WEBVTT

9cca12c0-0616-4693-8da0-0184f5d9a6d2-0
00:00:00.120 --> 00:00:00.720
Thomas, welcome.

f05f44fe-347c-4dcb-8d09-34300619507d-0
00:00:01.280 --> 00:00:05.099
If we could just start, what is
the link—it seems to be an ever

f05f44fe-347c-4dcb-8d09-34300619507d-1
00:00:05.099 --> 00:00:08.680
increasing link—between business
and mathematical sciences?

fcb32242-ffe2-47b5-8fd2-cfb1496e94bd-0
00:00:09.000 --> 00:00:13.480
So business looks at, is driven
by innovation, right?

2c473863-7b63-47d1-96e6-6c7a65d638cc-0
00:00:13.480 --> 00:00:14.800
So what is innovation?

ec712a54-480e-4922-8ed9-08689b0adc5a-0
00:00:14.800 --> 00:00:19.600
It's technology, and technology
seems to be speeding up, right.

c8664c45-399d-41c7-8fcf-00e10d477d29-0
00:00:19.600 --> 00:00:24.010
We know that's happening, and if
we can understand how technology

c8664c45-399d-41c7-8fcf-00e10d477d29-1
00:00:24.010 --> 00:00:28.220
works, if we can understand the
sort of mathematical structure

c8664c45-399d-41c7-8fcf-00e10d477d29-2
00:00:28.220 --> 00:00:31.962
behind it, then we can
understand why it's speeding up,

c8664c45-399d-41c7-8fcf-00e10d477d29-3
00:00:31.962 --> 00:00:35.838
speed it up more and and
influence the direction in which

c8664c45-399d-41c7-8fcf-00e10d477d29-4
00:00:35.838 --> 00:00:36.440
it grows.

dcea2ff5-13ae-4688-9afc-c3dffd434f84-0
00:00:37.280 --> 00:00:40.008
Tell me more, if you would,
about the work that you're doing

dcea2ff5-13ae-4688-9afc-c3dffd434f84-1
00:00:40.008 --> 00:00:42.601
at the London Institute for
Mathematical Sciences and how

dcea2ff5-13ae-4688-9afc-c3dffd434f84-2
00:00:42.601 --> 00:00:43.720
that relates to business.

df07bf48-ea43-4f7e-857d-f28392ce745b-0
00:00:44.040 --> 00:00:47.767
So we work in a bunch of things
ranging from string theory,

df07bf48-ea43-4f7e-857d-f28392ce745b-1
00:00:47.767 --> 00:00:51.682
which is a theory of everything,
to artificial intelligence of

df07bf48-ea43-4f7e-857d-f28392ce745b-2
00:00:51.682 --> 00:00:55.223
new discoveries, but also
something called the theory of

df07bf48-ea43-4f7e-857d-f28392ce745b-3
00:00:55.223 --> 00:00:56.280
human enterprise.

4c83fa2c-4430-43b4-8902-7f024b0f71df-0
00:00:56.880 --> 00:01:01.551
In particular, we've looked at
the mathematical structure of

4c83fa2c-4430-43b4-8902-7f024b0f71df-1
00:01:01.551 --> 00:01:05.840
innovation, of how technologies
work. And a technology,

25cd73e3-e07a-4e82-bfba-5eb539bb46c9-0
00:01:05.840 --> 00:01:09.384
and the simplest definition of a
technology comes from Brian

25cd73e3-e07a-4e82-bfba-5eb539bb46c9-1
00:01:09.384 --> 00:01:11.360
Arthur at the Santa Fe
Institute.

2b268625-5b05-400a-bb3c-53d5e83bbffd-0
00:01:11.680 --> 00:01:15.575
Technology is a combination of
more primitive technologies,

2b268625-5b05-400a-bb3c-53d5e83bbffd-1
00:01:15.575 --> 00:01:19.536
which harnesses the scientific
principle. And then when that

2b268625-5b05-400a-bb3c-53d5e83bbffd-2
00:01:19.536 --> 00:01:23.237
technology, it becomes useful
and becomes recognized, it

2b268625-5b05-400a-bb3c-53d5e83bbffd-3
00:01:23.237 --> 00:01:25.120
becomes a building block for 

5ea76e08-515e-47dd-b84c-516892c0e365-0
00:01:25.200 --> 00:01:29.240
even, you know, more advanced
technologies like a touchscreen.

7b09ac74-050f-464d-84bb-4fa3941b7510-0
00:01:29.240 --> 00:01:31.577
A touchscreen is made as a
complex technology, but

7b09ac74-050f-464d-84bb-4fa3941b7510-1
00:01:31.577 --> 00:01:34.328
eventually people think, oh,
I've got a touchscreen, that's

7b09ac74-050f-464d-84bb-4fa3941b7510-2
00:01:34.328 --> 00:01:35.520
one of my building blocks.

b67c1ec7-2c45-419d-b398-5855e6a023b3-0
00:01:35.520 --> 00:01:37.000
I got this, overall combined
those.

