
00:46:46
race would be a classic example of a
00:46:49
competition between countries to see who
00:46:53
is stronger to basically amass power and
00:46:55
they might be neutralizing each other.
00:46:57
The balance of power may stay in
00:46:58
equilibrium even though both sides feel
00:47:00
the need to constantly uplevel their
00:47:03
arms, their power. Yeah. And so I think
00:47:05
that to use the the term that Mir
00:47:08
Shimemer used at the all-in summit, we
00:47:09
are in an iron cage. The US and China
00:47:12
are the two leading countries in the
00:47:14
world economically, militarily,
00:47:17
technologically. They both care about
00:47:20
their survival. The best way to ensure
00:47:22
your survival in a self-help world is by
00:47:27
being the most powerful. And so these
00:47:29
are great powers who care a lot about
00:47:31
the balance of power. And they will
00:47:32
compete vigorously with each other to
00:47:35
maintain the greatest balance of power
00:47:37
between them. and high-tech is a major
00:47:40
dimension of that competition and within
00:47:42
high-tech AI is the most important
00:47:43
field. So look, there is going to be an
00:47:45
intense competition around AI. Now the
00:47:48
question is how does that end up? I mean
00:47:50
it could end up in a tie or in it could
00:47:53
end up in a situation where both
00:47:55
countries benefit. Maybe open source
00:47:57
wins. Maybe neither side gains a
00:47:59
decisive advantage. that they're
00:48:01
absolutely going to compete because
00:48:03
neither one can afford to take the risk
00:48:05
that the other one will develop a
00:48:07
decisive advantage. Prisoners dilemma.
00:48:09
Nuclear proliferation is a good analogy.
00:48:11
I would argue nuclear deterrence led to
00:48:13
a more peaceful world in the 20th
00:48:15
century. I mean is that fair to say Sax
00:48:17
that ultimately what happened with
00:48:18
nuclear is that the actual underlying
00:48:21
technology hit you know an asmtote right
00:48:24
it plateaued right and so we ended up in
00:48:26
a situation where in the case of the
00:48:28
United States versus Soviet Union where
00:48:30
both sides had enough nukes to blow up
00:48:32
the world many times over and there
00:48:34
wasn't really that much more to innovate
00:48:36
so you know the the the underlying
00:48:38
technological competition had ended the
00:48:40
the dynamic was more stable and they
00:48:42
were able to reach an arms control
00:48:44
framework to sort of control the arms
00:48:46
race, right? I think AI is a little
00:48:49
different. We're in a situation right
00:48:51
now where the technology is changing
00:48:53
very very rapidly and it's potentially
00:48:55
on some sort of exponential curve and so
00:48:58
therefore being a year ahead even 6
00:49:00
months ahead could result in a major
00:49:02
advantage. I think under those
00:49:03
conditions both sides are going to feel
00:49:05
the need to compete very vigorously. I
00:49:08
don't think they can sign up. This is a
00:49:09
system of productivity right for an
00:49:11
agreement to slow each other down. I
00:49:12
just don't nuclear was not a system of
00:49:14
productivity. It was not a system of
00:49:15
economic growth. It was a system of
00:49:17
literally destruction. And this is quite
00:49:20
different. This is a system of making
00:49:21
more with less which unleashes benefits
00:49:24
to everyone in a way that perhaps should
00:49:27
be calming down the conflict in the
00:49:29
potential. You got to admit that there's
00:49:30
a there is a potential dual use here.
00:49:33
There's no question that the armies of
00:49:35
the future are going to be drones and
00:49:36
robots and they're going to be AI
00:49:37
powered. Yeah. And as long as that's the
00:49:39
case, these countries are going to
00:49:40
compete vigorously to have the best AI
00:49:42
and they're going to want their leaders
00:49:45
or national champions or startups and so
00:49:47
forth to win the race. What's the worst
00:49:49
case, Saxs, if if China wins the AI
00:49:52
race? What is the worst case scenario?
00:49:54
Ask what it means first. Ask Sax. That's
00:49:56
literally what I'm asking. Like what
00:49:58
would that scenario be? Would they
00:49:59
invade America and they dominate us
00:50:01
forever? What does it mean to citizen?
00:50:03
Yeah. What does it mean to win? Yeah. To
00:50:06
me, it would mean that they achieve a
00:50:08
decisive advantage in AI such that we
00:50:11
can't leaprog them back. And an example
00:50:14
of this might be something like 5G where
00:50:17
Huawei somehow leaprogged us, got to 5G
00:50:20
first and disseminated it through the
00:50:22
world. They weren't concerned about
00:50:24
diffusion. They were interested in
00:50:26
promulgating their technology throughout
00:50:28
the world. If the Chinese win AI, they
00:50:30
will sell more products and services
00:50:32
around the globe than the US. This is
00:50:34
where we have to change our mindset
00:50:35
towards diffusion. I would define
00:50:37
winning as the whole world consolidates
00:50:40
around the American tech stack. They use
00:50:42
American hardware in data centers that
00:50:45
again are are fundamentally powered by
00:50:47
American technology. And you know just
00:50:49
look at market share. Okay? If we have
00:50:51
like 80 to 90% market share that's
00:50:53
winning. If they have 80% market share
00:50:55
then we're in big trouble. So it's very
00:50:58
simple. It means like yeah but if the
00:50:59
market grows up by 10x it doesn't matter
00:51:02
because the world will have every
00:51:03
individual in every country will now
00:51:05
have more they will have a more
00:51:06
prosperous life and as a result it's not
00:51:09
necessarily the framing about if we
00:51:11
don't get there first we are necessarily
00:51:13
going to lose I get that there's an edge
00:51:14
case of conflict or what have you but I
00:51:17
do think that there's a net benefit
00:51:19
where the whole world suddenly is in
00:51:20
this more prosperous state and this is a
00:51:23
classic example of a dual use technology
00:51:25
where there are both economic benefits
00:51:27
and there are military benefits. Yes,
00:51:30
GPS would come to mind in this example,
00:51:33
right? Like my summary point is just
00:51:35
that it's not all about a losing game
00:51:37
with respect to this quote race with
00:51:40
other nation states. But at the end of
00:51:41
the day, yes, there is risk, but I do
00:51:44
think that if the the pace of
00:51:46
improvement stays on track like it is
00:51:48
right now, holy [Â __Â ] I think we're in a
00:51:50
pretty good place. That's just my point.
