Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:04
Hello everyone, this is Tyler. We
0:06
want to hear from you about the show. In
0:09
the show notes, there's a link to a short
0:11
survey where we want to learn more about you.
0:14
Why do you listen to the podcast? What
0:16
do you think is under or overrated about the
0:18
show? Is there someone you would
0:21
just love to hear me interview? Your
0:23
feedback will help us understand our listeners
0:25
so we can produce better episodes as
0:28
well as future live events and listener
0:30
meetups. Again, please find the
0:32
link at the top of the show notes to
0:34
fill out the survey. And thank
0:36
you. We very much appreciate your feedback.
0:44
Conversations with Tyler is produced
0:46
by the Mercatus Center at George
0:48
Mason University, bridging the gap between
0:51
academic ideas and real-world problems. Learn
0:53
more at mercatus.org.
0:56
For a full transcript
0:58
of every conversation enhanced
1:00
with helpful links, visit
1:02
conversationswithtyler.com. Hello
1:07
everyone and welcome back to Conversations
1:09
with Tyler. Today I'm here
1:12
in San Francisco chatting with Michael Nielsen.
1:14
Michael is hard to introduce and also
1:16
difficult to prepare for because he knows
1:18
and has done so many different things.
1:21
He's from Australia, has a PhD
1:23
in physics, has written what is
1:26
perhaps the best-known text or co-authored
1:28
it on quantum computing, is
1:30
one of the leaders of the
1:33
open science movement, has co-authored with
1:35
Patrick Collison on progress in science,
1:37
has worked at Y Combinator, is
1:40
an extraordinarily prolific writer, reader, commentator,
1:42
tweeter, mentor to others, mentee, and
1:45
many other things and currently is thinking
1:47
about the fragility of civilization and much
1:49
more. Michael, welcome. Thank you so much,
1:51
Tyler. So you were saying there should have
1:54
been a metaculous on the opening question.
1:57
Why is the universe beautiful to human eyes?
2:00
Oh, selection? I have
2:02
no idea. I mean, selection is a
2:04
very attractive kind of an idea. I
2:06
mean, kind of to think not just
2:08
instinctively. But yeah, I don't know.
2:12
Why are there simple rules? Why do we have
2:14
simple rules governing the universe? In fact, why is
2:16
simplicity and arguably truth somehow associated to beauty? Physicists
2:20
tend to assert that this is the case,
2:22
but I don't think anybody really knows the
2:24
reason why. How beautiful do we in fact
2:26
think the universe is? The
2:28
people don't buy paintings of the universe. People like
2:30
you might, right? But it's not. Oh, I have
2:32
a painting. I have the Hubble Deep Fields on
2:34
my wall, of course. But the most expensive paintings
2:37
are not of the universe. There are people. There
2:39
are voting scenes, right? I don't think that's really
2:41
true. I mean, the James Webb Space Telescope was
2:44
about, I think, was $10 billion and is
2:46
arguably a machine for producing that
2:48
kind of image. It's got
2:50
to be one of the most important sort of
2:53
image factories and most expensive image factories ever made.
2:55
So I'm not sure I buy that. What's the
2:57
most beautiful image of the universe? The
2:59
image of – we
3:01
have a sort of a sequence of improved
3:04
images of the three-degree
3:06
microwave background is – I
3:09
don't know. Is it the most beautiful? It's
3:11
maybe the most extraordinary. It really is sort of a
3:13
photograph of the universe as a whole. You
3:15
can look at that and it says something about
3:17
structure out in creation. Why
3:20
do the sounds of the universe not
3:22
appeal to us so much, right? So
3:24
it's beautiful visually. But orally, it's –
3:26
you know, we create very complicated things
3:28
which we call music, which are beautiful.
3:32
Yeah, that's a great question. I
3:34
don't know why music is beautiful.
3:38
People have made attempts. You know, there's
3:41
things like sort of chirp sounds that might be
3:43
produced near a black hole and sort of ideas
3:45
like this. And you're right. They
3:47
tend not to be all that beautiful.
3:49
The only ones that I can think of that
3:51
sort of offhand, you know, it's
3:53
being produced by evolution. Birdsong is beautiful,
3:55
but we're actually quite closely related to
3:57
birds. So it's maybe not so surprising.
4:00
If I think about things like, what's
4:02
his name, Ron Seksmith, I think
4:04
his name is a composer in Toronto,
4:06
has made these musical pieces based on
4:09
the different periods in the solar system.
4:11
So the time that the Earth
4:13
takes one year to go around
4:15
the sun, but also then Mars and Jupiter and all
4:17
these. And they are
4:19
noticeably not particularly attractive musical
4:22
pieces. So that's a
4:24
good question. I wonder if the beauty of light isn't
4:26
part of the reason for the beauty of the universe.
4:29
So as human beings maybe were evolved to be
4:31
attracted to light, it gives you an integrated theory
4:33
of the beauty of the universe and beauty of
4:35
paintings. Vermeer, a great painter, he's
4:37
very attractive to people because of how he
4:39
uses light. When you look at the universe,
4:42
you're typically seeing signatures of light in many
4:44
cases. You look at the Milky Way, right?
4:47
It's pretty strange. We
4:49
see in such a tiny band
4:51
of wavelengths, we're really not
4:53
seeing almost anything. We're not seeing into the
4:55
infrared and the radio. We're not seeing into
4:57
the ultraviolet and the X-ray. So
5:00
a lot of what we view as
5:02
beautiful locally, I mean, it's got this
5:04
sort of evolutionary explanation again. Why
5:07
the large scale structure is
5:10
beautiful. Okay. I
5:13
can maybe attempt to sort of, I partially believe
5:15
this explanation, which is that we
5:18
do seem to be sort of programmed to
5:20
recognize and find attractive
5:23
instinctively novelty which
5:25
is associated to structure somehow. And
5:28
so we look and we see spiral galaxies
5:30
or things like that. It's reflecting something which
5:32
is interesting. We don't necessarily know quite what,
5:34
but maybe there's kind of an evolutionary explanation
5:36
for why that is attractive at least. But
5:39
I can apply that explanation to your question
5:41
about sound. It's equally as good and unfortunately
5:43
it just doesn't seem to hold there. So
5:47
I'm not that confident. Now you've written that
5:49
in the first half of your life, you typically
5:51
were the youngest person in your circle. And
5:53
that in the second half of your life, which
5:55
is probably now, you're typically the oldest person in
5:58
your circle. How would you model
6:00
that as a claim about you? I
6:04
hope I'm in the first five percent of my life, but
6:06
it's sadly unlikely. You're 50 now and you're just 200, right?
6:09
Which is plausible. Which is plausible. And you would now
6:11
be in the second half of your life. Yeah. I
6:14
mean, I can give shallow reasons. I can't give good
6:16
reasons. The good reason in
6:18
the first half was so much of the work I
6:21
was doing was kind of new fields of science, and
6:23
those tend to be dominated essentially for sort
6:26
of almost sunk cost reasons. People who don't
6:28
have any sunk costs tend to
6:30
be younger. They go into these fields. So there's
6:32
kind of early days of quantum computing, early days
6:34
of open science. They were dominated by people in
6:36
their 20s, and then they'd sort of
6:38
go off and become faculty members. They'd be the youngest person on
6:41
the faculty. Now, maybe
6:43
it's just because I found sort of San
6:45
Francisco and it's such an interesting cultural institution
6:48
or sort of achievement of civilization.
6:50
We've got this amplifier for 25-year-olds
6:53
that lets them make dreams
6:55
in the world. And that's for me anyway, for
6:57
a person with my personality, very attractive for many
6:59
of the same reasons. Well,
7:01
let's say you had a theory of your
7:03
collaborators. And other than yes, they're
7:06
smart, they work hard, but trying to pin down
7:08
in as few dimensions as possible, who's likely to
7:10
become a collaborator of yours after
7:12
taking into account the obvious? What's your
7:14
theory of your own collaborators? They're
7:17
all extremely open to experience. They're
7:19
all extremely curious. They're all extremely
7:21
prosocial. They're all extremely ambitious. They're
7:24
all extremely imaginative. And
7:27
do you think that ends up pairing you with
7:29
collaborators who are more different than you? So
7:32
a lot of collaborators are very similar, and then
7:34
other types are very different. So almost always as
7:36
well, I will select for somebody who has at
7:38
least one very strong skill, which I do not
7:41
have. And that's sort of enough diversity
7:43
from my point of view. And that
7:45
may account for some of the age differences throughout
7:47
your life. But also then there's also just this local
7:49
selection effect. I live in the Bay Area. There's a
7:51
lot of really amazing 29-year-olds around. It's
7:54
just incredible. I was told to ask,
7:56
what's the influence of Simone Weil on you? interesting
8:00
question. She's one of the
8:02
maybe the best examples of sincerity that
8:05
I know of. The fact that
8:07
she wrote what she wrote for herself, she
8:09
wasn't attempting to get published. It was just this
8:12
deep internal colloquia that was going on. And
8:14
it's reflected in every aspect of her life. She
8:16
went off to fight in Spain at a time
8:18
when women did not go off to fight in
8:20
Spain. Everything that she did, she did at 500
8:22
miles an hour. She's remarkable
8:25
as kind of a ... She's an extreme, very
8:27
extreme type of a human being in a way
8:29
that I find very interesting. If
8:31
you and she collaborated, what would it be on? I
8:35
don't think ... I'm sure she was
8:37
a difficult person. Although, her brother Andre Vile
8:39
was a very great mathematician. And you can
8:41
sort of see in some of the stories
8:44
about the two of them that she must
8:46
have quite liked the sort of scientist types.
