00:00:00
all right everybody welcome back to the
00:00:01
all-in Pod episode 174 almost up to 175
00:00:06
I believe there are a lot of folks
00:00:07
getting together next week at the all-in
00:00:09
meetups around the world to watch the
00:00:11
show unfortunately jcal had an oral
00:00:14
incident this week he will not be
00:00:18
joining us I guess he cracked a tooth
00:00:21
and couldn't make it so he is sitting
00:00:25
this one out very last minut decision
00:00:27
cracked her tooth eating a bison r about
00:00:30
the Salt Lick in hustin which is a true
00:00:32
story freir I think we found an image of
00:00:35
okay let's Jake out at the Salt Lick
00:00:36
cracking his tooth on a
00:00:37
[Music]
00:00:41
rib let your winners
00:00:44
ride
00:00:48
Rainman and instead we open source it to
00:00:51
the fans and they've just gone crazy
00:00:52
with
00:00:53
[Music]
00:00:56
it all right so we are going to announce
00:01:00
be all-in Summit 2024 back to
00:01:03
LA we had such a great time in LA it was
00:01:06
a great location for everyone we really
00:01:07
appreciated the facilities at UCLA this
00:01:11
year we're upgrading the experience once
00:01:12
again for everyone so the event's going
00:01:14
to be September 8 through 10th in La if
00:01:17
you haven't seen it please watch the
00:01:19
recap video from last year we put it up
00:01:21
on
00:01:22
YouTube and then there's all the
00:01:24
interviews from last year are also on
00:01:25
YouTube we're going to have another
00:01:27
amazing lineup of speakers this year and
00:01:29
have a much more kind of upgraded
00:01:31
elevated experience for everyone
00:01:33
applications are open today so go to
00:01:35
summit. Allin
00:01:37
podcast.co to put an application in to
00:01:40
join us this year as you know we had way
00:01:42
more applications last year than spots
00:01:44
and this year we're going to have all
00:01:46
sorts of upgrades we're GNA have
00:01:47
transportation to and from all the
00:01:49
events for everyone a concierge booking
00:01:50
service to help you book your rooms and
00:01:53
get you set up in La for the trip we're
00:01:55
doing a full day event on the Sunday
00:01:57
September 8th followed by two days of
00:01:59
content Dy about town on the first night
00:02:02
two blowout parties it's going to be
00:02:04
awesome so we're really excited about it
00:02:06
we've got a much bigger budget this year
00:02:08
and we're going to try and really make
00:02:09
the uh the experience awesome while also
00:02:12
delivering the great content so really
00:02:14
excited to do this again in La I
00:02:16
appreciate everyone's support of that we
00:02:18
had a lot of debate internally about
00:02:20
where to take
00:02:21
it and we settled on going with what we
00:02:25
know and trying to make it a little bit
00:02:26
better this year so anything else you
00:02:28
guys want to say about Summit and
00:02:30
there's only one kind of ticket this
00:02:32
year right that's right so we're not
00:02:33
doing like regular ticket and VIP we're
00:02:36
just doing one ticket for everyone and
00:02:38
everyone's gonna get this kind of
00:02:39
elevated experience she'll be cool sax
00:02:42
just show up would be awesome hope to
00:02:45
see you there I want to invite a couple
00:02:46
of speakers yeah you've got some good
00:02:48
speakers I saw yeah it's gonna be great
00:02:50
when are we gon to start announcing the
00:02:52
speakers freeberg we will do that in a
00:02:54
couple weeks we already have a bunch
00:02:57
booked and it's going to be awesome oh
00:02:59
that's really exciting that's awesome
00:03:02
yeah I know CP you wanted to take it
00:03:03
somewhere else I mean we were in debate
00:03:05
on Vegas maybe going to some other
00:03:09
country but uh you went to another
00:03:11
country where'd you go this week I was
00:03:12
in Paris France what were you doing
00:03:15
there there's like a important AI
00:03:18
conference and Jonathan Ross the founder
00:03:21
of grock and I kicked it off with like a
00:03:24
fireside chat for like 45 minutes at the
00:03:26
beginning just to get everybody amped up
00:03:28
and excited
00:03:30
cool yeah it was really anything
00:03:33
interesting come out of
00:03:34
it I think the sound bite that made an
00:03:38
impact on me was was from Jonathan so I
00:03:41
think he did a very good job of
00:03:43
explaining where Nvidia is really strong
00:03:46
which is really in learning and where
00:03:49
Nvidia is trying to build the
00:03:51
business which is an inference but the
00:03:54
problem is the features almost compete
00:03:57
with each other between learning and
00:03:59
inference and I think that that was a
00:04:00
good clarification and then the second
00:04:03
thing is he said that if he deploys his
00:04:04
road map he'll have 50% of the available
00:04:07
inference compute by the end of next
00:04:09
year wow based on all the needs of like
00:04:12
open Ai and meta and everybody else
00:04:14
which is a wait but that's a smaller
00:04:16
physical footprint is that right and
00:04:17
it's just because they can do higher
00:04:18
token per second less power higher
00:04:20
tokens per second across all the major
00:04:24
models so it would it would be
00:04:26
meaningfully less power and meaningfully
00:04:28
cheaper
00:04:30
so that was a pretty amazing sound bite
00:04:32
though his his whole explanation was
00:04:34
actually really good so I mean if you
00:04:35
listen to the spaces we did I thought it
00:04:37
was good I thought this was even better
00:04:39
so I'm trying to get a copy of
00:04:42
the of the talk because it was taped so
00:04:45
that we can kind of post it at a minimum
00:04:48
anybody who's interested in studying
00:04:49
Nvidia should listen to what he has to
00:04:51
say because I think the the nuances
00:04:53
between us and them are really profound
00:04:55
and they were pretty Stark the way that
00:04:57
he described them so it was cool it was
00:04:58
great to be in Paris frankly how does
00:05:00
Nvidia respond to this movement or this
00:05:03
notion that there's an entirely
00:05:04
different chip category focused purely
00:05:06
on inference do they have inference
00:05:08
focused chips in development or a
00:05:10
different road map that they're going to
00:05:11
have to kick up now I mean I think it's
00:05:13
very hard for them it's it is a
00:05:15
structural innovators dilemma and why is
00:05:18
that because they've made some
00:05:19
architectural decisions around things
00:05:21
like hbm which is high bandwidth memory
00:05:23
or specific kinds of extremely high
00:05:26
throughput optical cables and why these
00:05:29
things matter is that they've gone and
00:05:30
bought up the world Supply and they've
00:05:33
basically bare hugged these markets as a
00:05:36
structural part of their design now that
00:05:38
makes sense and it's very legal the
00:05:42
problem is that if you architect around
00:05:43
it and go through a totally orthogonal
00:05:46
design process we're not subject to any
00:05:49
of those supply chain considerations and
00:05:51
I think he explained that really well I
00:05:53
didn't understand it in nearly as much
00:05:55
detail as I did leaving so I think
00:05:57
that's an important takeaway it's also
00:05:58
really important for for all the
00:05:59
hundreds of Chip startups and for these
00:06:01
Venture investors who are ripping money
00:06:03
into these companies to know these
00:06:04
differences because if you are investing
00:06:07
in a company that is subject to you know
00:06:10
hbm in their design decisions or these
00:06:13
specific optical cables or a whole bunch
00:06:14
of other things you're going to be in a
00:06:16
really difficult spot because nvidia's
00:06:18
bought them
00:06:19
all so I thought that that was a that
00:06:22
was really interesting the other thing
00:06:23
that Nvidia does brilliantly is what
00:06:24
Intel did in the 80s and 90s when they
00:06:26
wanted to kind of Choke the market for
00:06:29
CPU competition which is that they were
00:06:32
able to define a metric that everybody
00:06:35
started to pay attention to which for
00:06:36
them was clock speed and if you remember
00:06:39
Intel would push Moore's Law and they
00:06:42
would push transistor density and clock
00:06:45
speed as the measurement of value now
00:06:47
for a lot of consumers none of us knew
00:06:50
any better and so every time we saw a
00:06:53
new chip we would go through the Pentium
00:06:55
upgrade cycle because we thought that's
00:06:56
what we were supposed to do it turns out
00:06:59
that clock speed doesn't really factor
00:07:01
into speed the way that you think about
00:07:03
it and experience it as an end user
00:07:06
using software and it's the same in Ai
00:07:09
and so we talked about that as well so
00:07:11
it's an incredibly important company
00:07:13
NVIDIA is I think they're performing
00:07:16
brilliantly they're running many of the
00:07:18
same plays that Intel did to try to
00:07:20
create total dominance but there's some
00:07:23
very important nuances
00:07:25
around power and around supply chain
00:07:28
decisions and Technology decisions
00:07:30
that frankly some people would think are
00:07:33
unsustainable and the more I understand
00:07:35
them I would be in that camp as
00:07:37
well so I thought it was a very
00:07:39
important conversation and hopefully we
00:07:42
can post it yeah it cool it and by the
00:07:45
way it's great to be in Paris too what
00:07:46
do you think of Paris you know I have a
00:07:48
new appreciation for what's going on in
00:07:50
these major cities I think that all
00:07:52
major cities are crumbling I don't and I
00:07:54
don't mean to be alarmist or very
00:07:56
dramatic about that it's just that I
00:07:59
spent aot lot of time with entrepreneurs
00:08:00
that are in
00:08:02
Paris and they struggle with the same
00:08:04
things that the people in San Francisco
00:08:06
do and then I had couple of
00:08:08
conversations with some folks that are
00:08:09
entrepreneurs in London who were
00:08:11
there and they suffer from the same
00:08:13
things lots of petty crime lots of
00:08:16
garbage lots of vandalism lots of
00:08:19
drugs and so no place is perfect it
00:08:22
turns out everybody is struggling from a
00:08:25
lot of experimentation gone wrong and
00:08:27
how these cities run that was that was
00:08:28
actually my
00:08:30
the biggest takeaway outside of the
00:08:31
actual AI conference itself but that's a
00:08:33
western City problem you're not I mean
00:08:35
have you visited many cities in China or
00:08:38
other parts of Asia and is that like a
00:08:40
western problem I have not been to China
00:08:43
in six years I've been to Hong Kong but
00:08:47
I've not been into China for six years
00:08:49
so I don't
00:08:50
know I would say that India has its own
00:08:53
vagaries so those issues are not the
00:08:55
same I've been to Singapore a couple
00:08:58
times my kids were just in Japan for
00:09:00
spring break so I saw a bunch of the
00:09:02
pictures those cities look relatively
00:09:05
the same very orderly well run clean but
00:09:08
yeah you're right these Western cities
00:09:10
are a bit chaotic yeah sax is this like
00:09:12
a western social liberal
00:09:15
manifestation how cities are being
00:09:17
challenged generally M like Malay has I
00:09:20
think Malay said this right he said like
00:09:22
all these Western social liberal beliefs
00:09:25
are having lifestyle impact in urban
00:09:28
centers well it's hard for me to speak
00:09:30
to what's happening in Europe but we
00:09:31
know what's happening in the United
00:09:33
States all the biggest cities are blue
00:09:36
and they're run by a certain ideology
00:09:39
and in those cities we've had now a
00:09:42
decade plus of decriminalization de
00:09:45
prosecution
00:09:47
decarceration and the impacts of this
00:09:49
are obvious we also at least in San
00:09:52
Francisco we have subsidies for people
00:09:55
who want to live on the streets and do
00:09:56
drugs all day I mean it's not prohib to
00:09:59
do that so all of our public spaces have
00:10:01
been taken over by people who are happy
00:10:03
to take the whatever it is $7 or $800
00:10:06
subsidy and do drugs all day you also
00:10:08
had this referendum in California that
00:10:12
essentially lowered a whole bunch of
00:10:15
felonies to misdemeanor basically if you
00:10:17
steal less than
00:10:20
$950 that's considered a misdemeanor now
00:10:22
instead of a felony and so there there's
00:10:24
been a huge increase in crime because of
00:10:27
that you know shoplifting just doesn't
00:10:29
get prosecuted anymore and misdemeanors
00:10:31
don't get prosecuted anymore we've had
00:10:32
all these sour daas who don't want to
00:10:34
prosecute misdemeanors so I think if
00:10:36
voters had understood that
00:10:40
lowering a lot of these crimes from
00:10:42
felonies to misdemeanors meant that they
00:10:43
wouldn't get prosecuted at all I don't
00:10:45
think they would have done it but that's