00:51:51
Okay. Some positivity. Okay. Look, I I
00:51:54
hope that the AI race stays entirely
00:51:56
positive and it's a healthy competition
00:51:58
between nations and the competition
00:52:00
spurs them on to develop more prosperity
00:52:02
for their citizens. But as we talked
00:52:04
about in the AI summit, there's two ways
00:52:06
of looking at the world. There's kind of
00:52:08
the economist way that Jeffrey Saxs was
00:52:10
talking about and then there's the
00:52:12
balance of power way or realist way
00:52:13
which Mir Shmer was talking about. And
00:52:16
when economic prosperity and survival or
00:52:21
balance of power come into conflict,
00:52:24
it's the realest view of the world that
00:52:26
it's the balance of power that gets
00:52:27
privileged. And I just think that's the
00:52:29
way that governments operate is that
00:52:31
prosperity is incredibly important. We
00:52:33
want economic success, but power is
00:52:37
ultimately privileged over that. And
00:52:39
this is why we're going to compete
00:52:40
vigorously in high-tech. That's why
00:52:42
there is going to be an AI race. Okay,
00:52:44
perfect segue. We should talk a little
00:52:46
bit about what was the topic of
00:52:48
discussion yesterday. I had a lunch with
00:52:51
a bunch of family offices and capital
00:52:52
allocators uh government folks here in
00:52:54
Singapore and they were talking about
00:52:56
our discussion last week about the big
00:52:59
beautiful bill and the debt here in the
00:53:02
United States. It's permeating
00:53:04
everywhere. The two conversations at
00:53:06
every stop I've made here is the big
00:53:09
beautiful bill and the balance sheet of
00:53:12
the United States as well as tariffs.
00:53:13
So, we need to maybe revisit our
00:53:16
discussion last week. Chimath, you had
00:53:18
uh and Freeberg did a an impromptu call
00:53:21
with Ron Johnson over the weekend, which
00:53:23
then spurred him going on 20 other
00:53:26
podcasts to talk about this. Steven
00:53:28
Miller from the administration has been
00:53:30
tweeting some corrections or his
00:53:33
perceived corrections about the bill.
00:53:35
And Sax, uh, I think you've also started
00:53:37
tweeting this. Where do we want to
00:53:39
start? Maybe. Well, I think there are
00:53:41
just a couple of facts that should be
00:53:43
cleaned up because Okay, so facts from
00:53:46
the administration, their view of our
00:53:48
discussion. Well, even though I was
00:53:50
defending the bill last week on the
00:53:52
whole, I wasn't saying it was perfect. I
00:53:53
was just saying it was better than the
00:53:54
status quo. Yeah, you were clear about
00:53:56
that. Yeah. Yeah. But even even I in
00:53:58
doing that was conceding some points
00:54:01
that I think were just factually wrong.
00:54:03
And the big one was that I said I was
00:54:05
disappointed that Doge the Doge cuts
00:54:07
weren't included in the big beautiful
00:54:08
bill. What Steven Miller has pointed out
00:54:11
is that reconciliation bills can only
00:54:14
deal with what's called mandatory
00:54:15
spending. They can't deal with what's
00:54:16
called discretionary spending. And since
00:54:19
the Doge cuts apply to discretionary
00:54:21
spending, they just can't be dealt with
00:54:23
in a reconciliation bill. They have to
00:54:24
be dealt with separately. There can be a
00:54:26
separate recision bill that comes up,
00:54:28
but it can't be dealt with in this bill.
00:54:30
And just to be very clear, look, if the
00:54:32
Doge cuts don't happen through recision,
00:54:34
I'm going to be very disappointed in
00:54:36
that. I really want the Doge cuts to
00:54:37
happen, but it's just a fact that the
00:54:40
Doge cuts cannot happen in the big
00:54:42
beautiful bill. It's not that kind of
00:54:44
bill. And I think it's therefore wrong
00:54:46
to blame big beautiful bill for not
00:54:48
containing Doge cuts when the Senate
00:54:51
rules don't allow that. You know, it all
00:54:52
goes back to the the Bird rules. There
00:54:55
are only specific things that can be
00:54:57
dealt with through reconciliation, which
00:54:59
is this 50 vote threshold, and it has to
00:55:02
be quote unquote mandatory spending.
00:55:04
discretionary cuts are dealt with in
00:55:05
annual appropriations bills that require
00:55:07
60 votes. Now look, this is kind of a
00:55:09
crazy system. I don't know exactly how
00:55:11
it evolved. I guess Robert Bird is the
00:55:13
one who came up with all this stuff and
00:55:15
maybe they need to change the system,
00:55:16
but it's just wrong to blame the big
00:55:18
beautiful bill for not containing the
00:55:19
Doge cuts. That's just a fact. Okay, so
00:55:21
the other thing is that the BBB does
00:55:24
actually cut spending. It's just not
00:55:26
scored that way because when the
00:55:29
bill removes the sunset provision from
00:55:31
the 2017 tax cuts, the CBO ends up
00:55:35
scoring that as effectively a spending
00:55:37
increase. But tax rates are simply
00:55:40
continuing at their current level. In
00:55:42
other words, at this year's level. So if
00:55:45
you used the current year as your
00:55:47
baseline, okay, and then compared it to
00:55:50
spending next year, it would score as a
00:55:54
cut in spending. So it's just not it's
00:55:56
not correct to say this bill increases
00:55:58
spending. It does actually result in a
00:56:01
mandatory spending cut, but it's not
00:56:04
getting credit for that because we're
00:56:06
continuing the tax rates at the current
00:56:09
year's rates. Do you believe Sachs that
00:56:12
this administration which you are part
00:56:14
of in four years will have spent will
00:56:17
have balanced the budget. Will it have
00:56:19
reduced the deficit or will the deficit
00:56:21
continue to grow at 2 trillion a year?
00:56:24
What is your belief because there's a
00:56:26
lot of strategies going on here. Yeah.
00:56:27
My my belief is that President Trump
00:56:29
came into office inheriting a terrible
00:56:32
fiscal situation. I mean basically that
00:56:34
he created and that Biden created. They
00:56:36
both put 8 trillion. They both put 8
00:56:38
trillion on the debt. It's a big
00:56:40
difference. It's a big difference to add
00:56:41
to the deficit when you're in the
00:56:43
emergency phase of
00:56:45
CO for that. Sure. It's emergency
00:56:48
spending. It was never supposed to be
00:56:49
permanent and then somehow Biden made it
00:56:51
permanent and he wanted a lot more.