8:48
Maybe we would have found something to
8:50
collaborate on. Why is Charles
8:52
Sanders Perr still an important thinker? I don't know
8:54
enough about Pierce to be able to answer that
8:56
question. You and I
8:58
were both fans of Olaf Stapleton, who wrote
9:01
the dual classics, Last and First Men and
9:03
Star Maker. What's the
9:05
biggest analytical mistake he made in those
9:07
narratives? So a lot of
9:09
implausible things happen, right? But those are too simple
9:12
to point to. Where is his understanding of the
9:14
social world going wrong? He
9:16
was both, certainly to some degree, a socialist
9:18
and certainly a pacifist. Though in World War
9:21
II, he switched out of pacifism. He did.
9:23
Yeah, as in many people. I
9:26
find myself as I read those books
9:28
actually becoming a little bit more sympathetic.
9:30
I'm not a priori, particularly sympathetic to
9:32
them. And I start to think
9:35
he has this very long view of
9:37
history, I mean, much longer than most people who
9:40
say they have a long view of history. And
9:42
I think he sort of he sees some of
9:44
his pacifism in that light. It's kind
9:46
of questions about what's actually good for a species
9:48
or in fact, not even a single species, but
9:50
across multiple species is it good to be pacifist?
9:52
And there's a really interesting point of view. It's
9:54
hard to reconcile with kind of a selfish gene,
9:56
kind of a point of view. But of course,
9:59
this is an ongoing problem in evolutionary biology,
10:01
it actually seems like group
10:04
selection doesn't quite work, but
10:06
something at that level has to be a little bit
10:08
true. So if you take that seriously,
10:11
then maybe his pacifism, which just seems
10:13
like sort of an outright mistake, maybe
10:15
it's actually justifiable in some way. Actually,
10:17
I'm not answering what I've answered. I've
10:20
answered the inverse of your question, which is
10:22
to justify the bits that I, a priori,
10:24
find most implausible. But yeah, I think those
10:26
are mistakes. I worry that he too quickly
10:29
assumes collective action problems are solved, which
10:31
is close to your answer. So he
10:33
thinks the League of Nations can be
10:35
effective for a long period of time,
10:37
which I suspect was not really contingently
10:39
possible. And he has this
10:41
Hegelian sense, what Hegel would call a national
10:43
spirit, for him is a
10:45
civilizational or certain stage of
10:47
man spirit that so shapes how
10:50
people think. And
10:52
I hang out with a lot of economists. I
10:54
think that's much stronger than the economists believe, your
10:57
overall view of the world and what's important. But
10:59
I don't think it's nearly as strong as
11:02
Stapleton believed. So the way
11:04
in which collective spirit rules millions, billions
11:06
or trillions of beings, I
11:08
feel he's overestimating the efficacy of that. The
11:12
comment about the League of Nations is really interesting. I think
11:15
there's this spirit at the time, lots of
11:17
people wanted this idea to work, lots of
11:19
his friends would have wanted it. I
11:21
think it's a sort of a shallow kind of a
11:23
mistake that he made there. That your comment about collective
11:26
action problems seems much more to the
11:28
heart of it. I think he didn't
11:30
really believe in them
11:33
or actually sort of understand just how difficult
11:35
they are to solve, how difficult it is
11:37
to supply public goods and these
11:39
kind of things. He always does away with
11:41
it sort of narratively and it's just assumed
11:44
away without really a mechanism being
11:46
given. He assumes a
11:48
lot away in those books. When the
11:51
problems are interesting, he usually doesn't and that problem
11:53
is interesting and he still assumes it away. I'm
11:55
not very sympathetic to that at all. But I'm
11:57
not sure how big a mistake League of Nations...
12:00
Was so clearly it didn't work
12:02
and i just criticized him for it but
12:04
if you think about eighteen fifteen up through
12:07
the first world war almost a century you
12:09
have an unprecedented degree of peace in much
12:12
not all of europe. And
12:14
everyone has just lived through that and they may be
12:16
thought that was not possible and maybe that is itself
12:18
still a bit of a mystery. And
12:21
then there's world war one and you feel
12:23
you can get back to some version of what you
12:25
had in the league of nations appears
12:27
to be the closest path to doing that and
12:30
it might have been more possible at the time.
12:32
I'm just saying this thing is a gap between
12:34
aspiration and what actually happened with the league and
12:36
then later with the united nations i think you
12:39
had the hopes and then you had the
12:41
what actually happened. Yeah there's a gap with
12:43
the as prototypes you know over the next
12:45
few centuries maybe these things are terrific maybe
12:47
we learn a lot from them things like
12:49
that i don't know whether the montreal protocol
12:51
have been possible without the united nations probably
12:53
not. I have a very concrete
12:56
question for you and this is to clear up
12:58
a confusion of mine so i've asked experts in
13:00
quantum computing what's the status of quantum computing right
13:02
now some of them say we already have it
13:04
some of them say the others will tell you
13:06
we already have it but we don't. Others will
13:09
say we're on the verge of having it and
13:11
there's two or three other answers i
13:13
hear often people who nominally would seem to know
13:15
what they're talking about. So
13:18
let me ask you michael what is actually
13:20
the status of quantum computing right now i
13:22
am i'm the wrong person i am determined
13:24
leave very agnostic about this i stopped i
13:26
worked on it from ninety ninety two two
13:28
two thousand seven. Actually
13:30
i do keep up with friends i'm
13:33
going to have coffee after this with
13:35
somebody who's still on the quantum train
13:37
it's very impressive progress each year it
13:39
is an extremely difficult problem it's not
13:41
solved there's no way it's definitely not
13:43
solved. But the fact that there's sort
13:46
of order hundred cubit systems which you can apparently manipulate
13:49
as you will suggest to me we just
13:51
wait it's going to happen we don't know
13:53
what it will mean. What's your
13:55
maximum likelihood estimate for
13:57
the first year when it will do something useful.
14:01
Useful to me or useful to symbolization? Useful
14:03
to anyone. And the most interesting thing
14:05
would be to discover that quantum mechanics was wrong from
14:07
my point of view. The other most interesting thing is
14:10
probably discovery of new materials. How
14:12
would it discover new materials? Just by being
14:14
able to do simulations very, very rapidly. It's
14:17
very hard to do simulations of stuff down at
14:19
the quantum scale. The way that we have are
14:21
pretty terrible and often produce wrong results. The fact
14:23
that we may actually have a very high throughput
14:26
way of doing lots and lots of
14:28
simulations which give correct results. It's
14:30
like being able to do a thousand times as many
14:32
experiments as before. That will just speed things up
14:34
insofar as there's anything to discover. I can't tell
14:36
you what we'll discover. Will there be
14:38
quantum money? Will all money be quantum money in this
14:41
world whenever it comes? I actually don't know. I mean,
14:43
there's this old idea of Steven Wiesner, which he called
14:45
quantum money. It's meant to
14:47
be uncountifedible. I don't know. But
14:50
isn't everything else countifedible if
14:53
quantum computing is up and running and
14:55
thus you need quantum money to protect
14:57
against just your counterfeiting? Most
14:59
of the 19th century monies, they were often counterfeit.
15:01
We don't know the exact percentage, but we believe
15:03
it was quite high. Yeah.
15:05
Well, it's still true in the world
15:08
today, never mind the 19th century, as we've talked about
15:13
before. I won't be surprised if
15:15
we end up with systems like that. It's hard
15:17
to make it stable. That's the issue. But my
15:20
guess is that in the long run, we actually
15:22
will find ways of making
15:24
quantum systems surprisingly stable. That's
15:26
speculation on my part. But if I
15:28
come back in 100 years' time and that's true, we
15:31
may just have quantum coherence everywhere. Do
15:33
you think that leads to a mass privatization of
15:36
a lot of social activity? So something like AI,
15:38
we're in San Francisco, the private sector does it.