00:10:46
kind of where we are but are we seeing a
00:10:48
rebound from that now Sach is democracy
00:10:50
working because I know that in San
00:10:51
Francisco we had an election recently
00:10:53
and in the ballots it seemed to indicate
00:10:55
that voters were sick and fed up with
00:11:00
the the living situation in the city and
00:11:02
they're putting different people in
00:11:04
office and have an intention of changing
00:11:06
the city Charter and
00:11:09
changing the the laws in the city so
00:11:11
that we can reverse this trend is that
00:11:14
not where things are I mean like sitting
00:11:16
you know sitting where you sit like
00:11:17
don't you see San Francisco kind of
00:11:19
trying to write write itself and as a
00:11:21
result doesn't democracy work over time
00:11:24
yeah it's just that we live in a one
00:11:26
party State and that one party has been
00:11:28
captured by ideologues and so it's very
00:11:30
hard to to change it is true that in San
00:11:33
Francisco there was an election recently
00:11:35
and the more
00:11:36
radical supervisors were defeated we
00:11:39
have I think finally a good da in San
00:11:42
Francisco but these um radicals have had
00:11:45
a long time to entrench their power and
00:11:47
the party Machinery helps keep them
00:11:50
entrenched so yeah I think you can argue
00:11:52
that we've bottomed out because the
00:11:53
public has recognized the problem but
00:11:55
it's still very hard to fix and it's
00:11:57
going to take a while I think for the
00:11:58
first time in a long time time I'm more
00:12:00
optimistic than not given the people
00:12:03
I've seen that tolerated the conditions
00:12:06
historically have reversed course over
00:12:08
the last year and have said we got to
00:12:09
act and we gotta yeah you know do our
00:12:11
kind of Civic process well time with
00:12:13
tell you have to remember like when when
00:12:15
New York City was a chaotic mess in the
00:12:17
70s and 80s it took two different mayors
00:12:20
and it took almost a decade for that to
00:12:22
get cleaned up but it took the strength
00:12:24
of leadership first David dinkin set the
00:12:27
framework and then Giuliani really
00:12:29
executed Rudy Giuliani and between the
00:12:32
two of them and I think what people call
00:12:34
the broken windows Theory they hired a
00:12:36
lot more policemen and they just really
00:12:38
cleaned up all of these really bad areas
00:12:42
and then it had this knock on effect the
00:12:44
problem is when I'm staying at a hotel
00:12:47
in Paris or when I was speaking to some
00:12:49
of my friends who live in London it's
00:12:52
the same thing it's like the amount of
00:12:53
knife crime is so egregious as an
00:12:56
example that they tell you don't wear a
00:12:58
watch I mean I don't I don't wear
00:13:00
watches anymore when I go out but what's
00:13:03
crazy is they will tell you that you'll
00:13:04
get stabbed over a phone where it's like
00:13:07
you don't need to stab me for my iPhone
00:13:09
here just take the iPhone so there's a
00:13:11
level of lawlessness and petty crime
00:13:13
that's affecting
00:13:16
tourists that to me was like very
00:13:18
surprising and so even when I was
00:13:19
walking around Paris I was like should I
00:13:22
really
00:13:23
feel scared here and my reaction was no
00:13:27
I've never felt scared in Paris but then
00:13:29
what's ringing in the back of my ears
00:13:31
are These Warnings from people that are
00:13:33
that are that have lived there for 10
00:13:35
and 20 years who've had this stuff
00:13:37
happened to them that was really
00:13:38
surprising to me so that stuff in San
00:13:40
Francisco as far as I can tell or any
00:13:41
major city in the west hasn't really
00:13:43
been fixed because you haven't had a
00:13:44
shift in leadership yet where they've
00:13:46
really said okay we're going to double
00:13:48
down on policemen and policing and we're
00:13:51
going to amp up the punishment that
00:13:54
hasn't happened yet well I was just in
00:13:55
Vegas this week and it doesn't feel how
00:13:58
was your trip give us the trip report
00:14:00
yeah I went to the Google next
00:14:01
conference but I will say first off like
00:14:03
no crime in Vegas like Vegas is Vegas
00:14:06
it's it's a very different
00:14:08
environment but I yeah I went to the
00:14:10
Google next conference I ended up going
00:14:12
out because at ohal we did a partnership
00:14:15
with uh with gcp so we're moving all of
00:14:17
our infrastructure and doing a lot of
00:14:19
model development on
00:14:21
gcp and it was great I think there was
00:14:23
30,000 people at this event which was
00:14:25
insane have you guys been to any of
00:14:27
these like AWS reinvent or
00:14:29
the Google next or any of these no I
00:14:32
have not what I didn't realize prior to
00:14:35
going to this was how much of an
00:14:36
ecosystem there is like you go they they
00:14:38
take over the whole Convention Center at
00:14:40
manderlay Bay and there's hundreds of
00:14:42
companies that have built services or
00:14:45
build services on gcp that are there
00:14:47
pitching cios and CTO of other companies
00:14:50
and so it's this real Network effect
00:14:52
type of business model that I hadn't
00:14:54
fully appreciated just how like deep the
00:14:57
network is and and how entrenched it is
00:14:59
I I also got to go to this thing where I
00:15:01
spent a lot of time with c and CTO at
00:15:04
other companies big
00:15:06
companies and a big takeaway for me that
00:15:08
I also didn't realize was how most of
00:15:10
them are not on a single cloud
00:15:12
provider almost everyone I spoke with is
00:15:15
on multiple clouds so they're all on
00:15:18
like AWS and AER and gcp and they use
00:15:23
different aspect of each cloud and
00:15:25
there's a lot of interoperability
00:15:27
between the clouds now which I had
00:15:29
previously my naive concept of this
00:15:31
because I'm obviously not deeply in the
00:15:32
space was that a company picks a cloud
00:15:35
and they move everything over to that
00:15:36
one service provider but it turns out
00:15:39
that they split and then they allocate
00:15:41
dollars based on the services and the
00:15:43
pricing so it was pretty clear that
00:15:45
everyone is it's like a race to the
00:15:48
bottom on pricing for basic
00:15:51
commoditizing services in the cloud like
00:15:53
data storage right and then whoever has
00:15:56
the better higher level Services Whoever
00:15:57
has access to better mod or has a lower
00:15:59
price on running models or on running
00:16:02
things like inference that's where the
00:16:05
money is made and the gross margin on
00:16:07
that is quite different and I think
00:16:09
that's where a company like Google
00:16:11
probably has a much bigger Advantage
00:16:13
because of the uniqueness and the
00:16:15
quality of the models that they've
00:16:17
developed but yeah I just didn't I
00:16:18
didn't really appreciate like how much
00:16:20
there was so so my other big takeaway
00:16:22
was the cloud is a wide open competitive
00:16:24
market right now like there is no AWS
00:16:27
lockin in this market there's no a zero
00:16:29
lock in there's no gcp lock in like you
00:16:31
guys yeah the lock in comes from the
00:16:33
developer services that they provide so
00:16:36
both Amazon and Google and I'm I'm
00:16:38
assuming Azure as well will give you
00:16:41
access to incredibly talented Engineers
00:16:43
who will help you build production code
00:16:45
yeah and the ecosystem on how so like a
00:16:47
lot of the clouds are competing for the
00:16:50
Consultants who actually do the work for
00:16:53
the big Enterprises so Cap Gemini deoe
00:16:56
PWC they have these massive booths
00:16:59
facilities and then go and build tools
00:17:01
for companies is also a key part of the
00:17:03
business strategy and yeah I mean it's
00:17:05
such a competitive market I did not
00:17:08
appreciate just like how much of an open
00:17:11
game this still was sure it was a great
00:17:13
great couple days in Vegas and I got to
00:17:16
see some of our friends we ducked in and
00:17:17
saw helmuth in the poker Go studio for a
00:17:20
minute some big tournaments going on at
00:17:22
the ARA it was fun we had a good time
00:17:24
okay let's move on so you know on the
00:17:26
docket today we were going to talk about
00:17:29
CPI the March CPI numbers came in from
00:17:31
the Bureau of Labor Statistics yesterday
00:17:34
higher than forecast at 3.5%
00:17:37
year-over-year this is up from 3.2%
00:17:40
year-over-year inflation reported by the
00:17:42
Bureau of Labor Statistics in February
00:17:44
higher than expected for three straight
00:17:46
months now let's pull up this chart next
00:17:48
as we all know the FED did 11 rate hikes
00:17:50
in 2022 and 23 and the intention was
00:17:54
that we would see inflation come down to
00:17:56
their target of 2% and they would start
00:17:58
to lower rate rates and obviously we're
00:18:00
not seeing that inflation is remaining
00:18:03
hot things are getting more and more
00:18:06
expensive I pulled together some of the
00:18:09
numbers on year-over-year price increase
00:18:12
housing is up 5.7% year-over-year
00:18:15
Transportation costs are up
00:18:17
10.7% year-over-year and the highest car
00:18:21
insurance is up
00:18:22
22.2% year-over-year
00:18:25
so the expectation has been that the
00:18:27
market will do rate Cuts Larry Summers
00:18:30
came out and said you have to take
00:18:31
seriously the possibility that the next
00:18:32
rate move will be upwards rather than
00:18:35
downwards I guess sax let me let me ask
00:18:38
you first for your commentary on how
00:18:41
this plays into the election cycle I
00:18:42
know Biden was expecting a rate cut or
00:18:46
several rate Cuts going into the
00:18:47
election which usually helps the case of
00:18:48
the incumbent maybe you can comment on
00:18:51
where the the election cycle might be
00:18:53
influenced by this inflation report well
00:18:56
I think it's definitely bad news for
00:18:57
Biden I was expecting fed put this year
00:19:00
and really the whole Market was the
00:19:02
market was expecting three rate Cuts
00:19:03
this year which would have been really
00:19:05
good news for for Biden and I think this
00:19:08
latest inflation print was kind of a
00:19:09
dagger to the heart of of that
00:19:12
expectation and I think the Wall Street
00:19:14
Journal put it best it said here
00:19:16
Wednesday's report had been hotly
00:19:19
anticipated because fed leaders had been
00:19:21
willing to play Down stronger than
00:19:24
anticipated inflation readings in
00:19:26
January and February as as reflecting
00:19:29
potential seasonal quirks but a third
00:19:32
straight month of above expectations
00:19:34
inflation data erodes that story and
00:19:38
could lead fed officials to postpone
00:19:39
anticipated rate Cuts until July or
00:19:41
later so it's not just the fact that
00:19:44
this inflation print was higher than
00:19:46
expected it was also higher than
00:19:47
expected in January and February but
00:19:50
people were willing to kind of Overlook
00:19:51
that saying well maybe it was just kind
00:19:53
of a you know quirky reading but now
00:19:56
we've had three straight months it's
00:19:57
pretty clear that the narrative that we
00:19:59
had going into this year which was that
00:20:02
inflation was on its way down that it
00:20:04
peaked at 9% uh as the official rate I
00:20:07
think in 2022 and then it was going down
00:20:10
every month through all of 2023 and I
00:20:13
think the expectation going into 24 was
00:20:16
it would keep going down we get these
00:20:17
rate Cuts well after three straight
00:20:19
months of inflation being more
00:20:22
persistent and stickier Than People
00:20:24
expected I think that narrative is
00:20:25
basically dead so I think the narrative
00:20:29
now is it's going to be rates are going
00:20:32
to be higher longer and I think Larry
00:20:34
Summers is reflecting that view Larry's
00:20:36
going further he's not just saying that
00:20:38
rates are going to be you know at the
00:20:40
current rates for longer he's saying
00:20:41
they might actually
00:20:42
increase and that's a risk yeah last
00:20:45
summer he said and this was last summer
00:20:47
so almost a year ago he was saying the
00:20:49
Market's got it totally wrong we're
00:20:51
actually going to need much higher rates
00:20:52
than the Market's anticipating for much
00:20:54
longer as well than the market is
00:20:56
anticipating and so did Mario drogy both
00:20:58
said the same thing no one paid
00:21:00
attention to him everyone ignored it and
00:21:02
assumed that this was going to be a
00:21:03
quick rebound to normalization it's
00:21:05
clear that you know once again Larry has
00:21:07
proven himself to be fairly
00:21:09
preci understanding where are headed
00:21:11
Larry Summers has for the most part been
00:21:13
spot on from this on this whole
00:21:15
inflation question going all the way
00:21:16