00:56:52
Remember build back better? He wanted a
00:56:54
lot more. So, you know, it's it's tough
00:56:57
when you come into office with a what is
00:56:59
$2 trillion annual deficit. So, to my
00:57:01
original question, now look, hold on.
00:57:03
Would I like to see the deficit
00:57:04
eliminated in one year? Yeah,
00:57:06
absolutely. But there's just not the
00:57:07
votes for that. Well, I asked you for
00:57:09
there's a one vote margin here in the
00:57:10
House and the Democrats aren't
00:57:13
cooperating in any way. So, I think that
00:57:15
the administration is getting the most
00:57:18
done that it can. This is a mandatory
00:57:20
spending cut and I think the Doge cuts
00:57:23
will be dealt with hopefully through
00:57:24
recision in a subsequent bill. I'm
00:57:27
asking you about four years from now.
00:57:28
Will we be sitting here in four years?
00:57:30
Will Trump have cut spending by the end
00:57:32
of this term? In another three and a
00:57:34
half years, will we be looking at a
00:57:36
balanced budget? potentially is that the
00:57:38
goal of the administration or will we be
00:57:41
at 42 44 45 trillion at the end of
00:57:45
Trump's second term? David said, listen,
00:57:47
if you want that level of specificity,
00:57:48
you're going to have to get Scott Besson
00:57:49
on. Okay, this is just not my area. I'm
00:57:51
not going to pretend to have that level
00:57:52
of detailed answers. But what I believe
00:57:54
is that the Trump administration's
00:57:56
policy is to spur growth. I think that
00:58:00
these tax policies will spur growth. I
00:58:02
think that AI will also be a huge
00:58:04
tailwind. It'll be a productivity boost.
00:58:07
I think let's stop being doomers about
00:58:08
it. We need that productivity boost and
00:58:10
I think that the net result of those
00:58:12
things will be to improve the fiscal
00:58:14
situation. Do I want more spending cuts?
00:58:16
Yeah, but look, we're getting more than
00:58:18
was represented last week. Let's put it
00:58:20
that way. Okay, fair enough. Sax, thank
00:58:21
you for the cleanup there. Chimath, our
00:58:23
bestie Elon was on the Sunday shows and
00:58:26
he said, "Hey, the bill can be big or it
00:58:28
can be beautiful. It can't be both." He
00:58:30
seems to be, I'll say, displeased or
00:58:34
maybe not as optimistic about balancing
00:58:36
the budget and and getting spending
00:58:38
under
00:58:39
control, but he he still believes in
00:58:41
Doge, obviously, and and and hopefully
00:58:43
Doge continues. You seemed a little bit
00:58:45
concerned last week. A week's passed.
00:58:49
You've heard some of Steven Miller's
00:58:51
opinions. Where do where do you net out
00:58:53
seven days from our big beautiful budget
00:58:55
bill debate last week, a week later?
00:58:58
Well, I mean, I think Steven's critique
00:59:00
of
00:59:01
how the media summarized the reaction to
00:59:06
the bill is accurate.
00:59:09
And I think it's probably useful to
00:59:11
double click into one thing that Saxs
00:59:13
didn't mention, but that Steven did. A
00:59:15
lot of this pivots around the CBO, which
00:59:17
is the Congressional Budget Office, and
00:59:19
how they look at these bills, and
00:59:22
there's a lot of issues with how they do
00:59:26
it.
00:59:28
In one specific case, which Sax just
00:59:30
mentioned and Stephen talked about, is
00:59:31
that they have these arcane rules about
00:59:35
the way that they score things. And what
00:59:38
they were assuming is that the tax rates
00:59:43
would flip
00:59:44
back to what they were before the first
00:59:48
Trump tax cuts, which obviously would be
00:59:51
higher than where they are today. What
00:59:55
that would mean in their financial model
00:59:58
is we were going to get all that money
01:00:00
now to maintain the tax cuts where we
01:00:03
are. They now then would look at that
01:00:05
and say, "Oh, hold on. That's a loss of
01:00:07
revenue. Why are all of these things
01:00:09
important?"
01:00:11
I downloaded the CBO model, went through
01:00:14
it, and what I would say is at best it's
01:00:17
Spartan, which means that I don't think
01:00:20
a financial analyst or somebody that
01:00:23
controls a lot of money will actually
01:00:25
put a lot of stock in their model. I
01:00:28
think what you'll have happen is people
01:00:29
will build their own versions bottoms
01:00:32
up. Do you trust it, the the CBO's
01:00:35
version of this, or do you largely trust
01:00:36
it? I don't think the CBO really knows
01:00:38
what's going on to be totally honest
01:00:39
with you. Okay. I think that there are
01:00:42
parts of what they do which they're also
01:00:45
opaque on. Nick, I sent you a tweet from
01:00:50
Goldman Sachs. So, here is what Goldman
01:00:52
put out. Now, the point is when you
01:00:54
build a model, what you're trying to do
01:00:56
is net out all of these bars, okay?
01:00:58
You're trying to add the positive bars
01:00:59
and the negative bars, and you figure
01:01:00
out what is the total number at the end
01:01:02
of it. Now, in order to do that, when
01:01:05
you see the bars on the far right,
01:01:07
that's a 20 $34. That's very different
01:01:09
than a 20 $25. The CBO doesn't disclose
01:01:13
how they deal with that. They don't dis
01:01:15
disclose the discount rate. So you can
01:01:17
question what that is. The CBO makes
01:01:19
these assumptions that, you know, as
01:01:21
Steven pointed out, are very brittle
01:01:22
with respect to the tax plan. That's not
01:01:25
factored in here. So those are the
01:01:27
issues with the way the CBO scores it.
01:01:30
So you have to do it yourself. Now,
01:01:32
Peter Navaro published an article which
01:01:34
I think is probably the most pivotal
01:01:36
article about this whole topic.
01:01:39
Peter of tariff fame. Yeah. Yeah. Here I
01:01:42
think he nails it right in the bullseye,
01:01:44
which is the bond market needs to make a
01:01:48
decision on one very critical assumption
01:01:51
when they build their own model. Okay,
01:01:53
so let's ignore the
01:01:55
CBO's kind of brittle math and the Excel
01:01:58
that they post on their website. People
01:02:00
are going to do their own because
01:02:01
they're talking about managing their own
01:02:02
money. But Navaro basically points to
01:02:05
the critical thing which is listen those
01:02:07
CBO assumptions also include a fatal
01:02:09
error which is they assume these very
01:02:12
low levels of
01:02:13
GDP. What you're probably going to see
01:02:16
in Q2 is a really hot GDP print. If I'm
01:02:20
a betting man, which I am, I think the
01:02:21
GDP prints going to come in above three.