15:41
No government is really close to doing it,
15:43
right? You have
15:45
to pay high salaries, hire the most talented people.
15:48
So if AI and quantum computing are done
15:50
by the private sector, what
15:52
is government in that world? I
15:54
don't know. I mean, it's
15:57
an interesting fact, right, that work on nuclear weapons
15:59
was at... actually nationalized in I think 1948
16:01
or something like that. So,
16:04
potentially that's just one answer, right? Because sort of
16:06
contingently... But that seems more of a brute force
16:09
thing. Sure. But then what does it
16:11
say? Open AI is done. I'm just saying
16:13
that's a potential outcome. I think quite a
16:15
plausible potential outcome. I don't think it's likely,
16:17
but it's not 99% unlikely either. Yeah,
16:20
I mean that's certainly... I mean it's kind of a
16:23
very Neil Postman point of view. You have this basically
16:26
you... Or almost Larry Lessig,
16:28
Coder's Law. You just keep building
16:30
more and more sort of governance infrastructure
16:32
into the technology, and you're moving
16:34
it out of the hands of the population and
16:36
into the technology. And that seems to be certainly
16:38
the story of the last hundred years and very
16:41
likely the story of the next hundred years. Is
16:44
the status of linear algebra rising? That's
16:48
a great question. It probably has, yeah.
16:50
It's prominent in quantum, right? It's
16:53
prominent in AI. Digital
16:55
is built on, you know, matrix multiplication. It's
16:59
prominent for a lot of reasons. What
17:01
should we infer from that about the whole nature of the
17:03
world? So if differential equations
17:05
were rising in status to a similar degree,
17:07
we might infer one set of
17:10
things. But linear algebra, you almost feel
17:12
a bit more grounded, don't you? When
17:16
I took that class, I felt I understood it. I
17:19
never quite know what your status questions mean, Tyler. I
17:22
don't know what it means for something to rise in
17:24
status. Well, AI now seems more
17:26
important than it did five years ago, and
17:29
matrix multiplication is a big part of that.
17:31
If quantum computing happens, as you're predicting it
17:33
will, well, that I think would also
17:35
make matrix algebra rise in status.
17:37
Like, oh, this is a really important tool.
17:39
It's behind all our quantum money. Do
17:42
you mean it's going to have more money go
17:44
to it, more power go to it, more glamour
17:46
go to it? Are people going to regard this
17:49
as all of those? But
17:51
you would revise your ideas about the fundamental
17:54
nature of the universe, just like our current
17:56
understanding of quantum mechanics. It might be incorrect,
17:59
but at least in the short term. short run, it seems
18:01
like probability theory is somehow more
18:03
important than Einstein might have thought.
18:07
And as you know, he famously asserted God
18:09
is not playing dice with the universe, perhaps
18:11
incorrectly. I mean, the
18:14
people who remake this
18:17
understanding are very good at
18:19
ignoring status. But
18:21
others aren't. But others aren't. I think
18:23
I'm inclined to think. I mean, maybe.
18:25
I don't care. I
18:28
actually just don't care that much. I
18:30
mean, if you're searching for comparative advantage in doing creative work,
18:33
you want to know where status is, but mostly so you
18:35
can avoid it. Yeah, absolutely. Be short
18:37
status, as Peter Peele has to say. Yeah,
18:39
exactly. That's what we'll put. Yeah. Is
18:42
there any chance Roger Penrose is right and the human
18:44
brain is some kind of quantum computer? I
18:46
would love it if he was right. I
18:48
think the answer, unfortunately, is not really. It's
18:51
certainly possible that there's some very interesting structure in
18:54
there that is quantum mechanical in some really interesting
18:56
way. I mean, lots of structure in there is
18:58
quantum mechanical. The reason why atoms are stable has
19:00
to do with quantum mechanics, like all these sorts
19:03
of things. But like an interesting,
19:05
unsuspected way would be... That would be terrific and
19:07
I think is not completely out of the question,
19:10
but it probably doesn't affect anything about consciousness or
19:12
anything like that. I would be very surprised if that
19:14
were the case. How
19:16
are we going to make progress toward a theory of
19:18
quantum gravity, a general understanding of
19:21
everything? We seem to be stuck. Many
19:23
people hate string theory. Many people hate
19:25
Everett, many worlds. Those seem
19:27
to be two major contenders. Where
19:30
are we at and what's going to happen next? One
19:33
fun reason for working on quantum computing
19:35
is you're trying to build the largest
19:37
scale fully
19:39
quantum coherent systems that have ever been built. Whenever
19:42
you push on into a new regime like that,
19:44
there's some chance that things break down. If
19:46
something was to break down there, that would be fantastic. Because
19:49
we'd learn a lot. Because we'd learn a lot. The
19:51
problem in some ways in physics has been that the fundamental
19:53
theories have been just too successful for the last 50 years.
19:56
Yes, you're right again. It's very attractive for sort
19:58
of a few years. but over 50 or 60
20:02
years it's terrible. And I think
20:04
that's certainly part of the issue with
20:08
quantum gravity. Does it bother
20:10
you that so many people hate string
20:12
theory, think it's now low status, think
20:14
it's not aesthetic, think it's unintuitive? Does
20:17
that carry any weight with you or do you want
20:19
to be like short status again on this one? I
20:22
mean there's the question of sort of
20:24
inside and outside. Well, there's a question
20:26
of inside and outside the profession. There's also the question
20:28
of inside and outside the group of people who know
20:30
something and those two are not exactly the same group
20:32
but there's a lot of overlap. So
20:35
I mean outside it affects funding
20:37
a little bit. Well, we actually need to be quite
20:39
a bit. And so in that sense it matters. But
20:42
internally I think I'm
20:44
more interested in the question of just how
20:46
much diversity of opinion is there. People
20:48
pursuing lots of different ideas. One of the
20:50
things that I've noticed over many years is
20:52
I just find mathematicians when I talk to
20:55
them. It's such a healthy culture because each
20:57
mathematician is really, well, a lot of them
20:59
are very unique. They've got their own sort
21:01
of particular path and their set of beliefs.
21:03
Physics, theoretical physics often seems just a little
21:05
bit more monotone. They can sum
21:07
themselves up in a few words when they're
21:10
talking to their professional colleagues and that's not
21:12
so healthy. So I'm really
21:15
not so interested in the question you asked. I'm much more
21:17
interested in the question of how do you generate that kind
21:19
of diversity. And do you
21:22
feel that ultimately the final theory of a
21:24
universe or metaverse ought to be simple? Who's
21:28
declaring ought here? But
21:31
when someone presents a theory to you, do you ever say,
21:33
no, that's too complicated. It might be an intermediate theory
21:36
at some level, but it's not going to be the
21:38
final theory because I hear this from many people.
21:41
A lack of satisfaction. You
21:43
want surprise. It's the same when I've, you
21:46
know, the little tiny pieces of economics I've
21:48
learned when I hear about, I don't know,
21:50
Ricardian comparative advantage or something like this. Just
21:53
a nice little element of surprise. You're getting a
21:55
free lunch somehow. So I'm more interested in that
21:57
than I am maybe in the question of simplicity.
22:00
What makes for physicists who age well? I
22:05
spent quite a bit of time thinking about this as actually
22:07
in my late 20s and went to
22:09
look and see what seemed to
22:11
distinguish older physicists who had aged
22:13
well and and older physicists who had maybe
22:16
gotten a bit too complacent. So I can
22:18
tell having younger mentors was really the key.
22:20
And why is that important? I
22:23
don't know. I mean, I have theories. This was
22:25
an empirical observation. Yeah, but what's your best theory?
22:27
What I think is probably the case, it's
22:29
almost a network effect. Basically, if
22:32
there is some slight sort of downhill slide
22:35
and most of your friends, you
22:37
know, are not quite at the edge anymore,
22:39
that's going to infect you. But
22:41
if you still have mentors who are 25, 28,
22:44
extremely active, and they're active in
22:47
the latest ways, you get
22:49
to partake of kind of the positive network effects.
22:51
And so I think that's why it's very important
22:53
not just to have not to have people who
22:55
work for you, lots of 70 year
22:57
old physicists have 23 year old students, but
22:59
actually to have 23 year olds, 28
23:02
year olds who you really learn from and
23:04
you regard as your mentors. And
23:06
holding constant your degree of power and
23:08
influence, what's the best way to attract
23:10
younger mentors? Find
23:12
people whose work you admire and befriend them. And
23:15
you think that works pretty well. Yeah. And just
23:17
being nice. I'm
23:20
not sure being nice is the right... You
23:23
have a lot of younger mentors. You're
23:25
known famously for being very
23:28
nice, right? This is partly
23:30
a question about your own self-awareness, but
23:32
has you being very nice helped you get more
23:34
younger mentors, or are they attracted to
23:36
other aspects of you? I am extremely disagreeable,
23:38
but in a polite way, I hope, and
23:41
a kind way, hopefully. So you are very nice then.