back to 2021 remember 100% 100% in q1 of
00:21:20
2021 Biden's first quarter in office the
00:21:23
big legislative push was for the $2
00:21:26
trillion covid relief bill this the so
00:21:28
called American Rescue plan and Larry
00:21:30
Summers said that it was it risked
00:21:32
inflation it was an inflationary bill it
00:21:34
was stimulatory yeah too didn't need it
00:21:37
the the economy was already coming back
00:21:38
and we didn't need it and Biden was
00:21:40
risking inflation but of course
00:21:41
inflation was only at 2% at that point
00:21:44
so the administration kind of poo pooed
00:21:46
Larry and said you know that's is Larry
00:21:48
Summers being Larry or whatever and sure
00:21:51
enough we were at 5% inflation by that
00:21:54
summer and you have to wonder if Biden
00:21:57
had listened to that advice
00:21:59
would he be in a different position
00:22:01
right now going into the election I
00:22:02
think probably he would have been and
00:22:04
you got to wonder for what I mean what
00:22:06
are that two trillion dollar accomplish
00:22:08
I mean covid was winding down I mean
00:22:10
these were the last bit of stimy checks
00:22:13
and payments to these Pharma companies
00:22:17
and I mean it was basically a grab bag
00:22:18
and it was passed on straight party
00:22:21
lines and it was just totally
00:22:22
unnecessary I mean the economy certainly
00:22:24
didn't need it and so here we are and I
00:22:26
think that a a lot of people including
00:22:29
the markets thought that Biden was kind
00:22:33
of out of the woods that this year we'd
00:22:35
see the final leg of inflation going
00:22:38
back to normal but that's not going to
00:22:39
happen and I think you know going beyond
00:22:41
the political ramifications I think
00:22:43
there could be several other knock on
00:22:45
effects that we should talk about I mean
00:22:48
one is for the
00:22:51
consumer this means that the cost of
00:22:53
borrowing is going to be higher for
00:22:55
longer that makes it harder to buy a
00:22:56
house mortgage payments are higher
00:22:58
that also means if there's fewer home
00:23:01
sale transactions that means the price
00:23:03
of housing could come down so there
00:23:05
could be a correction in that market
00:23:07
second if you want to buy a car your car
00:23:09
payment's higher and if you have loans
00:23:11
uh your your personal interest is going
00:23:13
to be higher and this is why I think
00:23:15
consumers feel like they're worse off
00:23:18
than the economic data would otherwise
00:23:20
reflect and Larry Summers actually had a
00:23:21
tweet storm about this about a month ago
00:23:24
where he calculated that if you included
00:23:26
the cost of borrowing in infl
00:23:29
that inflation was much higher than
00:23:30
people thought and that it actually
00:23:32
peaked not at 9% but at 18% so Larry had
00:23:36
an excellent tweet storm on that and he
00:23:38
said that the cost of borrowing which
00:23:40
used to be calculated in inflation but
00:23:42
is not anymore was the reason why
00:23:44
consumer sentiment about the economy was
00:23:47
depressed he said that that accounted
00:23:49
for about 70% of it so people should be
00:23:52
feeling better off because GDP growth is
00:23:55
good and unemployment is low but they're
00:23:57
not because inflation is so high and if
00:24:00
you include cost of
00:24:01
borrowing the inflation's even higher so
00:24:04
I think this is really bad news for
00:24:06
Biden but it it's a story that's been
00:24:07
going on now for three years it's not
00:24:10
just this one inflation print right it's
00:24:13
funny yesterday Biden said during a
00:24:16
press conference with Japanese prime
00:24:18
minister kashida that he's sticking with
00:24:21
his prediction of a rate cut before
00:24:23
Year's end but there might be a delay uh
00:24:25
which I think is obviously
00:24:28
you know based on what the data is
00:24:29
showing and what folks that seem to know
00:24:32
and have been pretty good predictors are
00:24:34
saying is unlikely we were supposed to
00:24:35
get one in April yeah and we were
00:24:37
supposed to get one in June oh yeah
00:24:39
remember the fed the FED Open Market
00:24:40
Committee in December stated that they
00:24:43
expected three rate Cuts this year and I
00:24:45
think the market in some cases we're
00:24:46
predicting as high as four or five right
00:24:49
and now we may not have any and we may
00:24:51
actually see a rate hike if you were to
00:24:54
follow some well if we if we see a rate
00:24:55
hike before the election I think Biden
00:24:57
is toast
00:24:59
yeah but I mean I think Summers gave
00:25:00
that maybe a 15 to 20% chance which is
00:25:03
it's still not the most likely scenario
00:25:05
but uh possible but possible no one's
00:25:08
counting on that in December chamat if
00:25:10
you're a Fed
00:25:11
governor and obviously the the the
00:25:14
Federal Reserve Board of Governors sets
00:25:16
the uh rates they meet and they they
00:25:18
make these decisions how influenced are
00:25:21
they politically how how influenced are
00:25:23
they by this election cycle and you know
00:25:25
historically this is meant to be a
00:25:26
fairly independent board but there's
00:25:28
been a lot of consternation over the
00:25:31
last few years that this board acts like
00:25:33
almost a political apparatus
00:25:35
particularly an election yours you mean
00:25:38
when they're not
00:25:40
plagiarizing well tell us what you're
00:25:42
talking
00:25:45
about it turns out that before I answer
00:25:48
your question that that was a joke Nick
00:25:50
you can please show the Tweet but there
00:25:52
was an article that came out that said
00:25:55
one of the FED Governors has apparently
00:25:58
massively plagiarized most of the work
00:26:00
that got her to the position of being
00:26:02
fed Governor yeah so the same thing that
00:26:04
happened to the president of Harvard so
00:26:06
TBD what happens to that fed Governor
00:26:08
but it seems you know my observation
00:26:10
there was just more that it's like the
00:26:12
preferred method of defenestration now
00:26:14
of academics and the appara right before
00:26:17
if you had right leaning views or
00:26:20
Centrist views you get run out of town
00:26:22
and get fired now all of those folks are
00:26:25
using these search engines to basically
00:26:28
f fig out really quickly that you
00:26:29
plagiarize large quantity of your
00:26:32
academic work and that your scholarship
00:26:36
is is a little bit more speculative
00:26:38
which then brings into question why do
00:26:39
you get the right to to help set the
00:26:42
policy of the most important economy in
00:26:43
the world well anyways so that that was
00:26:45
that joke but so I think that you're
00:26:48
asking the absolute right question I
00:26:51
think that the FED has become
00:26:53
increasingly
00:26:55
politicized and I do think that there
00:26:59
are actions that foreign governments
00:27:02
take to influence the election but I do
00:27:04
not think it's what most people think
00:27:07
when I say that most people ideas goes
00:27:09
to like some dark room conspiracy theory
00:27:12
and misinformation and I think that's a
00:27:15
little bit naive I think the more
00:27:17
obvious way in which other
00:27:20
governments can impact the US election
00:27:23
is exactly how Sac said which is they
00:27:25
know that if inflation ticks High and
00:27:28
interest rates are up and consumers are
00:27:29
under pressure and meaningfully
00:27:31
displeased with the economy they'll vote
00:27:32
out the sitting
00:27:34
Administration and so what can other
00:27:36
governments do they can do what they're
00:27:38
doing so
00:27:40
Nick I I love this like explanation in
00:27:43
three charts Nick chart number one
00:27:45
please what is this so this is a core
00:27:49
driver of inflation which is the price
00:27:51
of energy and what you see here is that
00:27:54
since the beginning of the year oil
00:27:57
prices have gone up which ultimately
00:27:59
will translate to into a higher cost of
00:28:02
doing business right running your
00:28:03
factory keeping the lights on keeping
00:28:06
your home warm getting from point A to
00:28:09
point B all of these things go up which
00:28:11
that ultimately get reflected in prices
00:28:13
which is what inflation is and so that's
00:28:15
a very bad Trend obviously we had a lot
00:28:18
of really good stuff happen through the
00:28:20
course of last year but it's largely
00:28:22
reversed itself so it's important to
00:28:24
double click into this and figure out
00:28:26
well who's driving this is it just a
00:28:29
random act or is it a systemic set of
00:28:31
decisions and it turns out it's systemic
00:28:34
so if you go to the next chart what this
00:28:36
shows you is what the US has been doing
00:28:39
so the United States has clearly created
00:28:42
an incentive to flood the economy with
00:28:45
oil which the laws of supply and demand
00:28:48
would say prices should go down and by
00:28:52
bringing prices further down they would
00:28:54
have further brought down inflation and
00:28:56
they would have pulled forward these r
00:28:58
cuts that we had been expecting from the
00:29:00
fed the problem is that the rest of the
00:29:02
world actually said no hold on and
00:29:05
that's what you can see on the next
00:29:06
chart so for every time we were trying
00:29:08
to increase production all of these
00:29:11
other countries have been cutting
00:29:13
production and this is the change in the
00:29:15
OPEC plus crude production targets of
00:29:17
2024 so what you see here is for every
00:29:21
barrel of oil the US would pump out of
00:29:23
the ground more than that one barrel of
00:29:25
oil would get actually restricted by all
00:29:27
of these countries on this list which is
00:29:29
just to show you that these countries
00:29:30
have a very specific view about how they
00:29:32
want that energy Market as and that's
00:29:35
just one market but it's an input a key
00:29:38
input into inflation and rates to work
00:29:41
so you're saying that the the OPEC has a
00:29:44
price per barrel price Target and
00:29:46
they're able to manipulate production to
00:29:48
maintain that price Target so while the
00:29:49
US is trying to lower the price per
00:29:51
barrel these other countries are are
00:29:53
able to have a stronger influence and
00:29:55
balance out production to get when you
00:29:57
when you make these Cuts Like This
00:29:58
prices will go up because OPEC plus is
00:30:00
meaningfully bigger the sum of all the
00:30:02
parts of OPEC plus is meaningfully
00:30:04
bigger than just the United States on
00:30:05
its own yeah and so the United States
00:30:08
May amp up a few hundred thousand
00:30:11
barrels a day but if OPAC cuts by a few
00:30:14
million barrels a day you see what you
00:30:16
saw in that first chart which is the
00:30:17
price will go up and that's largely why
00:30:19
prices have gone up so I think why this
00:30:21
happens is because when they see the
00:30:23
United States trying to manage inflation
00:30:27
and manage Global rates right because us
00:30:29
rates are not just for the United States
00:30:31
they're the they're the trends Setter
00:30:32
for everything else I think that in part
00:30:35
there's a decision that this rate
00:30:37
philosophy the economic philosophy the
00:30:39
political philosophy if they wanted to
00:30:42
support that what they would do is
00:30:44
actually also keep prices where they
00:30:45
were or increase production but by
00:30:48
reigning in production what you're
00:30:49
essentially doing is voting against that
00:30:51
entire apparatus right so OPEC and all
00:30:55
of those countries and those decisions
00:30:56
are not made by an oil Minister let's be
00:30:58
they're made by the leaders of those
00:30:59
countries whose decision then gets
00:31:01
reflected by the minister of petroleum
00:31:04
or oil in all of these countries is a
00:31:06
vote against Biden and the United States
00:31:09
government and the politicization
00:31:10
potentially of the fed and then the last
00:31:12
comment is just that when you look at
00:31:14
the reverse repo rate which is a
00:31:15
fundamental measure of liquidity what
00:31:17
you also see is liquidity very quickly
00:31:20
getting sucked out of the system which
00:31:22
is also generally a system health check
00:31:25
that the FED uses to make sure that the
00:31:26
system is operating as required so in
00:31:29
totality what I see is that there is a
00:31:35
structural disillusionment with the
00:31:37
current Administration and where people
00:31:41
can make decisions in their own best
00:31:43
interests versus the collective
00:31:45
interests of the United States in a
00:31:47
broader framework that the US may
00:31:49
represent are voting against it and so
00:31:52
this is why the idea of a persistent
00:31:54
inflation rate is a lot more credible
00:31:57
than it was six months ago and I and I
00:32:01