01:02:24
Not quite four, but above three. And so
01:02:26
what Peter is saying here is, hey guys,
01:02:28
like you're estimating 1.7% GDP. why
01:02:32
don't you assume 2.2 two or why don't
01:02:34
you assume 2.7 or any number or really
01:02:37
what he's saying is why don't you build
01:02:39
a sensitivity so that you can see the
01:02:41
implications of that and I think that
01:02:43
that is a very important point okay so
01:02:46
where do I net out a week later Jason
01:02:48
it's pretty much summarized in the tweet
01:02:52
that I posted earlier today so over the
01:02:56
last week as people have digested it I
01:02:58
think that there are small actors in
01:03:00
this play and big actors the biggest
01:03:02
actor is obviously President President
01:03:03
Trump. But the second biggest actor is
01:03:05
the long end of the bond market. These
01:03:07
are the central bankers, the long bond
01:03:09
holders, and these macro hedge funds.
01:03:11
Why? Because they will ultimately
01:03:13
determine the United States's cost of
01:03:16
capital. How expensive will it be to
01:03:18
finance our deficits irrespective of
01:03:20
whatever the number is. It could be a
01:03:22
dollar or it could be a trillion
01:03:24
dollars. That doesn't matter right now.
01:03:25
The point is what is going to be our
01:03:26
cost of capital? And what's happened
01:03:30
over the last little while is that
01:03:32
they've steepened the curve and they've
01:03:33
made it more expensive for us to borrow
01:03:36
money. That's just the fact. So how do
01:03:40
we get in front of this? I think the
01:03:43
most important thing if you think about
01:03:44
what Peter Navaro said is this plan and
01:03:48
the bill can work if we get the GDP
01:03:52
right. Okay. So how do you get the GDP
01:03:56
right? And this is where I have one very
01:03:59
narrow set of things that I think we
01:04:01
need to improve. And the specific thing
01:04:03
that I'll go back to is today
01:04:05
America is at a supply demand tradeoff
01:04:08
on the energy side. What does that mean?
01:04:11
We literally consume every single bit of
01:04:14
energy that we make. We don't have slack
01:04:18
in the
01:04:19
system. We are growing our energy
01:04:22
demands on average about 3% a year.
01:04:27
So I think the most critical thing we
01:04:29
need to do is to make sure the energy
01:04:31
markets stay robust. Meaning there's a
01:04:35
lot of investment that people are
01:04:37
making. On Tuesday I announced a deal
01:04:40
that I did building a 1 gawatt data
01:04:42
center in Arizona. This is a lot of
01:04:44
money. This is little old me. But there
01:04:46
are lots of people ripping in huge huge
01:04:48
huge checks, hundreds of billions of
01:04:49
dollars. I think the sole focus has to
01:04:52
be to make sure that the energy policy
01:04:54
of America is robust and it keeps all
01:04:57
the electrons online. If there's any
01:05:00
contraction, I think it'll hit the GDP
01:05:03
number because we won't have the energy
01:05:05
we need and that's where things start to
01:05:07
get a little funky. So, I think where I
01:05:08
am is I think President Trump should get
01:05:11
what he wants. I think the bill can work
01:05:14
narrowly address the energy provisions
01:05:16
and I think we live to fight another
01:05:18
day. So
01:05:19
Freiriededberg cynical approach might be
01:05:21
we're working the refs here. The CBO is
01:05:24
not taking into GDP. This GDP has a
01:05:27
magical unicorn in it. AI and energy is
01:05:30
going to spur this amazing growth. But
01:05:33
the bond markets don't believe it
01:05:34
either. So, are we looking at just a
01:05:39
GOP, a party, I'll put the
01:05:41
administration aside, that is just as
01:05:43
recklessly spending as the Democrats,
01:05:45
and they want to change the formula by
01:05:48
which they're judged in the future, that
01:05:50
there's going to be magically all this
01:05:51
growth and growth solves all problems.
01:05:54
And what we really need to do to your
01:05:56
point I think two weeks ago that this is
01:05:59
just disgraceful to put up this much
01:06:00
spending and we have to have austerity
01:06:02
and we need to increase uh maybe the
01:06:05
discipline in the country and both
01:06:06
parties have to be part of that. I'm
01:06:08
asking you uh from the cynical
01:06:10
perspective maybe to represent or steal
01:06:11
me on the other side here.
01:06:14
We had a conversation with Senator Ron
01:06:16
Johnson after we recorded the pod last
01:06:19
week and he was very clear in a key
01:06:23
point which is that this bill addresses
01:06:27
mandatory spending. Just to give you a
01:06:29
sense 70% of our federal budget is
01:06:32
mandatory spending. 30% falls into that
01:06:35
discretionary category. The mandatory
01:06:37
spending is composed of the interest on
01:06:39
the debt which is now well over a
01:06:41
trillion dollars a year on its way to a
01:06:43
trillion five almost a trillion a year.
01:06:44
Medicare, Medicaid, Social Security and
01:06:48
some other income security programs. And
01:06:50
as Ron Johnson shared with us over the
01:06:53
years more and more programs have been
01:06:55
put into the mandatory spending category
01:06:57
and so you can get past the
01:06:59
filibustering in the Senate to be able
01:07:00
to get budget adjustments done. The key
01:07:04
thing he's focused on and Rand Paul is
01:07:06
focused on and I've talked about is the
01:07:08
spending level of our mandatory
01:07:10
programs. The big beautiful bill
01:07:13
proposes a roughly $70 billion per year
01:07:16
cut in Medicaid. Okay, and that sounds
01:07:19
awful. How could you do that to people?
01:07:21
In 2019, the year before COVID, Medicaid
01:07:25
spending was $627 billion. 2024 it was
01:07:29
914 billion. So the $70 billion cut gets
01:07:33
you down to 840. You're still roughly
01:07:35
call it 40% above where you were in
01:07:37
2019. So is that the right level? And
01:07:40
fundamentally the opportunity to cut
01:07:43
those mandatory programs, which I know
01:07:45
sounds awful, to cut Social Security and
01:07:46
cut Medicaid, but the reality is they're
01:07:48
not just being cut from a low level.
01:07:52
They're being cut from a level that's 60
01:07:55
plus% higher than they were in 2019. And
01:07:57
I gave you another example which is the
01:07:59
SNAP program, the food stamp program.