23:44
Well, people often find people who are
23:46
disagreeable actually quite difficult. But if
23:49
you look at all of the younger mentors I've had
23:51
in the last, say, seven or eight years, they're all
23:53
people who enjoy disagreement. They
23:56
say the thing that they think is obvious and you say, here's
23:58
another way of looking at it. And they're like,
24:00
oh, they want to engage. Some people get insulted or
24:02
they get threatened or they get annoyed when you do
24:04
that. And those people are not going to be –
24:07
they're not going to be good collaborators. They're not going
24:09
to be a match. As
24:11
the years pass, do you think your probability for
24:13
God existing is going up or down? Which
24:16
type of God are you referring to here?
24:18
Are you referring to like an Abrahamic God
24:20
or what? Not a particular
24:22
religion, but some explanation
24:24
that would seem to stand prior to
24:26
and outside of what we call physics
24:29
and would be mystical in some way?
24:33
It's been – that hasn't changed
24:35
since I was seven years old. But
24:37
that's weird that it hasn't changed, right? You've
24:39
learned a lot. Why shouldn't it change in
24:41
whichever direction? I had explained to
24:43
me three basic theories of cosmology when I was
24:45
seven, one of which was the Big Bang and
24:47
then there were two others which was the steady
24:49
state theory and a third whose name I don't
24:51
even remember anymore. And
24:56
they leave some questions unanswered. Why
24:58
is there anything? But as far as I can
25:00
tell, we haven't made any progress on sort of
25:03
those things in the 40
25:06
odd years since. Yeah, it's frustrating
25:08
actually that that's the case. I mean, I think
25:10
you're correct to say, you've learned a
25:12
lot. Why haven't you changed? And
25:14
my response to that is, I've learned a
25:17
lot. Gosh, it's really annoying that it hasn't
25:19
impacted that question more. I
25:21
think my pee has gone up a modest amount
25:23
over time. So when I was saying
25:25
in my young 20s, I thought
25:28
physics was going to make more progress than it
25:30
has at fundamental theoretical levels. And
25:32
the fact that it hasn't, it nudges me a bit
25:34
to wonder, well, these other
25:36
types of explanations that I was not so
25:38
keen on, maybe they're a
25:41
bit more important than I had thought. That
25:44
hasn't happened with you? No.
25:46
I think that that's an interesting – it
25:49
hasn't really. I think I just don't think 40 years
25:52
is very long. If it had been 100,000 years, so it's – But
25:55
it's all I've got in a sense. I mean, I'm – I
25:58
know. My
26:01
opinion time span is going to be 40 plus
26:04
something. I think my appreciation for God has gone
26:06
way up. I
26:08
appreciate the construction of the religions far more
26:10
than I did, what notions of
26:12
God do for people, vastly more appreciative.
26:15
But my probability, I don't think, has
26:17
really changed. And what about evolutionary frameworks
26:20
where there's some Darwinian process, some kinds
26:22
of universes within a broader metaverse,
26:24
they reproduce at greater frequencies, that
26:27
shapes the properties of what we live in.
26:29
Isn't that a kind of substitute for a
26:32
good explanation and that rises in probability just
26:34
a bit? No. I
26:36
mean, why not? You
26:39
know, you're relabeling what you mean by universe. If
26:42
you just sort of use a term that
26:44
means everything that is, then that hasn't changed.
26:46
Our model of what it might be has
26:48
potentially changed quite a bit over the last
26:50
few decades. But maybe there's a
26:52
simple theory for the metaverse, but we can never,
26:54
ever see it. It's like mastic religion. And
26:57
then our own universe, there's not a simple theory, but
26:59
we do know the parameter values we got are
27:02
enough to drag it across the finish line. And
27:04
that takes some of the burden off physics in
27:06
a way. Just like,
27:08
well, the platypus, it seems an unlikely creature, but
27:10
it has in fact survived. Okay. Some
27:14
of the things that came arbitrary. And being a
27:16
good Australian, you appreciate the platypus, right? Yeah, indeed.
27:18
The platypus is quite a bit, it's good for
27:21
fooling, visiting Americans about whether or not this animal
27:23
can exist or not. Yes. I
27:25
still don't find that compelling. I think because there's
27:27
always, we've always known that there seem likely to
27:29
be fairly contingent facts about the universe. I mean,
27:32
it shifts the level at which they are. It's
27:34
more interesting if the value of the
27:36
fine structure constant is actually a contingent fact. That
27:39
is interesting. Or if some of
27:41
the other couple of constants are changing over
27:43
time or sort of models like this. But
27:45
it's not particularly, it's still not getting
27:47
at the essential question from my point of view, which is
27:50
why is there something rather than nothing? What
27:53
I've long thought is an impossible question. So
27:56
we might have theories of parameter values or
27:58
be able to predict how things are going. interact or what happened
28:01
a long time ago. But
28:03
the Heideggerian question, I don't think
28:05
it's a meaningful question at all. Because
28:09
the word why is already embedded
28:11
in some context, which it sends
28:13
a self-undercutting query. The
28:16
open science, why do some
28:18
fields have preprint platforms and others not?
28:20
Is there an actual regularity or is
28:22
that random and path dependent? I
28:25
think a lot of that probably comes down
28:27
to individuals. One of my favourite things, years
28:29
ago, before they started to spread in biology,
28:31
I would often ask physicists
28:34
and biologists this question. Why
28:36
is there preprints in physics but not in
28:38
biology? The biologists
28:41
would say, well, biology is
28:43
so much more competitive than physics
28:45
that we can't possibly bear to share our
28:48
results too early. The physicists would
28:50
say, physics is so much more competitive than biology
28:52
that we have to share them as rapidly as
28:54
possible to get the word out. But with COVID,
28:56
didn't biology go the route? Clearly some parts of biology
28:58
go the route of physics. And it just seems like
29:00
it's just a cultural problem. It turns out it's a
29:03
little bit more like fashion or something like that. It
29:05
does need to be solved. Like if you look at
29:07
what was done in the early days of the physics
29:09
preprint server, some very clever things,
29:11
actually things which are reflected in some of
29:13
my favourite economists, some of their ideas were
29:15
done by Paul Ginsberg when he was starting
29:17
up the preprint server. He went very narrow.
29:19
He didn't try and solve the problem all
29:21
across all the fields. And he went and
29:23
kind of twisted the arms at some level
29:26
of some very high status, high
29:28
profile physicists to say, I would
29:31
like you to use this service, send me your
29:33
best paper. So on the first day, Andy Strominger,
29:35
who I think is at Harvard, was
29:38
on the preprint server and Ed Witten showed up
29:40
very quickly. So these are very prominent people. And
29:43
you get this sort of just a tiny community, but then
29:45
you can sort of, you can start to attach
29:48
other communities. But that
29:50
was just, that's a very contingent fact about history.
29:53
It could have happened in some sub-discipline
29:55
of biology as well. Why do
29:58
so many crummy journals survive? can be
30:00
quite expensive, you might also have to pay
30:02
to publish in them. They seem
30:04
terrible that if a good piece were in them, the
30:07
journal would not certify the piece. If
30:09
anything, the piece would help certify the journal. Why
30:11
can't we get out of that? Yeah, I mean there's
30:14
a complicated set of things going on. One
30:16
is that libraries pay not individuals
30:18
usually for subscriptions. They're not actually that
30:20
really the person getting the utility is
30:23
not the same as the person
30:25
making the buying decision. That's always bad. There's
30:28
also the fact that since the 1990s
30:31
and the rise of the Internet, we
30:33
get economies of scale. You don't subscribe.
30:35
Libraries don't subscribe to individual journals for
30:38
the most part. They subscribe to all
30:40
these giant bundles. It was actually a
30:42
terrific idea at some level.
30:44
It's a way of passing on economies of
30:46
scale in publication to the customers, but it
30:48
does go some way to explaining why these
30:50
crummy journals persist. But you think
30:53
in part libraries are inefficient at capturing rents
30:55
for themselves. They get this budget,
30:57
they spend it on bad journals. It
30:59
might be better for the world if they just
31:01
took the money home and bought ice cream. Or did
31:03
whatever. Yeah. There's many other things that
31:05
could be done with that money. It's
31:07
difficult for them to reason about having talked to many
31:10
librarians. They will do things like they will use impact
31:12
factor. I mean that's the
31:14
differentiator that they tend to use. So
31:16
they'll try and get all impact factor,
31:19
whatever it is, and above journals. That's
31:21
the kind of the way they seem to think.