think sorry just the last point and I
00:32:02
think it's just what sa says which means
00:32:04
that we're probably now unbalance 5050
00:32:06
between a hike and a cut and I think if
00:32:09
you don't see this thing change in the
00:32:10
next three months you're going to see
00:32:12
7525 for a hike of at least 25 basis
00:32:15
points and I do think David's right I
00:32:18
don't see how the sitting Administration
00:32:20
in any
00:32:21
government will be able to withstand the
00:32:24
onslaught
00:32:25
of electoral displeasure
00:32:28
if rates are four and a half five and 6%
00:32:31
going into November Toth if that is I
00:32:34
think everything you said makes sense
00:32:35
going into this election rates are high
00:32:39
and Andor going higher what does the
00:32:42
Biden Administration do the fact is they
00:32:45
are very likely going to come up with a
00:32:48
strategy to try and win votes that will
00:32:51
likely involve some sort of stimulatory
00:32:53
process would be my guess and that
00:32:56
stimulatory process may look something
00:32:58
like more student loan forgiveness more
00:33:00
debt
00:33:02
forgiveness she had a whole speech about
00:33:04
debt forgiveness and here's here's the
00:33:06
irony the actions that the
00:33:08
administration is likely to take to
00:33:11
support themselves in the election cycle
00:33:14
are likely going to further fuel
00:33:16
inflation and that's the biggest concern
00:33:18
and worry I have is that there's almost
00:33:20
this inevitable set of behaviors that
00:33:22
are coming down the pipe here over the
00:33:24
next few months because of the inflation
00:33:25
report that are actually going to make
00:33:27
things worse not better I mean to to
00:33:29
your point Biden just announced a new
00:33:31
student loan forgiveness plan remember
00:33:32
the Supreme Court invalidated his last
00:33:34
one so now he's trying again and it
00:33:36
would potentially give loan forgiveness
00:33:38
to something like 30 million borrowers
00:33:40
so yeah they're going to try and buy
00:33:42
votes but I think it's already very late
00:33:44
in the game to be trying to do that I
00:33:46
think that's what they've been doing
00:33:47
over the past few years I mean Biden has
00:33:50
been a huge spender you know he passed
00:33:52
the the covid relief bill he passed the
00:33:54
inflation reduction act which was a
00:33:56
ridiculous name for was it 750 billion
00:33:59
of of more
00:34:00
spending and uh there was the chips act
00:34:04
which yeah and there was the chips Act
00:34:06
and the infrastructure Bill and then
00:34:08
remember he wanted four and a half
00:34:09
trillion of build back B on top of that
00:34:12
and he wanted even more he didn't get it
00:34:13
any event I think that this is a classic
00:34:16
case of Biden spent the last three years
00:34:19
making his bed and now he's going to
00:34:20
have to sleep in it and if in fact
00:34:23
inflation is persistent and sticky and
00:34:26
not going away and he doesn't get rate
00:34:27
cuts
00:34:28
again this may cost him the election but
00:34:30
but look I want to move beyond the
00:34:31
political aspect of this and just talk
00:34:33
about some more of these KnockOn effects
00:34:35
because we talked about the impact on
00:34:37
the consumer there's also an impact on
00:34:40
investors there's an impact on the
00:34:42
government and there's an impact on on
00:34:44
the banks so let's just deal with the
00:34:46
banks for a second we've talked about on
00:34:49
the show how commercial real estate
00:34:51
developers are hanging on by their
00:34:53
fingernails right now you know they they
00:34:56
basically borrowed huge amounts of money
00:34:58
to buy commercial real estate when rates
00:35:00
were really low and they did so at high
00:35:02
valuations now there's been a huge
00:35:04
correction on that market and they
00:35:05
cannot pay back those loans and they
00:35:08
can't get refinanced so what's happening
00:35:10
well they're working with the banks to
00:35:12
restructure those loans to tack on the
00:35:14
interest they can't pay now to the end
00:35:17
of the loan and then the bank will hope
00:35:19
and pretend that they'll be able to get
00:35:21
paid back one day it's called extend and
00:35:23
pretend well the extend and pretend
00:35:25
strategy could work if you believe that
00:35:27
r cuts are coming really soon because
00:35:29
you just need to hold on until the rate
00:35:31
cuts are here and then you can get
00:35:32
refinanced at lower rates but if rates
00:35:35
are going to be higher longer that
00:35:36
strategy is not going to work and you're
00:35:38
going to see more and more real estate
00:35:39
sponsors throwing in the towel and banks
00:35:42
are going to end up foreclosing on these
00:35:44
properties and you're going to see more
00:35:45
and more fire sales there's a really
00:35:46
dramatic example just the other day in
00:35:48
St Louis a building that was bought a
00:35:51
few years ago for $200 million just fire
00:35:53
sailed at $2 million I mean 99%
00:35:56
reduction this is like like crash level
00:35:59
that's for the equity right yeah no this
00:36:01
is like the price of the deal I mean I
00:36:03
asse the equity was completely wiped out
00:36:06
yeah I guess they sold it for three and
00:36:07
a half million are you sure that's not
00:36:08
the equity that's got to be the equity
00:36:10
there's no way that's the no it's just
00:36:11
the price of the
00:36:13
building okay I don't know I mean with
00:36:15
the debt it's got to have the debt
00:36:16
attach way look I mean either way it's a
00:36:18
total wipe out for the equity holders
00:36:20
yeah but it's also going to be a wipe
00:36:22
out for the bank that made that loan
00:36:24
originally yeah or that bought that loan
00:36:26
when is packaged into cmbs or whatever
00:36:29
it was so I would expect that if rates
00:36:32
stay higher longer that's going to
00:36:34
create tremendous stress on Commercial
00:36:35
Real Estate and and therefore on the re
00:36:38
Regional banking system that made all
00:36:39
these loans to commercial real estate
00:36:42
developers so that's knock on effect
00:36:44
number one knock on effect number two is
00:36:46
is the government's cost of borrowing
00:36:48
the 10e rate's already risen to 4 and a
00:36:50
half% now and if rates stay higher
00:36:53
longer as more and more of the
00:36:54
government debt rolls it's going to end
00:36:56
up getting refinanced at higher rat
00:36:58
and the need to refinance all this
00:36:59
government debt is going to continue to
00:37:01
put pressure on the bond markets and I
00:37:03
think this is why Larry is fundamentally
00:37:05
pretty bearish on rates going down is
00:37:08
the government's financing costs are so
00:37:10
huge and they continue to grow we're
00:37:12
adding something like a trillion dollars
00:37:14
of new debt every hundred days there
00:37:16
have been some reports of this yeah 7.6
00:37:18
trillion is getting refinanced of old
00:37:21
debt in the next 12 months right 7.6 is
00:37:26
sitting around 2%
00:37:28
and now it's going to get bumped up to
00:37:29
close to 5% right which adds which adds
00:37:32
an incremental $210 billion a year of
00:37:36
Debt Service expense just on the debt
00:37:39
that's getting refinanced in the next 12
00:37:40
months not to mention the deficit not to
00:37:42
mention etc etc so right so the
00:37:45
government was extending and pretending
00:37:47
as well I mean they were hoping that
00:37:49
rates would go down fast enough that
00:37:52
when all of those massive government
00:37:53
debt rolls and needs to be refinanced
00:37:55
it' be done at lower rates so the
00:37:57
interest costs wouldn't swamp us we're
00:37:59
already I think at over a trillion
00:38:00
dollars now of annual interest expense
00:38:03
for the federal government we are yeah
00:38:04
and it's going to get it's going to rise
00:38:05
to 1.6 trillion so it's going to be 60%
00:38:09
greater than our defense budget just
00:38:10
servicing the old debt we're basically
00:38:12
in a classic debt spiral where we're
00:38:14
borrowing our annual deficit is we're at
00:38:17
2 trillion plus now so we're not paying
00:38:20
back the principle on the debt we're not
00:38:22
even paying back the interest on the
00:38:24
debt we're borrowing money to pay off
00:38:26
all the interest
00:38:28
and then some on top of that so that's a
00:38:31
huge problem for the federal government
00:38:32
and then I think the the last piece of
00:38:34
this is investors if you pull up this
00:38:37
Fred
00:38:38
chart on the FED funds rate over the
00:38:41
very very long term so what this chart
00:38:43
shows if you just zoom in on this is the
00:38:45
Fed funds rate basically over the last I
00:38:48
don't know like 75 years and you can see
00:38:51
that it peaked in the early 1980s when
00:38:55
vulker decided to break the back of
00:38:57
inflation by jacking up interest rates
00:38:59
to like 18% and it was very painful we
00:39:02
had a pretty severe recession and I
00:39:05
think it was 81 or 82 but then the
00:39:08
economy came roaring back by 83 and
00:39:10
Reagan got elected in ' 84 in any event
00:39:12
that began a period of declining and low
00:39:16
interest rates for investors again
00:39:19
starting in the mid 80s and it continued
00:39:23
all the way through I mean especially
00:39:25
through I think the post GFC period and
00:39:28
then this Co period where we had zero
00:39:30
interest rates for a long period of time
00:39:32
and you can see the last Spike we've had
00:39:34
was when the FED decided to jack up
00:39:37
rates to 5% in response to inflation so
00:39:39
the fact that we've got this persistent
00:39:41
sticky inflation suggests that this
00:39:44
whole you could call it really a almost
00:39:46
a 40-year period of declining in low
00:39:49
interest rates maybe over and when
00:39:51
interest rates are going down it creates
00:39:53
a huge Tailwind to the stock market and
00:39:57
the bond market because the discount
00:40:00
rate is lower so stocks and bonds are
00:40:02
going to be more valuable well if we're
00:40:04
in a in a long-term environment that's
00:40:07
more inflationary and rates have to stay
00:40:09
higher longer that's not great news for
00:40:12
investors it hurts the stock market and
00:40:14
it creates in in the private investing
00:40:17
world like VC and real estate it creates
00:40:19
a much higher hurdle rate so it's harder
00:40:22
for for private investors to justify
00:40:25
fundraising so you know I think think
00:40:27
it's too early to say that this 40-year
00:40:30
trend is over but I think it could be
00:40:33
and um and and the reason for that is
00:40:35
just the huge pressure of government
00:40:38
borrowing crowding out private investing
00:40:40
and this sort of everpresent need to
00:40:44
keep issuing more and more debt at
00:40:46
higher and higher rates to make it more
00:40:48
attractive is going to create I think
00:40:51
just very bad conditions for for
00:40:53
investors so Chim let me ask you if you
00:40:56
concur
00:40:57
given that today where we're sitting as
00:40:59
of recording right now towards the end
00:41:02
of market on Thursday April 11th the
00:41:04
NASDAQ is sitting at an all-time high
00:41:07
and the NASDAQ is up 1.6% today
00:41:10
following this inflation report
00:41:12
yesterday what's the disconnect in the
00:41:14
market on why investors are buying
00:41:18
equities given that we're likely going
00:41:21
to have persistent higher rates and as a
00:41:24
result multiples should compress
00:41:26
particularly in growth stocks
00:41:28
and why the nasdaq's kind of firing
00:41:31
up I think it's important to look at
00:41:34
what stocks are actually lifting the
00:41:35
NASDAQ just like what stocks are lifting
00:41:37
the S&P 500 and we've seen this there's
00:41:40
a massive dispersion there's like seven
00:41:42
or eight companies that matter and 500
00:41:44
odd companies that don't matter
00:41:46
much and so the reason why things go up
00:41:49
in an environment like that is
00:41:51
irrespective of the economy money
00:41:53
managers are paid to deploy money and I
00:41:56
don't know if you guys saw this but Nick
00:41:58
maybe you can find it I linked to it in
00:41:59
my newsletter last Sunday which was an
00:42:02
article about the fact that the global
00:42:04
amount of total available equities to
00:42:07
buy is
00:42:08
shrinking right so if you think about
00:42:11
that what that means is that the number
00:42:13
of dollars is growing right linearly and
00:42:16
in some years it seems exponentially but
00:42:19
then the and so then a certain
00:42:21
percentage of that Finds Its way into
00:42:23
the equity markets but when the equity
00:42:25
markets continue to shrink and there's
00:42:26
few and fewer things to buy prices will
00:42:30
still go up irrespective of the
00:42:31
underlying justification so I think that
00:42:34
the the thing is you can't look at these
00:42:36