01:08:01
Again, uh $15 billion of the 120 a year
01:08:05
that we spend on food stamps is being
01:08:06
used to buy soda and a whole another
01:08:09
chunk of that 120 is being used to buy
01:08:11
other junk food. So that they have
01:08:13
proposed in this bill to cut SNAP down
01:08:15
to 90 and it was 60 in 2019. So it's
01:08:19
still 50% above where it was in 2019. So
01:08:23
the key point that's being made by Ron
01:08:24
Johnson and others is that the spending
01:08:27
on these mandatory programs which
01:08:29
account for nearly threequarters of our
01:08:30
federal budget are still very elevated
01:08:33
relative to where we were in 2019. And
01:08:36
we are not going to get out of our
01:08:38
deficit barring a massive increase in
01:08:39
GDP without changes to the spending
01:08:42
level. Now I don't put the blame on the
01:08:45
White House. This bill passed with one
01:08:47
vote in the House. One vote. And so a
01:08:50
key point to note, and I've said this
01:08:52
from day one, and every time I've gone
01:08:54
to DC and every time we've talked about
01:08:56
Doge, I've said there's no way any of
01:08:58
this stuff's going to change without
01:08:59
legislative action from the Congress.
01:09:02
And here we are seeing Congress, for
01:09:04
whatever reason, you can listen to Ron
01:09:05
Johnson. You can listen to Rand Paul.
01:09:06
You can listen to others say, you know
01:09:08
what, we can't cut that deep. It is
01:09:10
going to be too harmful to our
01:09:12
constituents. We need to keep the
01:09:14
programs at their current levels or make
01:09:17
no changes at all or only modest
01:09:18
changes. And that's where we are. That's
01:09:20
the reality. Now, I do think that Navaro
01:09:23
did an excellent job in his op-ed for
01:09:26
whatever criticism we may want to lay on
01:09:27
Navaro for many other things. He pointed
01:09:29
out that the CBO projections in 2017 for
01:09:34
the next year's GDP growth numbers was
01:09:36
1.8 to 2% and it actually came in at
01:09:38
2.9%. a full one point higher because of
01:09:41
the tax and jobs act that was passed by
01:09:43
the Trump administration in 2017. So the
01:09:46
additional money that goes into
01:09:48
investments because lower taxes are
01:09:50
being paid fueled GDP growth. This is
01:09:52
what some people call trickle down
01:09:54
economics. People ridicule it. They say
01:09:56
it doesn't work. It's not real. But in
01:09:58
this particular instance, they cut taxes
01:10:00
and the GDP grew much faster than was
01:10:03
projected or estimated by the economists
01:10:05
at the CBO. So the argument that's being
01:10:07
made is that we are not capturing many
01:10:10
of the upsides in the GDP numbers that
01:10:13
are being projected. And I will be
01:10:15
honest about this. I don't think anyone
01:10:18
knows how much the GDP is going to grow.
01:10:21
We don't know the economic benefit and
01:10:23
effects of AI. We don't know the
01:10:26
economic benefits and effects of the
01:10:29
work that's being done to deregulate.
01:10:30
Another key point which is not talked
01:10:32
about by Navaro or anywhere else.
01:10:34
There's a broad effort to
01:10:37
deregulate standing up new energy
01:10:40
systems, deregulate industry and pharma,
01:10:42
deregulate banking. Besson talked about
01:10:44
this in our interview with him. All of
01:10:46
those deregulatory actions theoretically
01:10:49
should drive more investment dollars
01:10:51
because if you can get a biotech drug to
01:10:53
market in 5 years instead of 10, you'll
01:10:55
invest more in developing new biotech
01:10:56
drugs. If you can stand up a new nuclear
01:10:58
reactor in seven years instead of 30,
01:11:01
you'll build more nuclear reactors.
01:11:02
Money will flow. if you can um get a new
01:11:06
factory working because it's a lot
01:11:08
easier and faster to to build the
01:11:09
factory and cheaper, you'll build more
01:11:11
factories and production will go up.
01:11:13
People were really taken, by the way, by
01:11:15
your comment that you would shut up
01:11:16
about the deficit if we had like a
01:11:19
really great energy policy. We were
01:11:21
dumping a lot on top of it. I want to
01:11:23
build on the point that both Jamatha and
01:11:25
Freeberg made about growth rates. So,
01:11:27
there's a very important chart here from
01:11:29
Fred. This is the Federal Reserve St.
01:11:31
Louis. This is Federal Receipts. So
01:11:33
basically it's federal tax revenue as a
01:11:36
percent of GDP and this goes all the way
01:11:38
back to you know the 1930s 1940s. So if
01:11:42
you look in the postWorld War II period
01:11:44
you can see just eyeballing it that
01:11:47
there's a lot of variation around this
01:11:49
but the line is around
01:11:51
17.5% plus or minus 2%. And the
01:11:54
interesting thing is that this chart
01:11:58
reflects radically different tax rates.
01:12:01
So, for example, during some of these
01:12:04
periods, we've had 90% top marginal tax
01:12:07
rates. We've had 70% top top marginal
01:12:09
tax rates. So, yeah, under Jimmy Carter,
01:12:11
the top marginal tax rate was, I think,
01:12:14
70%.
01:12:16
We've had tax rates, you know, under
01:12:18
Reagan or Clinton in the 20s. So, the
01:12:21
point is that the the tax rate that you
01:12:24
have and what you actually collect as a
01:12:26
percent of GDP don't correlate. The most
01:12:29
important thing by far is just how the
01:12:32
economy is doing. If you look at the top
01:12:33
tick, it's around 2000 there. If you
01:12:35
just mouse over it, 1999 to 2000. Yeah.