82b1bf93-3673-420b-ae21-215adbf52037-0
00:01:37.280 --> 00:01:41.160
So technology has this recursive
structure, you know. Primitive

82b1bf93-3673-420b-ae21-215adbf52037-1
00:01:41.160 --> 00:01:44.920
ones make more complex ones,
just make yet more complex ones.

a435b8e1-fdf0-423e-a267-d16060ca393f-0
00:01:45.160 --> 00:01:48.231
And it's that recursive
structure that we can understand

a435b8e1-fdf0-423e-a267-d16060ca393f-1
00:01:48.231 --> 00:01:49.040
mathematically.

05d85b03-1cad-4e5b-b334-1558191fc387-0
00:01:49.400 --> 00:01:49.920
Fantastic.

f6434b6a-75b0-4447-b6c9-0d2869697654-0
00:01:50.560 --> 00:01:54.265
Tell me if you would, about your
collaboration with BCG Henderson

f6434b6a-75b0-4447-b6c9-0d2869697654-1
00:01:54.265 --> 00:01:56.680
Institute, particularly around
innovation.

037b42ca-70a8-4a8e-aeb1-29293808a05c-0
00:01:57.240 --> 00:02:01.869
So that came about several years
ago when I got an e-mail from

037b42ca-70a8-4a8e-aeb1-29293808a05c-1
00:02:01.869 --> 00:02:04.000
someone called Martin Reeves.

866b48ba-cb31-4332-9f51-b74e303bf256-0
00:02:05.200 --> 00:02:08.783
He said, look, is innovation
like evolution or is it

866b48ba-cb31-4332-9f51-b74e303bf256-1
00:02:08.783 --> 00:02:10.880
something different on its own?

0b3e426e-3e2a-4ba2-9bdd-227b66720519-0
00:02:11.600 --> 00:02:15.460
And we met and started thinking
about that, started working on

0b3e426e-3e2a-4ba2-9bdd-227b66720519-1
00:02:15.460 --> 00:02:19.198
it and we found, well, actually
innovation and evolution are

0b3e426e-3e2a-4ba2-9bdd-227b66720519-2
00:02:19.198 --> 00:02:20.239
different things.

2f7cb364-cdf9-4374-968c-bdd736ae0e04-0
00:02:20.240 --> 00:02:24.080
Innovation doesn't imitate
biology, how life advances.

a119074f-2134-441b-bf2f-addb32e47acf-0
00:02:24.840 --> 00:02:27.040
It has its own sort of
mathematical structure.

b5d6076b-e2bd-4d7b-b454-3316e21732bc-0
00:02:27.360 --> 00:02:31.096
And we dug into that. And it's
quite beautiful, the mathematics

b5d6076b-e2bd-4d7b-b454-3316e21732bc-1
00:02:31.096 --> 00:02:31.680
behind it.

a519bc22-1db8-4f9c-a239-ca6a1fe2c9da-0
00:02:32.440 --> 00:02:36.669
And in fact, we found that you
can, understand what the factors

a519bc22-1db8-4f9c-a239-ca6a1fe2c9da-1
00:02:36.669 --> 00:02:40.833
are that speed it up and you can
look to different sectors and

a519bc22-1db8-4f9c-a239-ca6a1fe2c9da-2
00:02:40.833 --> 00:02:44.600
influence how different sectors
grow relative to others.

6548fe78-77ab-4466-8bc9-7d0f1ac78b63-0
00:02:45.360 --> 00:02:48.374
How is that—and bear with me in
layman's terms here—how is that

6548fe78-77ab-4466-8bc9-7d0f1ac78b63-1
00:02:48.374 --> 00:02:51.200
applicable in the real world,
what you've been discovering?

63fc38ec-9cf2-4f18-a2fe-18b171f02593-0
00:02:52.040 --> 00:02:55.908
So you know a lot of people
think that the future happens to

63fc38ec-9cf2-4f18-a2fe-18b171f02593-1
00:02:55.908 --> 00:02:59.903
us. But Peter Thiel said to a
recent debate, he says you know,

63fc38ec-9cf2-4f18-a2fe-18b171f02593-2
00:02:59.903 --> 00:03:03.582
but you and I are creating it
right now. And you know the

63fc38ec-9cf2-4f18-a2fe-18b171f02593-3
00:03:03.582 --> 00:03:07.577
direction that technology takes
isn't just something you know,

63fc38ec-9cf2-4f18-a2fe-18b171f02593-4
00:03:07.577 --> 00:03:09.480
like we're watching the movie.

