Alan Turing’s Final Theory Was About Leopards

0:00 I've got a question for you, Michael.

0:01 Okay.

0:02 It's a bit of a It's a bit of a weird one, right?

0:05 Good.

0:05 [snorts] It's sort of a philosophical question in a way,

0:08 but maybe also maybe also deeply scientific.

0:12 [snorts] If you start out with an embryo

0:13 and it's just this perfectly symmetrical sphere of cells,

0:19 how does it ever decide where the head goes, right?

0:22 Why doesn't it just Well, how does it ever end up with any structure?

0:26 I mean, when it when there's already a little

0:28 bit of structure there, like I get it.

0:30 Maybe there's hormones that come from the head

0:32 cells that let the neck start forming,

0:35 but when you're a blastocyst, like just a ball, a symmetric ball of cells,

0:41 how does it decide, all right guys, final positions, you're the butt,

0:46 you guys are the toes, you're going to be the brain, get to work?

0:50 Ready, steady, go.

0:51 Which way's up and down?

0:52 Does it have to do with like local gravity or the parent's body?

0:56 I don't know.

0:57 How does it seed those?

0:59 How does it seed those, exactly.

1:00 But then also, I mean I mean I you

1:02 said if there's a little bit of structure there, maybe it makes sense.

1:05 But at the same time, when you look at physics, if you take I don't know,

1:09 like a glass of water for instance and you put a little drop of ink in it,

1:13 the process that happens there is diffusion

1:15 and diffusion is like the destroyer of patterns.

1:19 You don't get structure from physical processes.

1:23 So, what is it about biology that means that you end up with structure?

1:26 It's sort of It's a bit of a puzzle.

1:28 It's a bit of a puzzle.

1:29 Well, yeah.

1:30 It is It's really blowing my mind cuz I've seen a blastocyst,

1:34 a human one, through a microscope when it's like eight cells.

1:38 And you're like, wow, cool.

1:40 But how in the world does it start assigning roles to each cell?

1:46 And how do you make sure that the two cells on opposite ends

1:48 don't both decide that they're going to start forming the the brain tube?

1:51 Mhm.

1:52 Right.

1:53 Am I going to find out today?

1:54 I think you are.

1:55 I think you are.

1:56 You've got I'm going to There's going to be an answer.

1:58 There's going to be an answer.

1:59 Maybe not to your I mean, look, I'm going to go for slightly simpler animals.

2:02 I'm not going to not I don't need to get too excited.

2:05 I've only got the answer for slightly simpler animals.

2:06 But I am going to talk about how you possibly end up with structure in biology.

2:11 And the answer to this incredibly difficult question,

2:16 which I mean, was given by a really extraordinary person.

2:21 [music] This episode is brought to you by Cancer Research UK.

2:27 Here's something strange.

2:29 Your DNA contains more ancient viral fragments than genes.

2:36 The genes that build our cells make up only 2% of our DNA.

2:41 And for years, that is what scientists focused on.

2:44 They treated the rest, the ancient viruses and stuff, as junk.

2:48 But now we know that that hidden majority, sometimes called the dark genome,

2:53 influences how our biology works and how diseases like cancer behave.

2:58 It's a reminder that progress rarely comes as a single breakthrough.

3:03 It builds gradually.

3:04 Cancer Research UK plays a central role in that progress,

3:08 supporting decades of research into over 200 types of cancer,

3:13 work that's helped double survival in the UK over the past 50 years.

3:17 For more information about Cancer Research UK, their research breakthroughs,

3:21 and how you can support them, visit cancerresearchuk.org/therestisscience.

3:28 [music] There's a kind of strange hero who enters the story,

3:36 who ends up actually explaining a lot of the structure in biology.

3:39 It's not not somebody I think you would immediately expect.

3:42 It's Alan Turing, who Who's Turing?

3:45 best known for inventing the computer, for cracking Nazi cryptography.

3:51 He's the guy who came up with a lot of this stuff.

3:54 No kidding, cuz I associate him with like steampunk-y computers,

3:59 metal circuits, definitely not flesh and and and love.

4:06 [laughter] Definitely not flesh and love.