31:23
They're just using very imperfect proxy. They understand as
31:25
well as anybody that their proxy is imperfect, but
31:27
they don't have anything better to do as far
31:30
as I can tell. So it's
31:32
a very unfortunate situation. And
31:34
right now, how high are the
31:36
marginal returns to greater openness? So
31:39
put aside terrorists, manufacturing, new pathogens.
31:42
Put aside people figuring
31:44
out how to make their own nuclear weapons,
31:46
AI problems. So putting aside the very negative,
31:49
just if the good stuff were more open, how
31:52
much more rapidly would science progress? Per
31:55
se, I mean that's a very weak word.
31:57
You need to be much more specific. could
32:00
say the culture around Jupiter Notebooks in machine
32:02
learning, I think having those very openly available
32:04
and widely available really has driven a lot
32:07
of progress. You can just write
32:09
your Jupiter Notebook with your experiment, you make it
32:11
available to other people and that can really drive
32:14
a lot of progress. It's not the same as
32:16
making your journal article openly available. It's a much
32:18
more active kind of a material. Do
32:20
I think that that is an important
32:23
component in really significantly speeding up science?
32:25
Yes. But it's
32:27
not going to be too x. I think that there's much
32:30
larger than 2x possible and
32:32
this is a piece of that. Yes,
32:35
but it's not 2x on its own. No,
32:38
it's actually too undefined a term. Openness
32:40
is always with respect to what platform, with respect
32:42
to what set of institutions, with respect to what
32:44
set of norms. With the current sets
32:47
of norms and institutions that we have, it buys you
32:49
a little bit. I don't think it buys you that
32:51
much. But the norms and institutions, they're going to change
32:53
in response. The way in which people work will change
32:56
in response. The Jupiter
32:58
notebook example I gave is I think a good
33:00
example of that. Why are science
33:02
textbooks so expensive? Is it
33:04
marginal cost? Is it third party payment
33:07
problems? Is it something else? I
33:09
don't know. Is it instructor lock-in because the notes are
33:11
geared to the text they've worked with for 15 years?
33:15
Yeah, very
33:17
few professors make that much money
33:19
from the textbooks that they write. But they're often
33:21
very protective. I see people complaining
33:24
on Twitter that they're not going to get the
33:26
$400 check next year
33:28
for their textbook under a new open access
33:30
policy. And they're really up in arms about
33:32
this $400. That's interesting.
33:35
It's hard for me to empathize with
33:37
psychologically. Don't
33:40
understand. I
33:42
wrote this neural nets textbook which I put
33:44
online for free. And that
33:47
massively, that's really made a
33:49
really large difference to the impact which
33:51
it's had. In fact, even if
33:53
I just think purely financially, I wasn't doing it and
33:56
thinking, oh, financially this will be better off. But
33:58
The greater impact has actually benefited. the Me:
34:00
much more financially than any amount of royalties
34:02
ever. would you mean like giving talks are
34:04
being invited Skyn? I just in general people
34:07
know. B. O P Bar what you've done
34:09
what you're interested in and they're much more likely
34:11
to provide all sorts of different opportunities including jobs.
34:13
I think from the point of you have the
34:15
authors it really actually doesn't make that much sense
34:18
From Bolivia, the publishers, the hood it might might
34:20
might make more sense like that can make of.
34:22
I mean is that the devil tix? What market
34:24
is not huge but it's It is multi billions.
34:27
Uber well known article with Patrick
34:29
Collison on progress and science slowing
34:31
down and it's published at a
34:33
point say, right before M R
34:35
and A vaccines right before Gpc
34:37
for other developments. Or. How
34:40
well can we know? The progress of science at
34:42
any point in time isn't there often and everything
34:44
all at once. Effect. And
34:46
in fact those years we were building up
34:49
and investing in things that very suddenly than
34:51
flourished. Yeah
34:53
it's amazing to think about your different point of time
34:55
it's and which you could brighten trying right? the same
34:57
article sorry that the it for the principia you would
34:59
have been if the to to do the same thing.
35:02
That there's some question about. Like.
35:05
What? Certain. Types of institution.
35:07
My puzzle I see I I, I think
35:09
really, but the hired a bit in my
35:11
responses. gonna be something what? like I just
35:13
a guy is is not yet one hundred
35:16
percent clear, but I think it's very likely
35:18
to drive a lot of scientific progress over
35:20
the next few years and it's just a
35:22
cases and we're moving all about. So much
35:25
recognition and eventually also the actuators. That's the
35:27
way we operate in the world out into
35:29
these devices. Whale of a sudden it becomes
35:31
much more mutable and and hopefully improvable to
35:33
think the private sector wages as scientists. Are
35:36
a good proxy for progress and science. Science
35:40
it is. Iris is declining in value. You'd
35:42
think scientists would be paid less and less.
35:44
But. Over the last forty years. Mostly.
35:47
Those wages haven't fallen. I editing to get
35:49
Isaac Newton wasn't the the the richest person
35:51
to ever live but he probably did more
35:53
for human understanding than than and but there's
35:55
more of a market now thought the Isaac
35:57
Newton of today. Would. probably be pretty
35:59
wealthy Einstein, maybe, had
36:01
he done more media, right?
36:05
But he wouldn't have been wealthy for what
36:07
he did well. But still,
36:09
his wage would have reflected his
36:11
fame. He could have endorsed ski boots and
36:13
other things. Well, famously, he asked for $3,000
36:16
a year when he moved to the IAS,
36:18
and they gave him I
36:21
think he wasn't very good at negotiating. But just
36:23
say that the wages for private sector pharma
36:25
scientists, they seem to go up for quite
36:27
a while. And the drug
36:29
pipeline seemed slow. Should we have inferred from
36:31
that while we're building up to some big
36:33
things, some blockbusters, or not? Yeah.
36:37
This is a question for you, Tyler. It's not a
36:39
question for me. Well, that's what I'm asking, right. Good
36:41
reason to do so. I think I'm inclined to think.
36:44
And there's always this interesting balance. Actually,
36:46
AI is a really interesting example at
36:48
this point in time. There's this theory,
36:50
which has become widely believed by almost
36:52
everybody, that scaling is very important. Scaling
36:55
is a very capital-friendly story. So it
36:57
actually moves some of the power, the
36:59
negotiating power, from individual researchers, I think,
37:02
to centers of capital. But
37:04
it is just a story. I think it's quite interesting
37:06
that it, in some sense, gives
37:08
the individual researchers less negotiating
37:10
power, whether or not this
37:13
is going to eventually result in a
37:15
diminished ability to build personal brands and
37:17
then capture value from that. I don't
37:20
know. I'm really interested, actually, to see what
37:22
will happen over the next few years. It
37:24
used to be that the big companies published
37:26
a lot of papers very openly, and that
37:28
is gradually going away. And as
37:30
that goes away, it damages the individual researchers
37:32
because they're not able to build their brands
37:34
publicly in that way. They're not as easily
37:36
able to say, I am the person who
37:38
did whatever. But this is a small city,
37:40
doesn't everyone know? I had dinner
37:43
with a bunch of AI researchers last night.
37:45
They all seem to know each other's relative
37:47
importance. Yeah. I mean, and you
37:49
look at their salaries. So there's rumors. They're
37:51
doing fine. And these researchers can be offered
37:53
five to ten million dollars a year. Those
37:55
must be some of the highest science salaries
37:57
ever. And you're saying AI is a big
37:59
thing. such a big thing, it seems
38:01
to be coming down on the side of the
38:03
wages predicting something. I
38:06
think so much money is going
38:08
in and it's going in on the basis of brand
38:11
to some extent. We've hired such and such a
38:13
person who did such and such a thing and
38:15
if that can make your
38:17
company valuation go up by a few hundred million
38:19
dollars then offering them another extra million dollars a
38:22
year makes sense. But I don't have a grand
38:24
... This is all local
38:27
storytelling. It's not grand theory
38:29
of what's actually going on.
38:32
I haven't thought it through in enough detail
38:34
to have any confidence there. Now you're working
38:36
on what I think you call the vulnerable
38:38
world hypothesis, yes? That's what ...
38:40
Nick Bostrom, yeah, that's his term. What do you think
38:42
is the cost at which a
38:45
nuclear weapon could destroy a city? So
38:47
if that costs only $50,000, it seems
38:50
to me the world's in big, big trouble pretty
38:52
quickly. What's that
38:54
cost level where you get very, very nervous? So
38:57
if it's $10 billion, maybe things are
38:59
fairly safe. If it's $50,000, we're
39:02
done for. What's the
39:04
threshold? It's a cost and expertise,
39:06
but let's say the expertise is comparable. So somebody
39:09
who has $50,000 is probably able to get the
39:11
expertise as well. Right. Yeah,
39:13
that's not great. That's my fundamental
39:15
worry. With or without AI, that
39:17
just that cost becomes low. Honestly, I mean,
39:19
there's some question about what exactly you mean
39:22
by nuclear weapon. Is it portable? Is it
39:24
... It would render a mid-sized city uninhabitable.