things as a sign per se you know if you
00:42:38
look at Commodities as a different
00:42:40
example Nick I sent you
00:42:42
this Commodities are up broadly speaking
00:42:45
30% since the beginning of the
00:42:47
year the I think the reality is the
00:42:51
following countries are feeling
00:42:55
increasingly uncomfortable
00:42:58
with the decisions that the United
00:43:00
States are making whether that's its
00:43:02
position to keep funding the war in
00:43:04
Ukraine whether that it's the continued
00:43:07
instability in the Middle East whether
00:43:09
it's posturing with China whatever it is
00:43:12
and so every country is now making a
00:43:14
very different calculation about what
00:43:16
they want to do and I think how it gets
00:43:17
reflected is that they're optimizing for
00:43:19
their own economies and what that yields
00:43:22
is each of them trying to make as much
00:43:24
money as they can in the short term and
00:43:26
as as a country that is a net importer
00:43:30
we are going to feel the price of those
00:43:32
decisions as inflation and I think saaks
00:43:34
is right like that's where we have to
00:43:36
start figuring out how we rebalance
00:43:39
these decisions from a foreign policy
00:43:41
and economic perspective because if you
00:43:43
don't change something here all of these
00:43:45
countries are basically going to be like
00:43:46
we're on our own we're going to optimize
00:43:48
for ourself they're these countries are
00:43:50
basically voting in a way that says we
00:43:53
are not being led in a way that we trust
00:43:55
and that unfortunately Falls at the feet
00:43:56
of the sitting president yeah well look
00:44:00
I'm
00:44:01
hopeful as we all are that work in the
00:44:04
tech industry that new tools and Ai and
00:44:08
other Technologies would have a massive
00:44:10
driver on productivity which netn net
00:44:13
could help even this out that's the one
00:44:15
last great hope sitting here AI
00:44:19
obviously viewed as being highly
00:44:20
disruptive by non- technologists and
00:44:23
deeply concerning and particularly in DC
00:44:26
what is but what is that thing that Bill
00:44:28
Gates says things are overestimated in
00:44:30
the short term and deeply underestimated
00:44:32
in the long term I think that that's
00:44:34
that's what AI that's right that's right
00:44:36
and so AI is a bunch of no offense
00:44:39
probabilities and statistics today okay
00:44:42
and so we talked about this last time so
00:44:44
we should debunk this whole thing of
00:44:45
like there's some glowing brain
00:44:47
somewhere that's capable of doing
00:44:49
everything doesn't mean that we
00:44:51
shouldn't underestimate it being capable
00:44:54
of this in five or 10 years which is
00:44:56
probably the window in which we have
00:44:58
some form of AGI but today it's not that
00:45:00
and so the the great leaps in
00:45:02
productivity freeberg that you're
00:45:03
talking about are probably a 3 to five
00:45:05
year phenomenon not a three to five
00:45:06
month so what saak says is Right which
00:45:08
is how does the sitting Administration
00:45:12
buy votes today stimulatory be it'll
00:45:14
have to make a ton of promises the
00:45:16
problem is it can't really work its way
00:45:18
into the economy fast enough come the
00:45:20
November election and so that that's
00:45:23
where I think they're in a really tough
00:45:25
spot which means that then
00:45:28
and and I hate to B it bear or bad news
00:45:29
but this is where you have a lot of
00:45:31
these Wag the Dog kind of scenarios that
00:45:34
come to mind because then the only way
00:45:36
to really Galvanize a populace is to
00:45:39
distract them yeah well look let's talk
00:45:41
a little bit more about AI because I do
00:45:44
think that regardless of the near-term
00:45:48
and longer term implications there is a
00:45:51
peing of fear peing of Interest both
00:45:54
different spelling of the word Peak Adam
00:45:57
shiff yesterday proposed a bill sorry
00:46:00
earlier this week called the generative
00:46:01
AI copyright disclosure act in response
00:46:04
to I think over a hundred musicians
00:46:07
signing a letter saying they're
00:46:09
concerned that their copyrighted works
00:46:12
are being used to train AI models and
00:46:15
Nick if you'll pull up the text from the
00:46:16
bill that I pulled out here real quick
00:46:19
we can take just a quick look it's a
00:46:21
very short bill it's only five pages
00:46:23
long and it defines a generative AI
00:46:25
model uh that's a combination of
00:46:28
computer code and numerical values
00:46:30
designed to use artificial intelligence
00:46:32
to generate outputs in the form of text
00:46:35
images audio or video and that it
00:46:38
substantially uh incorporates one or
00:46:40
more models as a system and it's
00:46:42
designed for use by consumers and what
00:46:45
the bill then asks is that anyone
00:46:47
developing these models has to submit to
00:46:50
a register run by the federal government
00:46:54
a list of all of the data that they used
00:46:57
that might be copyrighted to train the
00:46:58
model and that this becomes a precursor
00:47:02
to being able to legally develop and use
00:47:04
AI models that you actually have to
00:47:06
register the training data so I have a
00:47:08
rant on this which I'm going to preserve
00:47:10
for the end given my rights as moderator
00:47:13
here today but I'd love the reaction
00:47:15
from saaks first on shift's proposed
00:47:17
Bill Jam moth and we'll run this one
00:47:19
around the horn pretty quick well I
00:47:21
first of all I think it's just it's too
00:47:23
soon for this legislation I mean the the
00:47:25
first thing that needs to happen is we
00:47:26
to get the question of fair use
00:47:28
arbitrated by the courts we've talked
00:47:30
about this before people publish lots of
00:47:33
information on the internet and I guess
00:47:36
technically it's copyrighted but it's
00:47:38
been published in a completely public
00:47:40
way and humans are obviously allowed to
00:47:43
read it and and use it to generate their
00:47:45
own work I think there's a question
00:47:46
about whether AIS are allowed to learn
00:47:49
from it as well and I would argue that
00:47:52
they should be to some degree under fair
00:47:54
use maybe if you're talking about
00:47:56
copyright write of songs I don't know if
00:47:57
that's different anyway these questions
00:47:59
need to be arbitrated and once we know
00:48:02
that then we'll know whether there's a
00:48:04
need for a rights Clearing House of some
00:48:07
kind which may not be a bad idea I
00:48:09
actually got pitched by a Founder
00:48:11
recently to create a rights marketplace
00:48:16
with copyright holders on the on the one
00:48:19
side or rights holders could be
00:48:20
musicians could be movie studios could
00:48:23
be anyone who owns content and then AI
00:48:27
on the other that want to license that
00:48:28
content for training purposes and I
00:48:31
actually thought it was a pretty good
00:48:32
idea
00:48:33
so my point is just that there are
00:48:36
entrepreneurs working on solving this
00:48:38
problem as well and I guess what I'm
00:48:40
saying is I'd like to see fair use
00:48:42
figured out and i' like to see what the
00:48:44
private Solutions are going to be before
00:48:46
we try to legislate this problem away
00:48:49
from really a point of ignorance because
00:48:52
we're so early in the development of
00:48:53
this market now I think one of the
00:48:55
questions you have to ask here is why
00:48:57
did Adam Schiff introduce this bill and
00:49:00
who is Adam Schiff Adam Schiff is a
00:49:02
powerful California congressman who is
00:49:05
the candidate for US Senate right now
00:49:08
and he's expected to win he's he's ahead
00:49:10
of Steve Garvey who's the Republican
00:49:11
candidate by like 30
00:49:14
points Schiff is in addition to being a
00:49:18
vicious partisan who's known for for
00:49:20
rushia gate he also is a prodigious
00:49:23
fundraiser and when he's not in front of
00:49:26
the camera he's very effective in closed
00:49:29
doors appealing to the key special
00:49:31
interests in California that matter and
00:49:34
my guess is that he's not carrying water
00:49:36
here for the hundred musicians who
00:49:37
signed that letter he's carrying water
00:49:39
for the big studios the big Hollywood
00:49:41
Studios including music labels but also
00:49:44
the movie studios and he's an LA based
00:49:47
Congressman yeah I mean that's who he's
00:49:49
representing now at the same time my
00:49:51
guess is that he's run this by the big
00:49:54
tech companies and he's probably got
00:49:55
their blessing some degree because like
00:49:57
Bill Gurley always points out they like
00:50:00
rules like this because of regulatory
00:50:02
capture they can comply with them but
00:50:04
the little guy the entrepreneur who's
00:50:06
trying to create something new has a
00:50:07
much harder time complying with these
00:50:09
regulations so this is basically a big
00:50:12
special interest group sort of deal that
00:50:16
I think he's he's putting forward do you
00:50:18
think this becomes law not yet but it's
00:50:21
Paving the way for something and it's
00:50:23
more like his promotional effort during
00:50:25
his campaign effectively well I
00:50:28
mean his they'll still try to move it
00:50:30
legislatively but yeah I mean the the
00:50:33
big La powers that be like this and they
00:50:37
love this yeah
00:50:40
Jam any reaction to this bill I
00:50:44
think there was an article in The Verge
00:50:47
that I
00:50:48
posted it turns out that open AI
00:50:50
transcribed over a million hours of
00:50:52
YouTube videos to train
00:50:55
gp4 my take away when I read this was
00:50:58
well if Google's not going to sue open
00:51:00
AI then this is a mood
00:51:04
point because you're talking about one
00:51:06
of the most valuable companies in the
00:51:08
world with one of the most valuable
00:51:10
sources of training data in the
00:51:12
world and so if if that if it's true
00:51:15
that there was a copyright violation and
00:51:17
Google just lets us go what's anybody
00:51:20
else supposed to do what chance does
00:51:22
these group of singer songwriters have
00:51:24
and you know no offense to Adam shift
00:51:26
but he's not going to be able to do
00:51:28
anything so if this multi-trillion
00:51:31
dollar company doesn't draw a line in
00:51:33
the sand then none of this matters and
00:51:36
everybody's going to train on everything
00:51:38
I think like my big Takeaway on all this
00:51:42
is that there's a real deep
00:51:48
misunderstanding on what
00:51:52
these models really are and what they're
00:51:55
doing the assump
00:51:57
I think
00:51:58
naively is that they take in a buch of
00:52:01
data and then they spit that data back
00:52:05
out and in some cases maybe slightly
00:52:08
transformed but the truth is that the
00:52:10
models don't actually hold the data that
00:52:14
they're trained on they develop a bunch
00:52:17
of synthesized predictors that they
00:52:20
learn what the prediction could or
00:52:23
should be from the data and this is
00:52:25
similar to how humans operate
00:52:27
can you pull up the article Nick that I
00:52:30
posted this article is from last May and
00:52:33
you guys may remember this case I can't
00:52:35
remember if we talked about it on the
00:52:37
show but Ed Sheran was
00:52:39
sued by the estate of the writer of uh
00:52:44
the Marvin Gay song Let's Get It On for
00:52:48
infringement in his song Thinking Out
00:52:50
Loud and he actually prevailed in court
00:52:53
where he went through with the jury how
00:52:56
he comes up how he runs his creative
00:52:58
process how he came up with the song and
00:53:01
how he does his work and as you know Ed
00:53:04
Sheron and nearly every musician or
00:53:06
every artist listens to other people's
00:53:08
music looks at other people's art and
00:53:11
they synthesize that data they
00:53:13
synthesize that knowledge and that
00:53:14
experience into their own creative
00:53:17
process to
00:53:19
Output what in their mind is a novel
00:53:22
creation and I think that AI Works in a
00:53:25
similar vein in that it's not actually
00:53:27
storing the images it's not actually
00:53:29
storing the music of third parties it's
00:53:32
learning from that process we all like
00:53:35
musicians learn on classical music they
00:53:37
listen and play other pop artists music
00:53:40
and then they learn things that they
00:53:41
like things that they don't like things
00:53:43
that fit together well certain
00:53:46
intonations syncopations and that's how
00:53:48
they develop their own tracks and so I
00:53:50
think that the um assumption in AI is
00:53:53
that it's almost guilty until proven in
00:53:56
inent is what's going to really
00:53:58
challenge this technology getting
00:54:00
Mainstay appeal and being useful uh in
00:54:03
the way that it could and should be that
00:54:05
the de facto is that it's learning and
00:54:07