4:07 Your lucky, lucky wife, Michael.

4:10 Well, flesh flesh and and even procreation I don't associate much with.

4:14 totally.

4:15 No, totally.

4:16 Um but he was struggling with exactly this question.

4:19 This is like 1952, okay?

4:20 So, at this point in time, he is this absolute,

4:24 you know, god-like figure in the in the cryptography community,

4:28 in the sort of secret services,

4:31 but the rest of the world doesn't know who he is.

4:32 They just think he's this like crackpot old professor up in Manchester.

4:36 And he's thinking about exactly this question.

4:37 It's like, how is it possible that diffusion,

4:40 which is the process that sort of governs how liquid-y like things move,

4:45 how is it possible that when you put that in a biological being,

4:49 suddenly there's it's not following the same rules.

4:52 You get these patterns rather than being destroyed by destroyed by diffusion.

4:57 And he was looking in particular like the skin of animals.

5:01 So, leopards and spots, cows and their patches.

5:05 So, how does biology end up with these these kind of these features, right?

5:09 Not just a head and a tail,

5:10 but also front and back and left and right and these complex patterns.

5:13 You get like zebra stripes and and leopard spots.

5:16 And then he had this genius idea.

5:17 He was like, okay, well, what if it's not just one thing that's diffusing.

5:21 Okay, what if you have two things that are

5:23 diffusing simultaneously that are fighting against one another?

5:27 What what would happen then?

5:29 Okay.

5:29 So, I'm going to give you I'm going to give you a description.

5:32 Going to give you an analogy of what he was describing.

5:36 You kind of have to go with me a little bit on this analogy.

5:38 I've got I've got a couple ready for you,

5:39 but but just go with me on this analogy.

5:41 So, all right, imagine you've got this petri dish of water

5:45 and I'm going to put a drop of ink in there.

5:47 Normal diffusion, it would just spread out and it would all be gray, okay?

5:50 And it would be random.

5:51 And it would be random, exactly.

5:54 But this is special ink.

5:56 This is like a biological ink, okay?

5:58 And so, it can copy itself.

6:00 So, ink makes more ink, all right?

6:04 But the ink is like it's slow, it's thick, it's gloopy.

6:08 It diffuses, but it's really going to take its time about it.

6:11 But the key thing is that it's like it's making more ink as it goes, right?

6:14 Like a bacteria would, for instance.

6:16 Right.

6:16 Now, if it was in in there on its own, if it was just that ink making more ink,

6:20 then the whole thing would be black very quickly.

6:23 But what if instead of just making new bits of ink,

6:27 this ink also spits out some eraser?

6:31 Eraser?

6:32 Eraser, like ink eraser.

6:35 Okay, so this is like a very strange stuff ink.

6:39 As I said, you have to go with me, all right?

6:41 It's very strange stuff ink, but it can spit out All right, with me, okay?

6:44 It spits out both versions of itself and the thing that can kill itself.

6:50 So, does this mean like it like cell division?

6:52 Like it splits into more ink and eraser in the same localized point?

6:59 Yes, think of it that way.

7:00 Yes, Yes.

7:01 Same localized point,

7:02 it gets more ink so that the amount of total ink increases,

7:06 but you also get eraser, which can delete the extra ink.

7:09 Okay, this is like super theoretical.

7:12 So, if it didn't diffuse,

7:13 if it just sat there and it was spitting out ink and eraser, ink and eraser,

7:16 ink and eraser, the two would cancel each other out

7:18 and you would just have this completely clear petri dish.

7:21 Yeah.

7:22 But what Turing was thinking was, okay, well,

7:25 what if the ink is really thick and gloopy and slow,

7:29 but the eraser is thin and slippy and can diffuse really quickly.

7:34 Is there a way that this eraser could diffuse faster than the ink,

7:39 spread out across the petri dish,

7:41 and then end up creating this little moat around the ink as it forms?

7:48 Effectively like, could the ink build its own cage, essentially?

7:54 Right.

7:54 If that were the case, and he he did this all mathematically, right?