871b145d-e2f4-43da-b692-c116e904a016-0
00:03:09.480 --> 00:03:10.720
We're actors in the movie.

e38830d3-dac2-42a1-854a-c1f61124da53-0
00:03:11.080 --> 00:03:16.066
So, you know, the technologies
that we push, you know the way

e38830d3-dac2-42a1-854a-c1f61124da53-1
00:03:16.066 --> 00:03:20.891
that we think of how, of the
modules that make up, the ones

e38830d3-dac2-42a1-854a-c1f61124da53-2
00:03:20.891 --> 00:03:25.475
that we promote, have a profound
effect on the future of

e38830d3-dac2-42a1-854a-c1f61124da53-3
00:03:25.475 --> 00:03:26.360
technology.

bb4c992c-486c-45a7-aa8b-eea0aac951ab-0
00:03:26.960 --> 00:03:30.813
So, you know, small choices we
make now get amplified as

bb4c992c-486c-45a7-aa8b-eea0aac951ab-1
00:03:30.813 --> 00:03:34.600
technologies, you know, give
birth to new technologies.

ac5a99c5-59cf-4b7c-beca-1f944ee39d6c-0
00:03:35.080 --> 00:03:37.480
You've also been looking at AI
assisted discoveries.

20c37bdb-9036-4600-8a5b-a014ac7716e5-0
00:03:37.480 --> 00:03:39.120
Tell me about, tell me about
that.

61b53ab1-7abc-4d76-b9e4-ecdb040220cd-0
00:03:39.600 --> 00:03:41.320
So a lot of people.

b119072e-3fa3-4e74-b980-51ab8dca4347-0
00:03:41.640 --> 00:03:43.480
So first of all, I'm a great AI
skeptic.

3cbbd29f-f18d-4dfd-9749-7ee004054ac6-0
00:03:43.520 --> 00:03:46.440
So in general, I think AI is
like Explain? Well, I think, you

3cbbd29f-f18d-4dfd-9749-7ee004054ac6-1
00:03:46.440 --> 00:03:49.549
know, what's amazing about AI is
not how much it's done, it's how

3cbbd29f-f18d-4dfd-9749-7ee004054ac6-2
00:03:49.549 --> 00:03:50.680
little it's done, right.

afa5c665-c2f1-4e0f-8f4f-adbc03ebc8d0-0
00:03:50.680 --> 00:03:54.040
I mean, basically AI right now
is automated curve fitting.

0e02fad6-8c7b-4c9f-b25e-a1dfe27402e3-0
00:03:54.440 --> 00:03:58.711
But here's where I think it's
going to make its first steps in

0e02fad6-8c7b-4c9f-b25e-a1dfe27402e3-1
00:03:58.711 --> 00:04:00.000
the creative world.

7d281fad-247a-49fe-9a43-d6f5e123f790-0
00:04:00.520 --> 00:04:02.320
So this is something a lot of
people think about.

88aec95b-a437-43a8-826f-d8709d15edbb-0
00:04:02.320 --> 00:04:03.520
Well, AI is not creative.

f26d5414-9336-49a9-aa8c-fff7ab6515c2-0
00:04:03.520 --> 00:04:04.800
It's not really creative, right?

775d49e0-d333-4ec0-96d5-da902e346b62-0
00:04:04.800 --> 00:04:07.360
And maybe it can, you know,
construct silly poems.

5b2a80e5-b51a-4637-a081-5a35ebf1a02f-0
00:04:07.720 --> 00:04:12.343
But I think, counterintuitively,
the first time we're going to

5b2a80e5-b51a-4637-a081-5a35ebf1a02f-1
00:04:12.343 --> 00:04:16.012
see creative AI is in
mathematics and theoretical

5b2a80e5-b51a-4637-a081-5a35ebf1a02f-2
00:04:16.012 --> 00:04:16.600
physics.

31bc2c7e-4995-4f6b-9c21-5401daf9b7b2-0
00:04:16.800 --> 00:04:20.090
Now you may think that's nuts,
but there are two reasons for

31bc2c7e-4995-4f6b-9c21-5401daf9b7b2-1
00:04:20.090 --> 00:04:20.360
that.

f5db4c62-f7d2-43a5-a919-cf99e33f4255-0
00:04:20.880 --> 00:04:24.755
Well, the the main reason is
that and this sounds, this is a

f5db4c62-f7d2-43a5-a919-cf99e33f4255-1
00:04:24.755 --> 00:04:28.440
mad thing, that in mathematics
there are no coincidences.

8198ae60-eb68-42dd-95c8-cabb67ee881f-0
00:04:29.320 --> 00:04:32.120
And I mean that there are no
false positives, there are no

8198ae60-eb68-42dd-95c8-cabb67ee881f-1
00:04:32.120 --> 00:04:32.880
false negatives.