39:27
Okay. For a few decades, at the
39:29
very least. It's sort of a multi-megaton, but not
39:31
... It would make the headlines. It would make the headlines.
39:34
Very small nuclear bomb will make the headlines if
39:36
detonated. Yeah, I mean, if we're
39:38
getting down to hundreds of thousands of dollars. The
39:40
thing is ... The issue is ... Nuclear
39:43
weapons are terrible, but they're not
39:46
civilization threatening directly. Sure, but enough
39:48
of these go off. Life as
39:50
we know it is over. They can be destabilizing.
39:52
They can certainly be destabilizing. That's ... It would
39:54
be like the fall of the Roman Empire, maybe
39:56
worse. Yeah, that's what it starts to
39:58
seem like. But
40:00
in what year do you think that the cost will
40:02
be low enough that that happens? I
40:05
mean at this point I don't have a good
40:07
sense. I suppose I'm actually more concerned about other
40:10
threats. But if this is a- Fire
40:12
safety is the obvious thing. I wouldn't call it
40:14
a certain threat, but if you simply think technology
40:16
will advance. Here's the thing, like
40:18
the Nuclear Nonproliferation Treaty means that we actually
40:20
have a lot of controls of the ability
40:23
to produce fissile material. But that's like League
40:25
of Nations, right? Ultimately? Yeah.
40:28
It's really coming out of that same sort of-
40:30
Yeah. We have this cartel of
40:33
whatever it is, 10 countries or something. It'll
40:35
fail when the cost is low. It will
40:37
eventually fail, but that's many decades away. But
40:39
you have reason to think we're going to last a thousand years
40:42
in a civilized state. Not every
40:45
person dead, but- I think getting
40:47
off planet Earth and establishing a
40:49
civilization elsewhere is very, very important.
40:53
Very hard for economic reasons. Yes. But
40:56
utterly crucial. Robots in
40:58
a sense make it harder because you could send
41:00
robots to Mars to do whatever might be economically
41:02
useful there. It means you never work hard on
41:04
having humans do it. Yeah,
41:07
that's true. I mean, we're pretty curious.
41:09
But the robot will take perfect footage.
41:13
Whatever is there, the robot will send back
41:15
to us. You'll have your- whatever is the
41:17
current version of Apple Vision Pro on. Okay.
41:19
Right? It will seem very realistic.
41:22
You're an economist. I'm a romantic, I think. It might
41:24
be the difference. But we'd have to settle them at
41:27
scale. 20 people on Mars limping
41:29
along. Oh, we're talking about like a million
41:31
people, not 20 people. You want- But
41:34
if we can do a million, we can do a billion, I would think. Sure, sure, sure.
41:37
I'm saying you want to get- it's still not going to be
41:39
self- What's the right term? It's not
41:41
going to be an autarky or whatever the right term
41:43
is. It's not going to be completely self-
41:45
Sustaining, yeah. Sustaining, but at a million people,
41:47
it's doing a lot of the-
41:49
it has a lot of the civilizational infrastructure. And so
41:51
I think that's the right sort of scale. Casey Hanmer
41:53
has a nice book. I think it's, you know, How
41:55
to Build a City of a Million People on Mars
41:58
or something like that. It's
42:00
way too optimistic in many of its assumptions,
42:02
but he's got the right scale. Economics
42:05
aside, what's the main scientific constraint that has to
42:07
be overcome? Is it gravity? Is it effective radiation
42:09
on the human body? Is it water? I mean,
42:11
to some extent, we're not going to know until
42:14
we go. There was this great experiment done a
42:16
few years ago where there was a pair of
42:18
twins. One went up
42:20
into space for a year. The other one stayed
42:22
on Earth. So that was the first time we
42:24
actually got to do a controlled, somewhat
42:26
controlled study where we see what the impact of being
42:28
in space for a long period of time does to
42:31
a human body. I mean, they
42:33
just discovered so many things. This is still
42:35
below the Van Allen belts as well. So
42:37
we don't know the answer to those questions. There's
42:41
a whole bunch of problems. The regolith on
42:43
Mars is terrible for human beings. I'm sure
42:45
that the low gravity is going to be
42:48
bad for them. What else
42:50
is this? There's shortages of nitrogen, fortunately, which
42:52
we don't really have on the moon. So
42:54
you're making me think civilization as
42:56
we know it won't last a thousand years. No,
42:59
I also have a lot of faith
43:02
in long run economic growth. Basically
43:04
at the moment for us to go to Mars is very,
43:07
very, very expensive given the return or
43:09
to establish a permanent human presence in
43:11
space. If we continue to have economic
43:14
growth, the relative cost is just going to keep going
43:16
down. At some point, it's actually not
43:18
going to be that difficult. Does
43:20
a vulnerable world mean near universal
43:22
surveillance? Unfortunately,
43:24
I think probably yes. Doesn't
43:28
that then become the great point of vulnerability? Oh,
43:31
absolutely. Yeah. It's
43:33
how you could ban universal surveillance from here
43:35
on out forever. Would you press that button?
43:37
No, I mean, I think, you know, the
43:39
history of justice to some extent, the
43:42
term surveillance, it's funny, you know, it's, it
43:44
has negative connotations. People, you know, they
43:47
think of Bentham and the Panopticon and
43:50
the archipelago and the Stasi
43:53
and all these things. But in fact,
43:55
our ability to supply justice is dependent
43:57
upon having a good understanding
44:00
of what has occurred in the present and in the
44:02
past. Maybe it
44:04
needs opaqueness as well. It's this optimal mix
44:06
of surveillance and opaqueness that
44:09
you actually have some latitude to break
44:11
certain laws, to misbehave, that keeps the
44:13
system stable, limits the abuses
44:15
of power, limits
44:17
how much power the powerful have over us.
44:23
There has to be some sort of
44:25
Madisonian point of view
44:27
where you're bringing the powerful
44:29
institutions into conflict with each other.
44:31
We do that very imperfectly at
44:34
the moment. Ideas like search warrants
44:36
and things like this. There's supposed to be checks and
44:38
balances. It seems like the
44:41
organizations which do the surveillance are too
44:43
powerful. They don't have a
44:46
strong enough checks on them. I
44:48
don't know whether, just as a practical
44:50
matter, the United States is capable of doing this
44:52
well. I'd be much more comfortable if it was
44:54
certain other countries. In a
44:57
strong AI future, where do the
44:59
economies of scale lie? Say
45:03
within your lifetime, not 500 years from now.
45:06
I'm not even going to point to what you're pointing to. Well,
45:10
we're all trying to figure out how AI will shape
45:13
the future, right? One model
45:15
is everything is supplied competitively, maybe
45:17
a bit like fast food today. I
45:19
suspect that's not true, but it could be true. There's
45:22
the oligopoly model. There's the one company races
45:24
ahead of the others, and then its own
45:26
AI does R&D at an accelerated pace, and
45:29
they stay ahead forever. Or
45:31
there's one country, one company, one
45:33
something, controls all the chips. Where
45:36
do you see the monopoly power evolving? Because it's
45:38
essential, I think, to predictions of the model. When
45:42
I talk to people who know much more than I do,
45:44
they all point at ASML as having been surprisingly
45:47
hard to duplicate. It's
45:49
the ability to sometimes do the lithography, but
45:51
do the lithography at scale, which seems to
45:53
be very, very hard. So we should be
45:55
long Netherlands? Probably. Yeah,
45:57
yeah. That would be an amazing conclusion. and
46:00
return to the Dutch Renaissance. So
46:03
it's like agriculture and lithography. And
46:07
drawing on 17- Human service things maybe.
46:09
That's right. Yeah, yeah, yeah. Yeah. This
46:12
is a return of the myth. Yeah, yeah. What
46:14
do you think of the Netherlands as a country? Oh,
46:17
I love it. I've never spent a- well,
46:19
actually, I spent a month there once. Who am
46:21
I kidding? That's a lot of time. That's
46:23
a fair amount of time in Leiden, actually. It's
46:26
a lovely place. It has many problems, of
46:28
course. I mean, their altitude is
46:30
not great. In some ways, actually, in some
46:32
ways, it's good. Yeah, an interesting sort of
46:34
a test case. They show
46:36
what a strong, determined civilization can do
46:38
in response to nature. Yeah, I like
46:40
the flatness of it. I like the
46:43
water being everywhere. Not everywhere, but most
46:45
parts, at least, of Western Netherlands. I
46:48
find that very attractive. This
46:50
is- stereotype I sometimes encounter people sort of
46:52
view it as being a little orderly.