01:12:38
Yeah. We get like just under 20% of
01:12:40
federal receipts% of GDP and tax rates
01:12:43
were quite low back then. The reason why
01:12:45
is we had an economic boom. So look, the
01:12:47
point is the most important thing in
01:12:49
terms of tax revenue is having a good
01:12:52
economy. And this is why you don't just
01:12:54
want to have very high tax rates because
01:12:57
they clobber your economy. So this point
01:12:58
that Navaro was making in that article,
01:13:02
it actually makes sense. I mean 1.7% is
01:13:04
a pretty tepid growth assumption, we
01:13:06
should be able to grow a lot faster. And
01:13:08
if we have a favorable tax policy, you
01:13:10
can grow a lot faster. Now, if you go to
01:13:13
spending, can you pull up the Fred chart
01:13:14
on spending? What you see here is that I
01:13:17
mean it's been kind of going up but
01:13:18
let's say that since the 19 mid1970s or
01:13:21
so federal net outlays as a percent of
01:13:24
GDP so basically spending was around 20%
01:13:26
of GDP and then what happened is during
01:13:29
co it went crazy went all the way up to
01:13:31
30% and now it's back down to you know
01:13:33
low 20s but it's still not back down to
01:13:36
20 and what we need to do is grow the
01:13:40
economy we have to grow GDP to the point
01:13:42
where federal net outlays are back
01:13:44
around 20%. If you could get tax revenue
01:13:48
to the historical mean of around 17.5%
01:13:50
or 17%, you get spending to 20%, then
01:13:53
you have a budget deficit of 3% which is
01:13:55
much more tolerable. And I think that's
01:13:57
best target under his 33 plan, right? Is
01:14:00
you get GDP growth back up to 3% and you
01:14:03
get the budget deficit down to 3%. All
01:14:06
right, Chimat, you had some charts you
01:14:07
wanted to share. Well, I think what's
01:14:10
amazing is if you take last week and now
01:14:13
again this week, we're all converging on
01:14:15
the same thing. The path out of this is
01:14:21
through GDP
01:14:23
growth. And I just want everybody to
01:14:26
understand where we are. And this is
01:14:29
without judgment. This is just the
01:14:30
facts. What this chart shows in gray is
01:14:34
the total supply of power in the United
01:14:39
States and the blue line is the
01:14:43
utilization. So what you build for is
01:14:48
what you think is a premium above the
01:14:50
demand, right? You'd say if there's one
01:14:51
unit of demand, let's have 1.2 units of
01:14:54
supply, we'll be okay. But as it turns
01:14:56
out, historically in the United States,
01:14:59
we've had these cycles where we didn't
01:15:02
really know what the demand curve would
01:15:04
look like. And so over the last number
01:15:07
of years, we've stopped really building
01:15:11
supply in
01:15:12
power. But what happened with things
01:15:14
like AI and all of these other things is
01:15:16
that the demand just continued to spike.
01:15:19
And so what this chart shows is we are
01:15:21
at a standstill sitting here today in
01:15:24
2025. On margin we're actually short
01:15:27
power which is to say sometimes there
01:15:30
are brownouts sometimes there is lack of
01:15:32
power because we didn't add enough
01:15:35
capacity. So that's where we are today.
01:15:37
So then we talk about all of these new
01:15:39
kinds of energy and this is just meant
01:15:41
to ground us in the
01:15:43
facts. If you tried to turn on a project
01:15:47
today sitting here in May of 2025,
01:15:51
here's what the timelines are. We all
01:15:54
talk about
01:15:55
SMRs, small modular
01:15:58
reactors. The reality is that if you get
01:16:02
everything permitted and you believe the
01:16:04
technology can be derisked, you're still
01:16:06
in a 2035 plus time frame. You're a
01:16:08
decade away.
01:16:10
If you have an unplanned NAT gas plant
01:16:13
today, the fastest you could get that on
01:16:15
is four years from now. If we tried to
01:16:18
restart a mothball nuclear reactor, of
01:16:22
which there are only three we can
01:16:24
restart, that's a 20 27 to 2030 time
01:16:28
frame. So, let's give us the benefit of
01:16:29
the doubt. That's 2 years away. If we
01:16:32
needed to plan Nat gas plant, there's
01:16:35
already 24 gawatt in the queue which
01:16:37
can't get turned on. So where does this
01:16:40
end up? And this is where I think we
01:16:42
need to strip away all the partisanship
01:16:44
and understand what we're dealing with.
01:16:47
We have ready supply of renewable and
01:16:51
storage options
01:16:53
today. It's the fastest thing that you
01:16:55
can turn on. It allows us to turn on
01:16:58
supply to meet the demand and
01:17:00
utilization. So I just think it's
01:17:03
important to understand that we must not
01:17:05
lose energy. We cannot lose the energy
01:17:07
market because that is the critical
01:17:09
driver of all the GDP. All right. Nepon
01:17:11
steel and the US steel merger got
01:17:13
cleared by President Trump. This was
01:17:16
something that was being blocked by
01:17:17
Biden obviously for national security
01:17:19
reasons. Nepon is going to acquire steel
01:17:22
for 14.9 billion. Biden blocked that as
01:17:24
we had discussed. On Friday, Trump
01:17:26
cleared the deal to go through calling
01:17:28
it a partnership that will create 70,000
01:17:30
jobs in the US. And on Sunday, Trump
01:17:32
called the deal an investment sync. It's
01:17:34
a partial ownership, but it will be
01:17:35
controlled by the USA. Chim, there seems
01:17:38
to be uh a reframing of this deal and
01:17:41
that the United States is going to
01:17:43
benefit from it, but it's not a sale.
01:17:46
Let's set some context. The United
01:17:48
States is always on the wrong side of
01:17:49
these deals. Okay? We've been on the
01:17:51
wrong side for 20 years. Meaning, we
01:17:53
show up when an asset is stranded
01:17:56
or completely run into the ground. For
01:18:00
example, we did the auto bailouts at the
01:18:03
end of the great financial crisis. If
01:18:05
it's not a company and there's toxic
01:18:07
assets, we set up something called TARP.