7:57 And he kind of demonstrated that you can have

7:59 this where there is a moment where there's an equilibrium where

8:02 the amount of ink being made exactly matches the rate

8:05 at which the eraser is wiping out the borders, okay?

8:08 So, it's not like a stalemate, it's not like, oh, it just stops.

8:14 If you zoomed in, the ink's still having

8:16 like still dividing into more ink and more eraser,

8:19 still kind of reacting with what's going on a Now, okay,

8:22 I I accept that is the most mathematically

8:24 accurate version of what Turing was thinking of.

8:26 But I accept it's a bit abstract.

8:28 So, I've got a slightly more human example for you, if you like.

8:32 Okay.

8:32 Well, first of all, tell me when

8:34 when in history was Turing having these thoughts?

8:37 This is '52, 1952.

8:39 1952?

8:40 So, not even that long ago.

8:42 Cuz I'm imagining these great ideas that someone can just have in an armchair.

8:47 And I'm thinking of how Einstein did that, too.

8:48 He was like, what would it be like if I was riding on light?

8:53 And he's just sitting, you know, in his chair thinking.

8:55 And here's Turing going, zebra stripes.

8:59 Let me think about this.

9:00 Let me think about eating an eraser diffusing around in a petri dish.

9:04 All right.

9:05 So, very cool.

9:06 But tell me this like more human version or this more biological one.

9:11 All right, because you you do actually get stuff quite

9:14 often that can make its own make copies of itself,

9:18 make versions of its own self, and also the thing that kills itself.

9:22 I was going to say there's got to be chemical reactions that are similar.

9:25 Well, okay.

9:26 At the time, nobody thought that there were.

9:29 Everyone thought that you don't get, you know,

9:31 it doesn't make sense with in in terms of entropy.

9:33 But there are analogies.

9:35 So, forest fires is a really good analogy of this.

9:38 Because if you think about it,

9:39 when you get a little bit of fire, actually fire makes more fire.

9:43 But [snorts] also, if it's in a forest setting, the more fire there is,

9:47 the more likely that you are to get the thing that kills the fire,

9:50 which is helicopters carrying water, okay?

9:53 Right.

9:54 So, The existence of something creates more of that same

9:58 thing and also the thing that kills it.

10:01 That's right.

10:02 So so the ink and eraser analogy,

10:04 if you imagine that you are looking at a forest

10:06 or sort of top down on a forest, okay?

10:08 You for some reason you get a little bit of fire here and there and it starts

10:12 spreading and spreading and spreading and then you

10:14 get in the helicopters who can move much faster.

10:16 The the forest fire is is spreading but it's

10:18 spreading quite slowly and the the the helicopters

10:21 can move much quicker and they can

10:23 encircle this forest fire and basically create

10:26 a moat around it so that in the end you have patches of fire that are

10:31 burning that are being controlled from the outside

10:35 by these helicopters while the fire is burning inside.

10:38 Yeah, which is which is literally what happens.

10:41 Which is literally what happens.

10:42 Fire perimeters are set up, yeah.

10:44 Exactly.

10:45 So but Turing didn't have this more

10:47 concrete analogy because there weren't helicopters.

10:49 Well, [laughter] I don't know.

10:50 When was the helicopter invented?

10:52 Well, Da Vinci came up with a version of the helicopter

10:55 if you want to go all the way back.

10:56 When were they used for fire fighting, I guess?

10:58 Yeah, good point.

10:59 Okay, there were helicopters in Turing's time.

11:02 I don't think he was I don't think

11:03 that was I don't think that was going through him.

11:05 I think he was happy to stay abstract, yeah.

11:07 One analogy that people did use around the time

11:10 around the '50s was of rabbits and foxes.

11:12 I sort of I don't I object to this analogy for other reasons but mainly

11:18 because the ink is thick and gluey and slow and rabbits are quite fast, okay?

11:22 Yeah.

11:23 But imagine for a moment the rabbits are not fast.

11:25 [laughter] Okay, I'm doing that.

11:27 I'm imagining that rabbits are slow.

11:28 Okay, and rabbits sort of stay near the burrow, okay?

11:31 Rabbits make more rabbits but they also allow foxes to exist, okay?