46:55
I've heard people say it's dull. I
46:58
think some of the most interesting experiences of my
47:00
life were there. I
47:03
went to a- it was like a jamboree
47:05
in the field a
47:08
week- what was it? Five days long called Hacking
47:10
at Random in 2011, where
47:12
some people from Anonymous spoke, a whole bunch
47:14
of cryptographers spoke. It was really sort of
47:16
hacker culture. It was just intellectually
47:19
wild in the most interesting way.
47:21
It grew out of Dutch hacker
47:23
culture. I think there's a lot of- that
47:27
spirit of the Dutch Renaissance is
47:29
still visible. How
47:31
would you describe the quality of those conversations?
47:33
What were they like? Different than
47:35
what's in San Francisco? Oh, yeah. How are
47:37
they different? They're
47:41
not captured by capital to the same extent. The
47:43
conversations in San Francisco, particularly with younger people,
47:46
tend to be extremely idealistic and
47:48
often very pro-social. But
47:51
then later, there's this sort of negotiation that goes on where
47:53
they need access to capital. They need capital to make their
47:55
dreams come true. A certain amount
47:57
of compromise is made, although they also offer
47:59
it. and keep a lot of their original
48:01
sort of pro-social and idealistic character. In
48:04
the Netherlands, in those particular events, there
48:07
have been less of that. They also have less access to capital.
48:10
Yeah. If someone's going
48:12
to travel to the Netherlands, they have a tech background.
48:15
Like what should they do or what advice do
48:17
you have for them? How should they try to
48:19
learn more from the Netherlands? It's been years since
48:21
I've been there. So I'm not the right person.
48:23
But since you have good memory through spaced repetition.
48:25
Yeah, maybe. I mean, yeah,
48:28
I love going to the museums
48:30
in Amsterdam, just partially. Rembrandt is
48:33
maybe my favorite painter. That's a...
48:36
Yeah. It's hard actually to think of
48:38
anything else but I think of the Netherlands. Yeah. Other
48:41
than Rembrandt's late self-portraits, which I think are some
48:44
of the most extraordinary things I've ever done. Let's
48:46
say we all had better memories. How
48:49
big is the social gain there? Is there
48:52
any social gain at all? So you've been
48:54
an advocate of spaced repetition for improving your
48:56
memory. It works for medical students. It probably
48:58
works for languages. But are there
49:00
social gains, especially with AI coming?
49:03
I'm not actually... I wouldn't say I was an advocate. But
49:05
you do it. I do it. And
49:07
you teach other people how to do it. I get
49:09
benefits from it and some other people get benefits from
49:11
it. And I'm very enthusiastic if they do. And
49:14
if they don't, lots of people try it and are like,
49:16
this isn't working for me. And I'm like, well, stop doing
49:18
it. Same as if you listen to Bach and don't like
49:20
it, stop listening to Bach. To what
49:22
extent do I think... Yeah, there are these
49:24
interesting... There's a long sequence of
49:27
papers sort of trying to elucidate the
49:29
connection between deep sort of
49:31
practical expertise and the role of
49:33
memory, I suppose most famously. People
49:36
like Herb Simon and Andrew
49:38
Zarek and people like this have tried
49:41
to understand what relationship, if any. It's
49:43
a little bit murky. They all make
49:45
very strong claims about an
49:47
expert is somebody who's acquired sort of 50,000
49:49
chunks of information and things like this. They're
49:52
nice stories. They certainly seem to be born out, but
49:54
I don't know what the causal thing is. If I
49:56
talk to you about economics, you can tell me why
49:58
she's Not.... The getting number things
50:01
about economics but about a lot of different
50:03
things. but I don't know. Are you an
50:05
expert because you know those things are as
50:07
a downstream a something at a young. Is
50:09
it really downstream of something else? I'm sure
50:11
it's part of it. If you had a
50:13
magic memory, it might might help you. A
50:15
little bit but I suspect exceeds
50:17
downstream a something Else you determination
50:19
curiosity. Something like that to some
50:22
evidence that students learn better when
50:24
they take notes of what's being
50:26
said. Do. Field or something
50:28
for some people with memory bit similar
50:30
but until they have memorized it. It's.
50:32
Less real for them. I don't just mean that they
50:34
remember it more. But. The initial impact
50:37
somehow. Is created are defined
50:39
by the later act of memorization. Like.
50:41
People who take trips and until they photographed something
50:43
that I'm feel they seen it. Net.
50:46
Effect they probably didn't stamp named very likely. didn't
50:48
say it brass Yes. And he's part of a
50:50
reason why I take photos. I will look more
50:52
closely. That seems to be. Others I
50:54
will take notes, it's about. It is part of
50:56
the reason why I I despise repetitions. I it
50:58
provides me with another way of paying attention to
51:00
the world. I. Think the very
51:02
valuable rights like any any general purpose
51:04
strategy has which will cause you to
51:06
pay attention to the world. Is
51:09
incredibly valuable. When. So I
51:11
collect things like that I will be a white.
51:13
Why did I say yes to coming on the
51:15
podcast? A huge part of it is because I
51:17
know it's gonna make me pay attention in different
51:20
ways to I know you're going to ask me
51:22
questions that nobody else is gonna ask me. And
51:25
so for me, like the reason I despise repetition
51:27
and the reason I I will will come on
51:29
up a podcast with somebody like you who are
51:31
scary. Interesting questions? stick. It's kind of the same.
51:34
But in some other ways you're a fan
51:36
of non legibility has am I had other
51:38
margins and there's some tension. Because.
51:41
When you take the photo, when you remember
51:43
something, when you write it down. There's.
51:45
Less legibility and have. That
51:47
that there's no tension and all that you're
51:49
constantly expanding the legible. And and when you
51:52
do that, this is sort of penumbra of
51:54
eligibility that surround that moves. but it gives
51:56
you access to those other spices in it.
51:59
I got it. You travel in in
52:01
general, you make more of of the world's
52:03
culture. Your makes people much more legible to
52:05
yourself, but that and that expands what you're
52:07
able to to see as well. At the
52:09
at the edge of that. That's part of
52:11
the reason for doing it, but kind of.
52:13
the recent vote for wanting to make things
52:15
legible. What's. Underrated about travel. Other
52:17
than that, my god. almost different things. You
52:19
deny it's a saying this, but it's better
52:22
that somebody once said to me that travel
52:24
is the only education and it's really stayed
52:26
with me as expressing some deep trace, setting
52:29
mostly to the devoted, some credibly deep. absolutely.
52:31
What about his eight Billion People is the
52:33
only way to see that? Depth and breadth?
52:35
Yes, and it's just unbelievable. I mean, you'd
52:38
pick almost a random person. Anywhere you could
52:40
spend a great you with them is learning
52:42
things and that is can't and puts a
52:45
d that eight billion times. Yes,
52:47
it's very, very underrated. It's been very underrated
52:49
in my life. I haven't traveled nearly as
52:51
much as I said. Why? Is
52:54
that was at the Midway Aircraft
52:56
Carrier. Why is that so interesting?
52:58
There are many reasons. that and
53:00
San Diego right? it is. Yeah,
53:02
so for about ten years it
53:04
was probably the most dangerous object
53:06
in the world. A card. Nuclear
53:08
Weapons. I believe having five thousand
53:10
people on a specific you have
53:12
opened yet. people have been making
53:14
objects like that for centuries. many
53:16
centuries. The keep getting better at
53:18
it. So much Sir Bill knowledge
53:20
built into that environment is expresses
53:22
so much. Very. Deep expertise and then
53:24
you have for for a half thousand
53:27
people or completely dedicated to a single
53:29
purpose, they will care an enormous amount
53:31
about this purpose. They don't suffer like
53:33
a large organizations have all kinds of
53:35
blows and all kinds of problems. So
53:37
many of those problems are gone away.
53:39
Their that passionate because yeah we are
53:41
in a boat. It's not so easy
53:43
to empire builds like this. Real reasons
53:45
to trim Fab: You have amazing sort
53:47
of unity of purpose. What the captain
53:49
or what the admiral there has both
53:51
on on the boat say is that.
53:53
That goes, yeah, you're not arguing about
53:55
what the right corporate strategy is to
53:57
have incredible power. To you have incredible.
54:00
Relief In this purpose, it actually is
54:02
a high purpose in their case. you're
54:04
talking to some of the sailors, yet
54:06
they felt very strongly that as I'm
54:08
protecting our civilization that they cared about
54:10
a great deal the At Eight they
54:12
speak with so much pride about it.
54:14
So to my sort of the perfect
54:16
floating civilization in some regards is just
54:18
immensely interesting. If you something you
54:20
on throat and I quote. The. Great talent.
54:23
identifiers I know or know of
54:25
all seem very own idiosyncratic. the
54:27
rather like Missile and chefs. This.