01:18:09
What do we get? Not much in return. In
01:18:12
this, it's the
01:18:14
opposite. And I think that this strategy
01:18:16
has worked for many other countries
01:18:19
really well. So if you look at
01:18:22
Brazil, companies like Embraer and
01:18:24
Valet, which are really big Brazilian
01:18:26
national champions, have a partnership,
01:18:28
a pretty tight coupling with the
01:18:30
Brazilian government. The Brazilians
01:18:31
have a golden vote. If you look inside
01:18:33
of the UK, there's a bunch of aerospace
01:18:36
and defense companies, including
01:18:37
Rolls-Royce, that have a very tight
01:18:39
coupling with the UK government. They
01:18:41
have a golden vote. If you look in
01:18:43
China, companies like Bite Dance and CL
01:18:47
have a very tight coupling with the
01:18:50
Chinese government and the Chinese
01:18:52
government has a golden vote. And so
01:18:54
what are all of those deals? Those deals
01:18:56
are about companies that are thriving
01:18:58
and on the forward foot. And so I think
01:19:00
this is a really important example of
01:19:02
things that we need to copy. I've said
01:19:04
this before, but one part of China that
01:19:06
I think we need to pay very close
01:19:07
attention to is Hu Jin Tao in 2003 laid
01:19:12
out a plan and he said we are going to
01:19:14
create 10 national champions in China in
01:19:16
all the critical industries that are
01:19:18
going to matter for the next 50 years
01:19:20
including things like batteries and rare
01:19:22
earths and AI and they did it but for
01:19:25
those companies that allowed them to
01:19:26
thrive and crush it and I think that we
01:19:29
need to do that and compete with those
01:19:31
folks on an equal playing field. So in
01:19:33
all industries or in very specific
01:19:35
strategic ones because that would seem
01:19:37
like corrupting capitalism in free
01:19:39
markets would be the steelman. Yeah,
01:19:42
there's 10 industries that matter and
01:19:44
you steel is one. Okay. I think the
01:19:48
precursors for pharmaceuticals are
01:19:50
absolutely critical. Got it. I think AI
01:19:53
is absolutely critical. I think the
01:19:56
upstream lithography and EV deposition
01:19:58
and chipm capability absolutely
01:20:01
critical. I think batteries are
01:20:04
absolutely critical and I think rare
01:20:06
earths and the specialty chemical supply
01:20:09
chain absolutely critical. If you have
01:20:11
those five, you are in control of your
01:20:14
own destiny in the sense that you can
01:20:16
keep your citizens healthy and you can
01:20:18
make all the stuff for the future. So I
01:20:21
think if the president is creating a a
01:20:24
more expansive idea beyond US deal with
01:20:27
this idea of US support, maybe there'll
01:20:30
be preferred capital in the future to US
01:20:32
deal. But if he creates a category by
01:20:35
category thing across five or six of
01:20:37
these critical areas of the future, I
01:20:39
think it's super smart and we should do
01:20:41
more of it. Sax, what do you think?
01:20:44
interventionism, putting your thumb on
01:20:45
the scale, golden votes, a good idea for
01:20:47
America in very narrow verticals or let
01:20:51
the free market decide. What are your
01:20:52
thoughts on this golden vote, having a
01:20:54
board seat, etc. Well, it depends what
01:20:58
the free market, so to speak, produced.
01:21:00
And the reality is over the past 25
01:21:02
years is we exported a lot of this
01:21:04
manufacturing capacity to China. And I
01:21:07
don't think it was a free market because
01:21:08
they had all these advantages under the
01:21:09
WTO that we talked about on a previous
01:21:11
podcast. they were able to subsidize
01:21:13
their national champions while still
01:21:17
remaining compliant with the WTO rules
01:21:18
because supposedly they were a
01:21:19
developing country. It was totally
01:21:20
unfair. And what they would do is
01:21:23
through these subsidies, they would
01:21:25
allow these national champions to
01:21:27
essentially dump their products in the
01:21:28
global market and drive everyone else
01:21:30
out of business. They became the lowcost
01:21:31
producers. I think that as the president
01:21:34
just said recently, not every industry
01:21:36
has to be treated as strategic clothes
01:21:39
and toys. is we don't necessarily have
01:21:40
to reshore in the United States but
01:21:42
steel production is definitely strategic
01:21:45
steel aluminum and I'd say the rare
01:21:47
earth we have to have that capacity we
01:21:49
cannot be completely dependent on China
01:21:51
for our supply chain so some of these
01:21:53
industries have to be reassured and if
01:21:55
you need subsidies to do it I think that
01:21:57
you do it for national security reasons
01:21:59
first and foremost there are other sense
01:22:01
yeah yeah there are other industries
01:22:03
where the private market works just fine
01:22:05
and what we need to do to help those
01:22:07
companies is simply not get in their
01:22:08
Hey, with unnecessary red tape and
01:22:10
regulations. So, I would say empower the
01:22:13
free market when America is the winner.
01:22:16
And then in other areas where they're
01:22:20
necessary for national security, then
01:22:21
you have to be willing to basically
01:22:23
protect our industries. Freeberg, it
01:22:25
seems like the great innovation here
01:22:26
might also be the American public
01:22:29
getting upside. When we gave loans to
01:22:32
Celindra and Tesla and Fiser and a bunch
01:22:35
of people for batterypowered, you know,
01:22:37
energy under Obama, we just got paid
01:22:39
back in some cases by Elon. Other people
01:22:42
defaulted, but we didn't get equity.
01:22:44
What if we had instead of getting our
01:22:46
500 million back in the loan from from
01:22:47
Elon, which he paid back early and with
01:22:49
interest, if we got half back and we got
01:22:51
half in equity, RSUs, whatever, stock
01:22:54
options, warrants, this would be an
01:22:56
incredible innovation. So, what are your
01:22:57
thoughts here? because people look to
01:22:59
this podcast as, hey, the free market
01:23:00
podcast, but this does seem to be a
01:23:03
notable exception here of maybe we
01:23:05
should get involved and do these golden,
01:23:09
you know, share votes, board seats, you
01:23:12
know, maybe more creative um structures
01:23:15
in order to win faster. What are your
01:23:17
thoughts, Reaper? I don't like it. I
01:23:20
don't like the government and markets.
01:23:22
Keep the government out of the markets.
01:23:24
It creates a slippery slope. First of
01:23:26
all, I think markets don't operate well
01:23:28
if government's involved. It gets
01:23:30
inefficient and that hurts consumers. It
01:23:32
hurts productivity. It hurts the
01:23:34
economy. Second, I think it's a slippery
01:23:36
slope. You do one thing. question
01:23:38
though. If government non-intervention
01:23:40
results in all the steel production
01:23:42
moving offshore, if it results in all
01:23:45
the rare
01:23:46
earth processing and the rare earth
01:23:49
magnet casting industries moving
01:23:52
offshore, in fact, not just moving
01:23:54
offshore, but moving to an adversarial
01:23:57
nation such that they can just switch
01:24:00
off our supply chain for pretty much
01:24:02
every electric motor. Is that an outcome
01:24:05
of the quoteunquote free market that we
01:24:06
should accept? Well, then I think that's
01:24:09
where the government can play a role in
01:24:10
trade deals to to manage that effect. So
01:24:13
you can create incentives that'll drive
01:24:15
onshore manufacturing by increasing the
01:24:18
tariff or restricting trade with foreign
01:24:20
countries so that there isn't a cheaper
01:24:22
alternative, which is obviously one of
01:24:24
the plays that this Trump administration
01:24:25
is trying to do. I' I'd rather have that
01:24:27
mechanism than the government making
01:24:30
actual market-based decisions and
01:24:31
business decisions. You know how
01:24:32
inefficient government runs. You know
01:24:34
how difficult it is to assume that that
01:24:36
bureaucracy is actually ever going to
01:24:37
act and pick any best interest or any
01:24:39
good interest at all. They're just going
01:24:41
to [Â __Â ] it all up. So, I'd rather keep
01:24:43
the government entirely out of the
01:24:44
market. Create a a trade incentive where
01:24:46
the trade incentive basically will drive
01:24:48
private markets, private capital to
01:24:51
build that industry on shore here
01:24:53
because there isn't one and there's
01:24:54
demand for it because you've restricted
01:24:56
access to the foreign market. That I
01:24:58
think would be the best general
01:24:59
solution. tax and then I think it's a
01:25:01
slippery slope because then you could
01:25:02
always rationalize something being
01:25:03
strategic, something being security
01:25:05
interest in the United States. So then
01:25:06
every industry suddenly gets government
01:25:08
intervention and government involvement.