training is copyright infringement I
00:54:11
don't buy that you that stores the data
00:54:15
no that that just because you're
00:54:17
learning like a human it's not copyright
00:54:19
infringement and I'm sorry let me let me
00:54:21
make my last statement which is that I
00:54:23
do think that similarity is measurable
00:54:25
and that is copyright infringement so I
00:54:27
would rather than focus on how you train
00:54:29
the model I would encourage legislators
00:54:32
to find as Saks points out a better
00:54:35
process to Define what makes one thing
00:54:37
distinct from another run that analysis
00:54:41
and then use that analysis to determine
00:54:43
uniqueness and that analysis is
00:54:45
ultimately what should determine whether
00:54:47
there is copyright infringement versus
00:54:49
the how was the model trained exercise
00:54:52
which is rought with challenges as you
00:54:53
guys know how do you determine
00:54:55
weightings how do you determine whether
00:54:56
this was even relevant to the model we
00:54:58
don't even know if the models ignore the
00:55:00
data or use the data in their synthesis
00:55:02
or in their waiting it's all very opaque
00:55:05
and so it's almost like hey here's a
00:55:06
list of stuff that I read but I may not
00:55:08
have even paid attention to 95% of it
00:55:11
when I was reading it we don't know and
00:55:13
so I'd rather focus on is there
00:55:15
copyright infringement in the output
00:55:16
does the output mimic an original
00:55:19
copyrighted source and if it does then
00:55:21
that's where the infringement should lie
00:55:24
not on the how did you do it I think
00:55:25
that you're creating an impossible
00:55:27
standard is it 5% the same is it 8% the
00:55:29
same is that 14% where's your threshold
00:55:32
is it how that different than training
00:55:34
how's that different than training if
00:55:35
I'm JK rling am I not allowed to read
00:55:37
other authors copyrighted books for fear
00:55:40
that by reading it I will now use their
00:55:41
copyrighted material in my novel
00:55:44
book okay stick with that example but we
00:55:46
can also use the Edge shear an example
00:55:48
Nick can you play this video I could
00:55:49
play you every song in the pop charts
00:55:51
right now over four chords you name any
00:55:53
song and I'll play it in the four chords
00:55:55
so these These are CS fcgn messenger um
00:55:59
that's your friend um uh cuz you only
00:56:02
need the light when it's burning low
00:56:04
only miss when its starts only know you
00:56:07
love when you let her go I'm walking
00:56:10
away CRA David uh C David I'm walk
00:56:14
away the in last one Let It Be Beatles
00:56:19
when
00:56:23
I why is that important obviously it's
00:56:27
not illegal for Ed to have learned these
00:56:29
four chords but if he tries to produce
00:56:31
those songs and calls them something
00:56:33
else that's clearly somebody else's IP
00:56:36
similarly I think that if you don't give
00:56:39
people the right to say yeah of course
00:56:41
the letters AB c d e FG all the way to Z
00:56:44
are not my invention it's the alphabet I
00:56:46
don't control that I don't control words
00:56:47
and their meanings but when I put one
00:56:49
word in front of another in front of
00:56:51
another purposefully and I create
00:56:53
something that has value to me I should
00:56:55
have the right to protect protect that
00:56:57
and I think that's the fundamental
00:56:58
question should you not have a right to
00:57:00
protect it should you allow somebody
00:57:02
else to take that and in some opaque way
00:57:06
and undefinable way make a system that
00:57:09
could compete with that thing and have
00:57:11
no repercussions I think that that's
00:57:13
what people need to decide and I'm fine
00:57:15
by the way one way or the other yeah in
00:57:17
the Ed Sheran case he was saying the
00:57:19
exact copyrighted lyrics and he was
00:57:22
singing the exact copyrighted Melody he
00:57:24
was just playing the harmony with four
00:57:25
chord I that was kind of the point he
00:57:28
Wasing in the case of all the the
00:57:31
copyrighted material you're stating I
00:57:33
want to protect it from showing up in
00:57:34
the end product of someone's creative
00:57:37
process I shouldn't care about them
00:57:39
listening to my song in their creative
00:57:40
process everyone always references how
00:57:42
Oasis sounds like the Beatles right they
00:57:44
those guys love the Beatles they listen
00:57:45
to The Beatles so much of their music
00:57:47
mimics and and sounds like the Beatles
00:57:49
but the Beatles aren't going to file
00:57:50
infringement suits against them they
00:57:51
were inspired by The Beatles that was
00:57:53
part of their creative process and so I
00:57:55
would argue that the output should
00:57:56
really be where we assess this stuff and
00:57:58
it's going to take a lot of cases you're
00:58:00
creating an impossible standard where no
00:58:03
any individual person will have the
00:58:05
economic wherewithal to fight this and
00:58:07
so it's basically seeding all creativity
00:58:11
to the board and I don't think what is
00:58:14
what is the standard of measuring how
00:58:16
someone is training their creative
00:58:17
process your standard is I gotta go I
00:58:19
don't think I got to go listen I got to
00:58:21
go you got to tell me every time you as
00:58:22
an artist listen to a song and then I'm
00:58:25
going to deter determine whether or not
00:58:27
you are listening to copyright music and
00:58:29
making your own music that's basically
00:58:31
what we're asking for here with this
00:58:33
bill that's not I don't I'm not talking
00:58:35
about the bill I'm talking about this
00:58:36
idea that if you train a system that can
00:58:39
actually automate a task for you based
00:58:41
on some learning That's Unique you
00:58:43
should be able to uniquely protect your
00:58:46
ownership for example we have a
00:58:48
different set of ways in order to
00:58:50
protect processes that are unique to you
00:58:52
which is you can have a patent or you
00:58:54
can have a trade secret but if you R all
00:58:56
of this stuff away where some system can
00:58:59
in an automated way do anything for you
00:59:02
and replicate a whole bunch of work that
00:59:04
it sits on top of to be able to do that
00:59:07
without any repercussions I think that
00:59:09
that's a very important explicit
00:59:10
decision and what I'm saying is you
00:59:12
cannot expect individual people to go
00:59:14
and fight these little battles against
00:59:16
these big multi hundred billion dollar
00:59:18
companies if the big multi hundred
00:59:20
billion dollar companies and trillion
00:59:22
dollar companies don't fight the other
00:59:24
multi hundred billion dollar and
00:59:25
trillion dollar companies what they're
00:59:27
effectively saying isone of this is
00:59:29
worth anything copyrights are worth
00:59:30
nothing and I'm okay with that decision
00:59:33
but it should be an explicit one and
00:59:34
it's not going to happen the other way
00:59:36
where some five-page law from a guy
00:59:40
who's trying to get elected is not the
00:59:41
thing that's going to make a the
00:59:43
difference the difference is if Google
00:59:45
doesn't Sue open AI it means copyright
00:59:47
is worthless and by the way that article
00:59:49
didn't just point the finger at Google
00:59:51
it pointed the finger at meta where
00:59:53
there were explicit decisions to say
00:59:54
yeah no let's just violate the copyright
00:59:56
it's it's okay and if it's okay then
00:59:58
it's okay sax you're the independent
01:00:02
Arbiter here any summary well somebody's
01:00:06
going to litigate it somebody's litig
01:00:08
GNA litigate it you're you're right it'd
01:00:09
be surprising if Google didn't want to
01:00:12
given how valuable all their YouTube
01:00:13
content is and a competitor you know
01:00:15
slurped it up to train their model but
01:00:18
somebody will will litigate this it
01:00:20
could be these songwriters who knows and
01:00:22
then we're going to we're going to get
01:00:24
some messy arbitration fair use and it's
01:00:26
probably going to work its way up over
01:00:28
the different circuit courts you'll
01:00:30
probably get different judgments and
01:00:32
finally the Supreme Court will resolve
01:00:33
it and then we'll kind of know where
01:00:35
things stand and then and then there'll
01:00:37
be a legislative
01:00:39
fix I I don't know where it's going to
01:00:41
end up I mean I know what Jal would say
01:00:42
if he were here I'm sure he's like
01:00:44
chomping at the bit be part of this
01:00:45
conversation he would say that creators
01:00:48
deserve to get paid for their AI rights
01:00:52
and that if you want to make money
01:00:53
training a model on their cont you got
01:00:56
to compensate them I also I I I buy I
01:00:58
buy your argument too which
01:01:01
is you're not really violating their
01:01:04
their rights for the AI model to read
01:01:08
their content and analyze it which is
01:01:10
what it's really doing as long as it
01:01:12
doesn't output a violation of their
01:01:15
copyright and that's what all artists do
01:01:18
that's what all humans do the creative
01:01:20
process is one of initially synthesis
01:01:23
you learn from other people's works you
01:01:25
you learn from the environment around
01:01:27
you you learn from the world that you're
01:01:28
exposed to but there was compensation in
01:01:29
those other bottles there was not you
01:01:31
don't oh Oasis never paid the Beatles to
01:01:35
listen to their music they the radio no
01:01:37
that's not true they listened on the
01:01:39
radio they may not have paid but some
01:01:41
Advertiser paid for an ad the radio
01:01:43
station had to pay ASCAP and BMI there
01:01:46
was publishing rights there was all
01:01:47
kinds of other rights so that's not true
01:01:49
what you're saying I'm not what I'm what
01:01:51
I'm saying is it's not copyright just
01:01:52
because I train my process somebody got
01:01:54
paid so that singer songwriter did get
01:01:56
paid they may not have getting paid by
01:01:58
the listener but in this example
01:01:59
nobody's getting paid look I think
01:02:02
there's three potential outcomes here
01:02:03
okay outcome number one is that freeberg
01:02:06
wins and the courts decide that this is
01:02:09
fair use and these AI companies aren't
01:02:11
in violation copyright infringement is
01:02:14
measured in the work of art that is
01:02:16
produced not in the process by which it
01:02:18
is generated right okay so that's
01:02:20
solution number one I think solution or
01:02:23
outcome number two is that the courts
01:02:26
throw a big wrench into it or a bill
01:02:28
like shift bill which is largely
01:02:30
punitive throws a big wrench into it and
01:02:32
I think what happens then is new AI
01:02:35
companies have a really hard time
01:02:37
accumulating training data and so the
01:02:39
incumbents who already have all the
01:02:41
training data they may have to pay a
01:02:43
fine there'll be some slap on the wrist
01:02:45
but now you've created a real moat and
01:02:46
barrier to entry and so secretly they're
01:02:48
going to love it and it's going to be a
01:02:50
big problem I think for the open source
01:02:52
movement so that's outcome number two
01:02:54
and then outcome number three is you
01:02:56
create some sort of Rights Clearing
01:02:59
House kind of like uh exists in the
01:03:02
recording industry where if you want to
01:03:04
cover somebody's song or perform
01:03:06
somebody else's song there's already
01:03:08
predefined terms on which you can use
01:03:10
those rights and then there's a
01:03:12
compensation mechanism for that so you
01:03:14
don't have to go negotiate every single
01:03:16
deal that you want to do you can just
01:03:18
use the song you can use the rights and
01:03:20
there's again compensation for that that
01:03:23
would be like the compromise I could see
01:03:25
that alter
01:03:26
number three existing in which case the
01:03:29
incumbents probably like that because
01:03:31
they again they want there to be a
01:03:33
barrier to entry they want to make it
01:03:35
hard for the guy in the garage to Pony
01:03:38
up the the capital necessary to create a
01:03:40
new model but at least that would be a
01:03:43
somewhat functional system it wouldn't
01:03:44
be outcome number two which would
01:03:47
totally put the kabash on the use of
01:03:48
training data so in any event I don't
01:03:50
know which one of the three it's going
01:03:51
to be but those are the three potential
01:03:53
outcomes that I see Happening Here have
01:03:54
you guys seen UD
01:03:58
udio uh no it's like a new music geni
01:04:03
system it is frakin unbelievable is that
01:04:07
vod's thing or vod's another thing I
01:04:10
think vod's another thing this one just
01:04:11
came out it was apparently rumored for a
01:04:13
while ex deepmind employees started yudo
01:04:17
it's been this kind of like hot stealth
01:04:20
company