54:30
Is getting a fact? the tension between the opaque
54:32
and the legible? What? Why do you think that's
54:34
true? that their idiosyncratic. Precise
54:38
because the boundaries of knowledge at any given
54:40
time tend to be idiosyncratic almost by definition,
54:42
but they haven't been commodities yet. There is
54:44
a best person at making superconducting circuits in
54:46
the world. There is a best person. You're
54:48
young, you're at a good Thomas showing. I
54:50
think what you're pissed a similar yes, yeah,
54:52
you and you read shelling. He realized that
54:54
that some of the things he did he
54:56
did fairly well, and he must have been
54:58
remarkable to talk to him, but he's a
55:00
little bit illegible. That's right. even though it's
55:02
a very very he was when you would
55:04
speak to him as well. I'll bet he
55:06
was sent. That's devout part of the value,
55:08
right? You're You're like oh yeah, this business
55:10
actually out on the edge of civilization and
55:12
and I think that that that people who
55:14
are good at identifying people who are able
55:16
to expand boundaries like that's like that they
55:19
they need to have some sense of of
55:21
of that edge. And. Here's a
55:23
question you wanted me to ask. You quote:
55:25
You initially were skeptical of Emerge and Ventures,
55:27
but you've changed your mind and become enthusiastic
55:29
about it. What caused the switch, and what
55:31
would you change about Emerge and Ventures? with
55:35
biggest single thing is just empirical i've made
55:37
a bunch of he be grantees of him
55:39
as encouraged a bunch of people to apply
55:42
some seem use is given grants to integrate
55:44
also we haven't funded in the way i
55:46
might have expected or something or some to
55:48
somebody i have actually forgotten who it is
55:50
and yet they clearly had some sort of
55:52
socialists lose quite anti libertarian ideas as we've
55:54
given them eg a grant and i thought
55:56
that might have been a mistake of course
55:58
typically toilets like these He's trying to figure
56:01
out, do they actually believe in this idea?
56:03
Do they actually really care? Then
56:07
you don't mind. They're certainly
56:09
not coming to the MacCadis Center to
56:11
carry forward the libertarian flag. Somehow,
56:14
I think seeing so many people
56:16
who are
56:19
doing very worthwhile things, which
56:21
have very
56:23
little institutional chance of support being
56:27
amplified, that I
56:30
care about a great deal. It
56:32
seems like EV is one of the places
56:34
that is doing the best. What
56:37
do you think we can do to attract more
56:39
non-legible but excellent people? Find
56:43
other people like yourself. They're going to be
56:46
like you in this abstract way, but actually
56:48
very unlike you in other ways. I
56:51
think it's about people like Stuart Brand and
56:53
people like that in the past who've just
56:56
been wonderful at talent identification, but they're not identifying
56:58
the same kind of talent as you. To
57:01
some extent, they become human marketplaces as well.
57:04
They're actually at a crossroads.
57:06
They're connecting people to opportunity. That's
57:09
a very special type of a person. Maybe
57:13
five to ten people in my life who seem
57:15
like that. Well, really strongly. Here's
57:17
something else you wrote in my quote. I
57:19
internalized a lot of Ivan Ilyich,
57:21
John Holt, A.S. Neil, and Paulo
57:24
Freire as a kid. What
57:27
did you mean? You were talking, I think, in the context
57:29
of agency, but how did that shape you? At
57:31
the time, as a 12 or 13-year-old, it mostly
57:33
probably made me insufferable to my parents because I
57:35
hated school already. That's
57:38
a good thing, right? It gave me a real
57:40
way of expressing that, and they dealt
57:42
with me very patiently. I
57:44
think over the long term, the most
57:47
important of those was, for
57:50
many years I would have said Ilyich. I'd
57:53
maybe still say Ilyich. Basically,
57:55
his point is about the question, It's
57:57
about the question of what's the relationship? Between.
58:01
The human beings and institutions and
58:03
help paternalistic or those institutions towards
58:05
for humans to indy schooling society.
58:08
He really makes the point that
58:10
effect schools do not treat children
58:12
as as human and to some
58:14
extent adult denies from the most
58:16
basic points of agency and just
58:18
thinking about that kind of a
58:20
relationship What what relationship Trudeau institutions
58:22
have to individuals was very very
58:24
important to me As a. Teenager.
58:27
And and sort of I mean through my entire
58:29
life is at why you didn't join Open a
58:31
I and Twenty sixty. Six
58:34
assists and you regret that decision. As
58:36
such, as a hub the Isis, I
58:39
did consider going as they were getting
58:41
started. It would have been an an
58:43
interesting life choice and a lot of
58:45
doubts about the wisdom of pursuing artificial
58:47
general intelligence which wouldn't are not at
58:49
all resolved then though it just feels
58:51
yes time as part of the reasons.
58:54
Honestly, really the main reason at that
58:56
point though? it's. It's this point about comparative
58:58
advantage. It was like oh I oh it's happening.
59:00
It's become very fashionable year if you gonna wake
59:02
up in the morning and it turns out that
59:04
some institution is could mad team to pay you
59:06
to do whatever it is he doing. you should
59:08
actually think about whether or not you're in the
59:10
right one. a business, a synth think unfortunately for
59:13
creative work disk and eighty correlation between have able
59:15
you doing is and and and what you've been
59:17
paid often. but it's kind of it. It's the
59:19
the anti economist point of. yeah it's not, it's
59:21
not right, but I like this a simple model
59:23
in which that is right. Actually of the simple
59:25
him that very much kind of an economist. Model
59:27
so I have felt for more than a
59:29
decade that I was he had spit it
59:31
become sort of a rant to doesn't Live
59:34
in two Thousand and twelve When I decided
59:36
to write said writing my book of at
59:38
Noon as they become sort of this unstoppable
59:40
force in the world or very difficult to
59:42
say to stop it was clear was gonna
59:44
attract more and more capital, more and more
59:46
people and Minnesota also to some extent I
59:48
just filled with i i should go do
59:50
something else if that's the case. And.
59:52
The next twenty years, where do you think you're
59:54
comparative advantage will be? Such
59:56
as tell you after twenty years a separate
59:58
you face the year at something Up at
1:00:01
Now writes that the probably a step over
1:00:03
for you have a problem is of course
1:00:05
it is very helpful for motivational reasons to
1:00:07
have answers to that but your answers never
1:00:09
turn out to what did the correct answer
1:00:11
was probably my ability to write and your
1:00:13
twitter biographies says and I quote searching for
1:00:15
the new minister. What? Does that mean
1:00:17
with suspect you. Just.
1:00:20
Trying to find the deepest possible experiences
1:00:22
in the world, in people and things
1:00:24
and ideas and places. And.
1:00:27
Final Question: What do you think it is that you
1:00:29
will learn about next? City
1:00:32
love to ask us questions I've had
1:00:34
you asked for all. Learn more. it
1:00:36
much more deeply about religion. Than
1:00:39
I have been the past and that involves.
1:00:41
Travel. Going to church, reading books, talking
1:00:43
to people. All of this thing involves all
1:00:45
of those things here. I'm gonna see that
1:00:48
the highest of here in Istanbul and in
1:00:50
the near future that's gorgeous. I wanted to
1:00:52
see so almost all my adult life and
1:00:54
I have never been. Have you been to
1:00:56
Amritsar? know? That to
1:00:58
me is the most religious feelings site
1:01:00
I've ever visited, so I would recommend
1:01:03
going there. And. It's not a
1:01:05
hard trip in any way, and I have never
1:01:07
been to India. You Must go To Amritsar. And
1:01:09
and the old cliche, something like when it comes
1:01:11
to religion, every Indians a millionaire. It's not really
1:01:13
true. But. I still think
1:01:15
India is the best place to go to think about
1:01:18
religion. Death? That makes a lot
1:01:20
of sense there. I'd love to go to
1:01:22
Jerusalem as well. I think the somewhat similar
1:01:24
reasons. But. They are. You Think about tension.
1:01:27
The. Religious aspects of tension. But you sigma
1:01:29
I think more about tension, that about religion
1:01:32
per se. And
1:01:34
it's very useful for that. That. Michael Nielsen.
1:01:36
Thank you very much for thanks so much Tyler. Thanks
1:01:41
for listening to Conversations with Tyler!
1:01:43
You can subscribe to the show
1:01:46
on Apple podcasts. Spot a Fi
1:01:48
are your favorite podcast that the
1:01:50
few like this podcast. please consider
1:01:52
giving us a rating and leaving
1:01:54
a review. This helps other listeners
1:01:56
find the show on Twitter. I'm
1:01:58
at Tyler Cowen. The show
1:02:00
is and go and conjures. Until
1:02:02
next time. Please keep listening and
1:02:05
learning.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More