01:25:09
And then the third thing is I don't want
01:25:11
the government making money that the
01:25:13
Congress then says, hey, we've got more
01:25:14
money, we got more revenue, let's spend
01:25:15
more money because then they'll create a
01:25:17
bunch of waste and nonsense that'll
01:25:19
arise from having increased revenue. one
01:25:22
side and I will say I one thing where I
01:25:24
do think we do a poor job is we don't do
01:25:26
a good job to answer your question Jal
01:25:28
of investing the retirement funds that
01:25:30
we've mandated through social security
01:25:32
we should be taking the $4.5 trillion
01:25:34
that our social security beneficiaries
01:25:36
have had deducted from their paychecks
01:25:38
over many many years and those social
01:25:41
security future retirees or current
01:25:43
retirees are getting completely ripped
01:25:45
off because their money is being loaned
01:25:46
to the federal government. It's not
01:25:48
being invested. It's been loaned to the
01:25:49
government to spend money and run a
01:25:51
deficit and ultimately inflate away the
01:25:53
value of the dollar. We should have been
01:25:55
investing those dollars in some of these
01:25:57
strategic assets. So if ever there were
01:25:59
to be shares or investment that the
01:26:01
government does, it should be done
01:26:02
through strategic investing through the
01:26:04
social security or retirement program.
01:26:06
Similar by the way to what's done in
01:26:07
Australia where these uh these supers
01:26:10
are have created an extraordinary
01:26:12
surplus of capital. Same in Norway, same
01:26:15
in all the Middle East countries.
01:26:16
incredible sovereign wealth funds that
01:26:19
benefit the retirees and the population
01:26:21
at large. That's where the dollars
01:26:22
should be invested from. I do think the
01:26:24
fundamental focus priority right now
01:26:26
should be reforming social security
01:26:28
while we still have the chance. We have
01:26:30
until 2032 when social security will be
01:26:32
functionally bankrupt and everyone's
01:26:34
going to get overt taxed and kids are
01:26:36
going to end up having to pay um through
01:26:38
inflation for the benefits of the
01:26:40
retirees of the last generation.
01:26:41
Freeberg's right. We're on a seven-year
01:26:42
shock clock to when social security is
01:26:45
not funded. And by the way, this
01:26:46
opportunity to fix mandatory spending,
01:26:48
it was an opportunity to introduce some
01:26:50
structural reform in social security.
01:26:52
Another reason why I think that there's
01:26:53
a degree of discretzia in this bill,
01:26:55
particularly with how Congress had acted
01:26:57
and not addressing what is becoming a
01:26:59
critical issue because everyone wants to
01:27:01
get reelected in the next 12 months, 18
01:27:03
months. They've got elections coming up.
01:27:05
So, everyone's scrambling to not mess
01:27:06
with that because you can't touch it.
01:27:07
It's like, you know what, guys? This is
01:27:09
bankrupt in seven years. It's going to
01:27:11
cost us 5 10 times as much when we have
01:27:13
to deal with it when everyone runs out
01:27:14
of money. Deal with it now. Fix the
01:27:17
problem. And by the way, we should flip
01:27:18
all that money, $4.5 trillion into an
01:27:21
investment account for the retirees
01:27:23
where they can own equities and they can
01:27:25
make investments in the markets and they
01:27:26
can participate in the upside of
01:27:28
American industry and the GDP growth
01:27:29
that's coming. Instead, they're getting
01:27:31
paid 3.8% or four 4 and a.5% average
01:27:34
from treasuries that they own that, by
01:27:36
the way, are now have a lower credit
01:27:38
rating than they've ever had. You know,
01:27:40
it's crazy. I I'm I'm in complete
01:27:42
agreement with you and I think it's a
01:27:43
lack of leadership on Trump's part. If
01:27:45
Trump is going to criticize Taylor Swift
01:27:47
and Zalinski and Putin and everybody,
01:27:50
you know, all day long on Truth Social,
01:27:52
he can criticize Congress and the
01:27:54
Democrats and the Republicans on not
01:27:57
cutting spending. I think he should
01:27:59
speak up. I think he was elected to do
01:28:01
that. It was a big part of the mandate
01:28:03
and uh he should tone down the tariff uh
01:28:06
chaos and tone up the uh and lean into
01:28:10
uh intelligent immigration you know
01:28:12
recruiting great talent to this country
01:28:14
and he should be pushing to make these
01:28:17
bills uh control spending that's just
01:28:19
one person's belief for the chairman
01:28:20
dictator hapatia your zar David Sachs in
01:28:25
that Chris Brion white shirt very
01:28:27
beautiful and the sultan of science deep
01:28:30
in his wal E era. I am the world's
01:28:33
greatest moderator and as Freeberg will
01:28:35
tell you, executive producer for life
01:28:37
here at the All-In podcast. We'll see
01:28:38
you all next time. Bye-bye. Jason.com.
01:28:40
Love you boys. Bye-bye.
01:28:43
We'll let your winners ride.
01:28:45
Rainman David.
01:28:50
We open sourced it to the fans and
01:28:52
they've just gone crazy with it. Love
01:28:54
you. Queen of
01:28:56
[Music]
01:29:02
Kino besties are gone.
01:29:05
That is my dog taking your driveways.
01:29:10
Oh man, my appetiter will be. You should
01:29:13
all just get a room and just have one
01:29:15
big huge orgy cuz they're all just like
01:29:17
this like sexual tension that they just
01:29:19
need to release somehow.
01:29:23
Wet your feet. her feet. That's going to
01:29:26
be good. We need to get Murphy's
01:29:31
[Music]
01:29:36
our all in.