01:04:22
for a couple of months people have been
01:04:24
talking about it as rumored to to be the
01:04:26
best music AI but here is someone that
01:04:28
said Dune the Broadway musical and this
01:04:31
is what it outut go ahead of AR the
01:04:35
greatest leader we've ever seen they say
01:04:38
that he's a listen
01:04:42
gu
01:04:44
and onand back that's crazy you can hear
01:04:49
the layers of Music the layers of the
01:04:51
tracks the
01:04:53
smoothness the tempo obviously like
01:04:56
there's just so many layers to this Nick
01:04:57
can you pull up another one some of
01:04:59
these are just incredible because it
01:05:00
spans genres it spans Styles it's like
01:05:04
so beautiful like the way it's put
01:05:06
together every every piece feels
01:05:08
somewhat emotional it's Unique it really
01:05:11
doesn't feel like you're just listening
01:05:13
to someone printing copyright
01:05:17
music funky dance pop
01:05:20
[Music]
01:05:22
song create a birthday song for my
01:05:25
friend Irvin
01:05:31
create an Irish folk tune with violin
01:05:33
and
01:05:35
[Music]
01:05:41
instrumental it's pretty incredible and
01:05:44
I think it shows the
01:05:46
state-of-the-art is such now that this
01:05:49
is going to become a real challenging
01:05:51
question from a legislative point of
01:05:54
view given how far ahead these
01:05:56
technologies have gotten and I think
01:05:58
that musicians artists consumers are
01:06:00
going to start to use these tools in a
01:06:01
really prolific way given how good they
01:06:04
are now I heard from someone that just
01:06:07
sat for hours creating tracks that they
01:06:09
listen to on udio the other day and so
01:06:12
you're basically just playing your own
01:06:13
music all day long it's really
01:06:15
incredible anyway exciting time
01:06:17
hopefully all the lawsuits and the
01:06:19
legislators don't get in the way of all
01:06:20
this Innovation cuz it's moving really
01:06:22
fast and we should just kind of see
01:06:23
where it goes before trying to clamp
01:06:25
down totally mhm guys I want to talk
01:06:29
about as you know this like emerging
01:06:31
Trend that we've been talking about
01:06:32
offline about drones in Warfare recently
01:06:35
we saw small drones controlled by the
01:06:37
houis attack cargo vessels in the Suz
01:06:40
Canal it led to rather large scale
01:06:42
Supply disruptions economic value
01:06:44
destruction we've seen
01:06:46
videos I think you have one we can pull
01:06:48
it up and take a look of Ukrainian
01:06:51
controlled drones flying in to destroy
01:06:53
tanks and large warehouses and Russian
01:06:55
held territories causing tens of
01:06:57
millions of dollars of damage I think
01:06:59
we're seeing the acceleration of the
01:07:00
changing face of warfare technology lots
01:07:03
of these small form factor lowcost
01:07:05
autonomous physical agents and I think
01:07:08
there's a bunch of implications both
01:07:09
with respect to war but also how we
01:07:12
spend money on defense and how Silicon
01:07:15
Valley is involved in response um sax
01:07:18
I'd love to hear your your points of
01:07:19
view maybe you can highlight for us some
01:07:20
of the things you've been seeing in the
01:07:22
utilization of these new technologies in
01:07:24
in the war given you're our in-house War
01:07:27
correspondent and you obviously had do
01:07:29
some work in Silicon Valley go ahead
01:07:31
I've heard a lot about the impact of of
01:07:33
drones just by watching the coverage of
01:07:35
the Ukraine war I was listening to an
01:07:38
interview with an American mercenary
01:07:40
who's fighting on the side of Ukraine
01:07:42
and he described how ubiquitous these
01:07:45
drones were on the battlefield now he
01:07:46
said that you literally can't get out of
01:07:48
a trench to go to the bathroom because a
01:07:51
drone will basically uh get you and he
01:07:53
said they're buzzing around they sound
01:07:55
like mosquitoes because they're kind of
01:07:57
just everywhere on the battlefield both
01:07:59
Ukraine and Russia have them and several
01:08:04
months ago going into this year zalinsky
01:08:06
said that this would be the big game
01:08:07
changer for Ukraine they're going to
01:08:08
make a million drones I don't think it's
01:08:10
worked out that way it turns out that as
01:08:11
many drones as ukrainians have been able
01:08:14
to put on the battlefield the Russians
01:08:15
have even more because they're able to
01:08:17
mass produce them in factories they've
01:08:19
got bigger drones better drones and I
01:08:22
think the most important variable in in
01:08:24
the war so far with respect to drones is
01:08:26
that the Russians have pretty Advanced
01:08:28
electronic warfare and so they've been
01:08:31
able to jam a lot of the Ukrainian
01:08:33
drones whereas the reverse has not been
01:08:35
true for the ukrainians so I would say
01:08:38
that drones have been a huge factor in
01:08:40
the war but so
01:08:42
far the balance there is tipping towards
01:08:45
the Russians as it is in so many other
01:08:47
areas as well but to your larger Point
01:08:50
there's no question this is the the
01:08:51
future of warfare and you're seeing that
01:08:54
it's cre creating a lot of opportunities
01:08:57
for asymmetric Warfare so for example
01:09:00
with the hoodies they've been firing
01:09:03
cheap missiles and drones at our at our
01:09:06
ships in the Red Sea and we've been
01:09:08
having to spend $2 million air defense
01:09:11
missile shooting down $2,000 drones so
01:09:13
if if that continues and we don't have a
01:09:15
good response to this problem it's going
01:09:17
to really change the balance of power I
01:09:19
don't know if this was at our Summit or
01:09:21
if I heard it this later from some
01:09:22
senior person in the military who said
01:09:24
that aircraft car
01:09:25
are outdated technology like they don't
01:09:27
make sense anymore and we're going to
01:09:30
see a shift towards lots of small
01:09:34
autonomous drone like systems on the
01:09:37
battlefield taking out targets and as a
01:09:39
result if you Game Theory this out the
01:09:41
necessity for defense systems against
01:09:44
large amounts of swarming drone like
01:09:46
systems and then you know what do you
01:09:50
actually have to do with your existing
01:09:52
military architecture to play into this
01:09:56
kind of new Tactical system and then
01:09:59
that's going to require a massive
01:10:01
evolution in technology and is the
01:10:04
United States
01:10:05
prepared they should Silicon Valley play
01:10:07
they should be we should I maybe Nick
01:10:11
you can throw up this the first link
01:10:12
that I sent you but eight years ago I
01:10:16
came to this conclusion and I invested
01:10:17
in this company called s drone which is
01:10:19
autonomous drones in the Seas our
01:10:22
customers include the US Navy the
01:10:23
chairman of our company actually is is
01:10:25
Admiral Mike Mullen who is the former
01:10:27
chairman of the Joint Chiefs of Staff
01:10:29
but these drones are capable of going
01:10:32
into some of the most dangerous places
01:10:33
in the world and collecting enormous
01:10:36
amounts of surveillance data and
01:10:38
information that otherwise
01:10:41
takes the United States government lots
01:10:44
of time and lots of money and is pretty
01:10:47
scattershot and instead with these
01:10:48
things you can
01:10:50
be constantly in hotpots and if you go
01:10:54
to the next one so much that you know
01:10:55
we've deployed some of these drones in
01:10:58
the Middle East and there was a point at
01:11:00
which the Iranian Navy intercepted two
01:11:02
of our drones and picked them up and
01:11:03
this was kind of like a a global thing a
01:11:06
few years ago but I really believe in
01:11:08
this trend and I think that we have an
01:11:10
enormous responsibility to be funding
01:11:13
these things these are really
01:11:14
complicated systems to build obviously
01:11:16
and they take lots of time so these are
01:11:18
not overnight successes freedberg and
01:11:20
they take lots of money which is hard to
01:11:22
come by as well but these are absolutely
01:11:24
the right kinds of businesses because
01:11:26
eventually at a minimum you Building
01:11:28
Systems that can measure and collect
01:11:30
enormous amounts of really critical data
01:11:32
so that people can make better decisions
01:11:35
which hopefully is measured in Saving
01:11:38
life right and then over time you
01:11:41
actually get to a military that should
01:11:43
be much cheaper much safer and has fewer
01:11:47
people in general on our side and on our
01:11:49
enemy sides on the front lines which
01:11:52
means fewer casualties and I think that
01:11:54
that's the whole value of this entire
01:11:56
movement it's cheaper it's way better
01:11:59
it's much more useful and it gives the
01:12:03
respective armies that use these things
01:12:05
or militaries rather a fuller picture so
01:12:07
that you can make decisions that have
01:12:09
enormous consequences yeah and in
01:12:11
February the Department of Defense on a
01:12:14
panel in Arlington Virginia briefed
01:12:17
reporters and Industry folks on their
01:12:21
interest in growing their Partnerships
01:12:22
in Silicon Valley to access the
01:12:25
necessary technologies that are going to
01:12:28
rewrite the face of warfare I don't know
01:12:31
about you guys but I've seen kind of two
01:12:34
sides of this in Silicon Valley a good
01:12:36
number of venture capitalists and
01:12:38
entrepreneurs and technologists think
01:12:41
that it's morally incorrect to support
01:12:44
warfare technology and then there are a
01:12:46
few that are going blazing ahead with
01:12:49
supporting this new Evolution and these
01:12:51
Technologies do you guys see the same
01:12:53
and is there going to be kind of this a
01:12:55
big step to weaponize these things right
01:12:57
so s drone we have a very good big
01:13:01
thriving successful business these
01:13:03
machines aren't weaponized by any
01:13:05
stretch of the
01:13:06
imagination and so I think that
01:13:09
that comment is more
01:13:12
genuflecting and virtue signaling than
01:13:14
it is real because in order to even be
01:13:18
legitimately considered for some kind of
01:13:22
like armed
01:13:23
unmanned product you have to be deep
01:13:26
inside of the bowels of the dod and
01:13:29
Pentagon and have been working with them
01:13:30
for years to even be taken seriously so
01:13:33
the startups that are like assing those
01:13:35
deals are not even close to those deals
01:13:38
and the startups that are close to them
01:13:40
are probably being much more quiet and
01:13:42
thoughtful and Hush Hush because the
01:13:44
path to do that isn't legitimate unless
01:13:46
you've been in business with these folks
01:13:48
for 10 plus years yeah I think
01:13:51
it's it's more a commentary on the fact
01:13:53
that if this is where
01:13:56
an organization the Department of
01:13:58
Defense that has a trillion dollar
01:14:00
annual budget is going to be allocating
01:14:03
resources that technology is going to
01:14:05
need to come from somewhere and the
01:14:08
investors in Silicon Valley that are
01:14:11
making these Investments are likely
01:14:12
going to outperform those who are not
01:14:15
does that sound reasonable to you saxs
01:14:17
that there is this shift underway and
01:14:19
that again who have sh shied away from
01:14:21
defense technology are you know
01:14:24
necessarily going to be left out of a
01:14:26
you know kind of a new industry that's
01:14:27
emerging in Silicon Valley well I guess
01:14:29
the way I put it is that the US
01:14:31
government spends over $800 billion
01:14:33
every year on defense and it's not clear
01:14:36
what we're getting for that money
01:14:38
because in Ukraine for example we've run
01:14:40
out of artillery shells we've actually
01:14:41
we've run out of Patriots we've run out
01:14:43
of javelins we've run out of stingers so
01:14:45
we're spending all this money we're not
01:14:46
getting a lot for it and part of the
01:14:48
reason is because the defense industry
01:14:50
is Consolidated down to these five Prime
01:14:53
defense contractors who are basically an
01:14:56
oligopoly and we have this Cost Plus
01:14:58
procurement system where they just raise
01:15:00
their prices every year and the
01:15:01
government pays it
01:15:03
so I think all of us want the United
01:15:06
States to have an effective defense I
01:15:07
mean I want us to use our military power
01:15:09
more wisely I don't like all these
01:15:10
stupid Wars we keep getting in but I do
01:15:12
want the United States as an American to
01:15:15
be the most powerful country I do want
01:15:19
us to get the best value for our defense
01:15:21
dollars and the only way that's going to
01:15:24
change is if if the defense industry
01:15:26
gets disrupted by a bunch of startups
01:15:28
doing Innovative things and there's no
01:15:31
question that I think drones are the
01:15:33