1ce357d8-fee5-499f-a8b4-c6648b3c3885-0
00:04:33.440 --> 00:04:35.943
Whereas in, you know, in
ordinary life there are all

1ce357d8-fee5-499f-a8b4-c6648b3c3885-1
00:04:35.943 --> 00:04:38.918
sorts of false positives like.
You know, is this, you know, is

1ce357d8-fee5-499f-a8b4-c6648b3c3885-2
00:04:38.918 --> 00:04:41.799
this gene connected to that
disease or is that just a fluke.

379c446a-edc2-48d0-9e55-840de3c716e1-0
00:04:42.000 --> 00:04:45.131
So there are no flukes in in
theoretical physics and

379c446a-edc2-48d0-9e55-840de3c716e1-1
00:04:45.131 --> 00:04:45.840
mathematics.

e31665a0-50df-4bec-876a-96fdf1a1f7be-0
00:04:45.840 --> 00:04:48.937
So AI, which is really good at
detecting patterns, when it

e31665a0-50df-4bec-876a-96fdf1a1f7be-1
00:04:48.937 --> 00:04:51.720
detects those patterns, those
patterns have meaning.

6d26b00c-7ca6-404c-a9b0-92ff85c26c83-0
00:04:52.680 --> 00:04:53.520
And that's creative.

0afc064e-3d0e-4924-b000-cc773bee41d2-0
00:04:54.600 --> 00:04:57.434
Well, I mean, some people would
say pure mathematics is it's the

0afc064e-3d0e-4924-b000-cc773bee41d2-1
00:04:57.434 --> 00:04:58.960
height of human creativity,
right?

f1bb25e3-23e2-47cf-9c29-b1a6534ff93f-0
00:04:59.000 --> 00:05:02.000
Like, you know, how do we form a
new conjecture?

a2a2aa5a-9c6b-4009-9a92-3f47c97635fe-0
00:05:02.120 --> 00:05:05.780
You know, how do you, how did
Gauss prove the density of the

a2a2aa5a-9c6b-4009-9a92-3f47c97635fe-1
00:05:05.780 --> 00:05:07.040
prime numbers, right?

d53ab491-1d34-4e63-9b71-8d3db68b8434-0
00:05:07.040 --> 00:05:10.273
Well, there's a lot of new
properties of the mathematical

d53ab491-1d34-4e63-9b71-8d3db68b8434-1
00:05:10.273 --> 00:05:12.280
world and the physics world,
right?

8fc87c6a-8e39-44e1-93de-10a263f95cd3-0
00:05:12.280 --> 00:05:13.760
I mean, physics is broken.

698d672a-0ffe-4c49-a98a-7b896d1a4941-0
00:05:13.760 --> 00:05:15.080
We're missing big parts of it.

ef5966bb-ad91-4dba-ba87-62a0f2f2dec5-0
00:05:15.240 --> 00:05:16.920
Who's going to find those new
things?

a6141c68-388f-4b31-aa49-d2b91b0f4dba-0
00:05:17.000 --> 00:05:20.038
Where are those new conjectures,
those new ideas going to come

a6141c68-388f-4b31-aa49-d2b91b0f4dba-1
00:05:20.038 --> 00:05:20.280
from?

1ac34773-2603-498c-8a12-6899ffb3d50d-0
00:05:20.680 --> 00:05:23.160
And an AI can guide us, right?

ddcb29d9-68be-4e05-ab17-267f2b3d49c6-0
00:05:23.160 --> 00:05:25.765
It's not, it can't do it for us,
but it can guide us in the right

ddcb29d9-68be-4e05-ab17-267f2b3d49c6-1
00:05:25.765 --> 00:05:26.160
direction.

89c5e1e2-7e0e-4ce3-a490-afe211208092-0
00:05:26.160 --> 00:05:30.354
So we sometimes call this AI
assisted discovery and this is

89c5e1e2-7e0e-4ce3-a490-afe211208092-1
00:05:30.354 --> 00:05:33.780
something that the Bruce
Henderson Institute has

89c5e1e2-7e0e-4ce3-a490-afe211208092-2
00:05:33.780 --> 00:05:34.480
supported.

f52bbc18-b4f0-42d5-97ca-a6fdd0e3ad80-0
00:05:34.480 --> 00:05:39.497
We've had you know—actually, our
group in London is is leading

f52bbc18-b4f0-42d5-97ca-a6fdd0e3ad80-1
00:05:39.497 --> 00:05:43.240
the world in AI assisted
conjecture formation.

d628c7dd-8171-4390-9581-71a5512c6e4a-0
00:05:43.680 --> 00:05:44.920
Thomas, absolute pleasure.

99f9a0e7-9f5f-4a30-ab0e-c5c8468b4c42-0
00:05:44.960 --> 00:05:45.720
Thank you so much.

282c9c9a-4238-43b4-8901-2c9050329def-0
00:05:45.920 --> 00:05:46.320
Thank you.