future of warfare so your point about
01:15:35
autonomy I think that is where this is
01:15:37
going next is that right now the drones
01:15:40
are typically controlled by you know
01:15:43
somebody with like a VR heads called
01:15:46
teleoperated teleoperated and they're
01:15:49
fpv this firsters Vision drone and you
01:15:52
basically strap some sort of explosive
01:15:55
on to it and then you drive it into
01:15:57
whatever your target is yeah those types
01:16:00
of drones are easier to disrupt by
01:16:03
electronic warfare because if you can
01:16:05
disrupt the signal from the T operator
01:16:08
then the Drone basically doesn't know
01:16:10
what to do and so you're right the next
01:16:12
step here is autonomous systems that can
01:16:17
be programmed with a Target and can find
01:16:21
it on its own make decisions on its own
01:16:23
and then they also build in some
01:16:24
shielding against some of this like
01:16:26
jamming technology or this ew electronic
01:16:29
warfare technology so that's where all
01:16:31
these things that's where the
01:16:33
battlefield is of the future is headed
01:16:36
and in a in a weird way if you think
01:16:38
about humans becoming a smaller and
01:16:40
smaller piece of the battlefield and
01:16:43
autonomous drones becoming a larger and
01:16:45
larger piece of it these wars become
01:16:48
resource Wars and basically technology
01:16:50
Wars that's right which which may or may
01:16:52
not be good I don't know but by the way
01:16:54
so so this is the point I wanted to make
01:16:55
if you pull up this chart if if that is
01:16:57
where Warfare is headed is a large
01:17:00
number of small autonomous systems that
01:17:03
go and find a Target and try and execute
01:17:05
a mission and you know in the case of
01:17:08
Ukraine saying they want to have a
01:17:09
million China might say I want to have a
01:17:12
100 million all of these systems are
01:17:14
dependent on lithium ion battery systems
01:17:16
today 79% of lithium ion battery
01:17:18
production comes out of China the US is
01:17:21
only 6.2% of global lithium ion battery
01:17:24
production and China has talked about
01:17:27
scaling up drone manufacturing to a
01:17:30
level that the US simply cannot even
01:17:32
contemplate in its industrial
01:17:34
architecture today so it seems to me
01:17:38
that if that is where Warfare tactically
01:17:40
is headed that China has a huge leg up
01:17:44
and is going to become a critical point
01:17:45
of dependency for the United States to
01:17:47
develop an
01:17:49
Arsenal necessary to be competitive in
01:17:52
this this kind of next evolution of
01:17:54
warfare I will say that like if you then
01:17:56
Game Theory this out like how do you
01:17:58
defend against these autonomous
01:17:59
electronic systems there's a technology
01:18:01
called EMP or electromagnetic
01:18:04
pulsing where as you guys know if you
01:18:07
run a very high current electric field
01:18:09
you can actually send out and emit a
01:18:11
pulse that then when it hits electronic
01:18:14
circuits far away induces a high
01:18:16
electric current in those circuits and
01:18:18
short circuits them so
01:18:21
emps are a defense system that allow you
01:18:24
to take electronic systems and this has
01:18:26
been you know kind of a a part of
01:18:28
warfare since probably the 1960s 50s but
01:18:32
targeted emps and targeted systems for
01:18:34
eliminating all of these autonomous
01:18:35
systems becomes a new defense
01:18:38
technology that I know several startups
01:18:41
in Silicon Valley that have been funded
01:18:43
with a lot of capital by folks that we
01:18:45
all know very well that are kind of
01:18:47
pursuing this we don't have a choice
01:18:49
because I think the point is that if you
01:18:51
nick you just post this photo we have an
01:18:53
enormous human capital problem with the
01:18:56
military which is there's just not
01:18:57
enough folks in listing anymore so we
01:18:59
don't have any choice except to automate
01:19:01
and become drone dependent just tell us
01:19:04
these numbers Jam this is a chart from
01:19:06
the Department of Defense that shows
01:19:08
military enlistments looking back from
01:19:10
about the 1970s through today and these
01:19:14
are two lines that are just going from
01:19:16
the upper left to the lower right and so
01:19:19
we're at alltime lows with respect to
01:19:20
the number of people that actually want
01:19:22
to join the military and so if we are
01:19:25
supposed to as saak said be this very
01:19:29
sharp fighting force then all of the
01:19:31
money that we're spending needs to get
01:19:33
allocated into things that can be
01:19:35
remotely
01:19:36
operated we don't have a choice yeah I
01:19:39
mean so freeberg you you asked the
01:19:40
question how do we defend against these
01:19:41
drones yes there's EMP technology if you
01:19:46
want to use an EMP you better make sure
01:19:47
there's not any planes nearby because it
01:19:49
will take them out of the sky you better
01:19:50
make sure by the way there are now
01:19:52
there's you can Target emps in a narrow
01:19:54
cone which is part of the the technology
01:19:57
you better make sure you don't hit any
01:19:58
of your own stuff because you'll fire
01:19:59
the electronics on those unless you EMP
01:20:01
Harden them which by the way will happen
01:20:03
I mean if if the US gets really good at
01:20:05
at sending out EMP pulses then our
01:20:08
opponents will start EMP hardening their
01:20:11
drones but any event yes it's a category
01:20:13
of Defense so is electronic warfare the
01:20:16
the one defense investment I've made is
01:20:18
actually in this idea of how do you
01:20:20
defend against drones and it's a startup
01:20:23
where I led the seed Dr called Allen
01:20:25
control systems and they have a product
01:20:26
called bullfrog which is a gun turret
01:20:29
that uses a standard m240 machine gun
01:20:32
but it's got a scanner next to it next
01:20:35
to the machine gun that is searching the
01:20:37
sky for enemy drones and it uses
01:20:41
computer vision to recognize them and
01:20:43
then it just shoots them down like very
01:20:45
quickly and it seems to me this is it's
01:20:48
maybe not the only way of doing it but
01:20:49
it's a really good way of doing it
01:20:51
because uh you can mount these things to
01:20:54
uh a vehicle I mean imagine in the
01:20:58
future that we have these autonomous
01:21:00
drones everywhere that are basically
01:21:02
assassination drones how you going to
01:21:04
stop them I mean how do you move the
01:21:05
president of the United States around I
01:21:06
mean in his Caravan you're going to have
01:21:09
to have you know his Caravan is going to
01:21:11
have to have some sort of electronic
01:21:14
warfare system on it to basically
01:21:17
prevent drones but it's going to need
01:21:19
some sort of last line of defense where
01:21:20
you can actually shoot a drone out of
01:21:22
the sky or stays in his b
01:21:25
so SX you don't you don't have any moral
01:21:27
trepidation around investing in defense
01:21:29
technology just to be clear well you
01:21:31
know I'd have to think twice about
01:21:32
investing in a in a weapon this is more
01:21:34
of a defense technology this is
01:21:36
defensive in nature where you're trying
01:21:38
to take out a drone that would be
01:21:40
attacking one of our bases or
01:21:42
installations or
01:21:43
convoys so this is fundamentally
01:21:45
defensive in nature and look I want
01:21:48
America to be the most powerful country
01:21:50
I'm an American I'm patriotic so I don't
01:21:53
want our troops be able to be hit by
01:21:56
enemy drones so no I don't have any more
01:21:58
problem with this I do have a problem
01:22:01
with all the stupid Wars the United
01:22:02
States gets into but that's on a policy
01:22:05
level I feel like that's separate right
01:22:07
from the question of do we want our guys
01:22:09
to be protected yes we do and chamat you
01:22:12
don't care you would invest in defense
01:22:13
technology companies no I care I think
01:22:16
that I would probably have to think
01:22:17
twice about making weapons and I'm not
01:22:19
sure that I'm really interested in that
01:22:20
but like I said eight years ago my
01:22:22
thought was we need to move to just
01:22:24
looking at those Trends we need to move
01:22:26
to an entire surface area of unmanned
01:22:29
autonomous vehicles and I really like
01:22:32
this idea of a class of machine that's
01:22:35
there to surveil and collect all the
01:22:38
important data so that you can make
01:22:39
better decisions and it's taken us eight
01:22:42
years but we're firmly entrenched now I
01:22:44
think the fact that both of you
01:22:46
highlighted degree of trepidation and
01:22:48
you guys are like fairly outside of the
01:22:51
middle bounds of how Silicon Valley
01:22:54
would think and make decisions indicates
01:22:56
like why so much of Silicon Valley has
01:22:58
deep concerns about investing in defense
01:23:00
technology and the US despite having the
01:23:03
best technology may be challenged by the
01:23:06
fact that our Silicon
01:23:08
Valley technologists and investors
01:23:12
are held back on
01:23:15
pursuing these Evolutions because of a
01:23:18
moral suasion and I think it is
01:23:20
something that the Department of Defense
01:23:21
has highlighted is that iar it's never
01:23:25
like we've been faced with that choice
01:23:27
and the reality is like if the choice is
01:23:30
stay small or scale up and be really big
01:23:34
you know it's hard to turn away from
01:23:37
that decision but that choice isn't a
01:23:40
realistic one this is why I was saying I
01:23:42
think it's you you can come into it with
01:23:44
the moral framing but I do think that
01:23:45
sax is Right which is if you believe
01:23:48
that America should be the most
01:23:50
important country in the world and you
01:23:51
have the ability to help it be that
01:23:55
eventually some of those decisions will
01:23:57
come in contact with this kind of
01:24:00
decision and usually though it takes
01:24:03
many many years before that's realistic
01:24:05
and to be honest from what I've
01:24:09
seen the opportunity to pursue that is
01:24:13
generally so financially remunerative
01:24:15
that everybody's like okay I think we
01:24:17
have to do it you know I don't have a
01:24:18
problem with investing in a company like
01:24:20
andrel I mean America needs defense
01:24:22
companies in order to maintain its
01:24:24
defense and the problem right now that
01:24:26
we have is there's these five Prime
01:24:27
companies that are Antiquated and really
01:24:31
expensive and we need startup disruptors
01:24:34
to that so I'm not like morally opposed
01:24:36
to it and in fact I think it's probably
01:24:38
a good thing I think my concern would be
01:24:40
I don't really want to be part of the
01:24:41
military-industrial complex I feel like
01:24:43
it
01:24:44
might like corrupt my views on things
01:24:47
like I don't want to you want war you'd
01:24:49
be Gering for war yeah like I don't want
01:24:50
to be I don't want to feel that's your
01:24:53
demand gen that's I right I don't want
01:24:55
to feel an incentive to be like these
01:24:57
generals who go on CNN and Fox News and
01:24:59
justify every single War that's my
01:25:02
biggest fear as well it's like it's one
01:25:04
thing to be investing in a bunch of
01:25:05
companies where we're sitting around
01:25:07
hoping for cheaper clicks on Facebook
01:25:09
and Google it's an entirely other thing
01:25:11
to sit on the board of a company and
01:25:12
hope for war that's a horrible place to
01:25:14
be I think on that point that's a great
01:25:16
point to wrap on guys this has been fun
01:25:19
a little more flat than it would
01:25:20
normally be with jcal we miss him we
01:25:23
wish him well we hope his tooth
01:25:25
recovers and um we will be back with
01:25:27
your it's like that scene in Dune where
01:25:30
he eats his tooth and releases that
01:25:32
poison gas that's right that's right get
01:25:36
better soon jcal please uh if you're
01:25:38
interested in the summit summit. Allin
01:25:41
podcast.co click subscribe on YouTube
01:25:43
I've been given notes to say that for
01:25:45
the chairman dictator chamath poopaa The
01:25:48
Raid Man David Sachs I am your sultant
01:25:51
of science Dave fredberg we will see you
01:25:54
guys next week love
01:25:55
you you
01:25:58
byebye let your winners
01:26:01
ride Rainman
01:26:05
David and instead we open source it to
01:26:08
the fans and they've just gone crazy
01:26:10
[Music]
01:26:18
withen besties
01:26:21
are that's my dog taking Drive wait wait
01:26:25
oh man oh man myit will meet me at we
01:26:30
should all just get a room and just have
01:26:31
one big hug orgy cuz they're all this
01:26:33
useless it's like this like sexual
01:26:35
tension that they just need to release
01:26:37
[Music]
01:26:42
somehow we need to get merch
01:26:48
[Music]
01:26:52
our I'm going all man