11:38 So the more rabbits you get,

11:39 the more rabbits you get but also the more foxes you get.

11:42 So so rabbits are effectively in a in a sense

11:45 creating both more of themselves and the thing that kills them.

11:48 Right.

11:48 Okay.

11:48 Here is the idea, if you can accept that the rabbits might be slow

11:51 and sort of stay near the burrow but that the foxes like can move around

11:55 much faster then what you end up with what you can end up with is

11:59 this dynamic equilibrium where you have the number

12:02 of rabbits that having more rabbit babies,

12:05 the rabbit babies are sort of creating this balance

12:08 with the amount of foxes that are eating them.

12:10 So you end up with these stable populations,

12:12 these little pockets of rabbits that are kind of surrounded by foxes.

12:17 So little pocket over there and a little pocket

12:19 over there and a little pocket over there, okay?

12:21 And essentially what you need

12:22 for this system is called a reaction-diffusion system.

12:25 So diffusion because you've got the foxes

12:28 or the helicopters or the eraser that's kind of diffusing through the system

12:32 and the reaction because you've got you know,

12:34 the the rabbits and the foxes are reacting together,

12:37 the fire and the water, whatever it might be.

12:39 But it relies on these two things, the activator,

12:42 something that makes itself but it's also slow

12:46 to diffuse and makes the thing that also kills itself.

12:50 So Turing was like, oh, you know what?

12:51 I reckon let's just play around with this.

12:53 Like let's just see what happens with this.

12:54 Let's just like write some equations, like see what happens.

12:58 You know, he came up with this this this mathematical system [snorts]

13:01 and it was like it's sort of like a mathematical

13:03 version of local love and long distance hate, okay?

13:07 So like the sort of the inhibitor spreads really fast and quickly

13:11 but in a in a local setting you can get a cluster where things are quite happy.

13:16 And he didn't just scribble this on a chalkboard, by the way.

13:18 He was like because he had invented the computer [laughter]

13:25 he saved a lot of many people.

13:27 He had access at Manchester University to this really really crude computer.

13:32 It's called the Ferranti Mark 1.

13:35 And so he wrote all of these computer programs

13:38 to simulate what would happen in this environment, right?

13:41 If you've got these two different chemicals or two

13:44 different processes that are fighting against each other in space.

13:47 And what he would do is he would start

13:48 off with like a soup of kind of random noise,

13:51 like little fluctuations here and there.

13:53 And then he would watch as the computer

13:57 spat out what looked like the perfect image

14:02 of the spots that you get on a leopard or on the or the stripes of a zebra.

14:08 What the computer was spitting out was I mean qualitatively

14:14 identical to what you end up seeing in animal skin, okay?

14:18 Wow.

14:19 Which is funny, right?

14:20 That it's like he's literally just having this in his brain.

14:23 So even if you start with this perfectly uniform soup,

14:28 a kind of gray embryo where nothing interesting is happening,

14:32 if you just get a tiny little variation and you've got

14:35 this process that's sitting there waiting to to kick in Turing

14:40 basically proved mathematically that you can get this order from chaos

14:44 just purely through the laws of physics and mathematics, okay?

14:47 Wow.

14:48 He also showed that the geometry makes a difference.

14:51 So he demonstrated that if you have like a big space that you've got

14:56 this process going on in a big space like the belly of a leopard,

14:59 for example then you end up with with spots and if you make it a much

15:03 narrower space then you end up with stripes like on the tail of a leopard.

15:08 And if it's too small altogether, you don't get any patterns at all, right?

15:12 So like on mice, for example,

15:14 they're usually solid colored where where whereas when

15:16 you have like larger cats, they have like much Gosh, that's true.

15:19 Yeah.

15:20 Okay, so this is all like, you know,

15:22 this is all like nice and nice and theoretical.

15:24 Would you like to know how the biologists to I would love to.

15:28 I mean poor Turing, right?

15:30 Like he's got this unbelievable glory from World

15:32 War and he can't tell anyone about it.

15:34 Poor Turing, there's a lot of reasons we could say poor Turing.

15:39 The guy died in 1954.

15:41 Mhm.

15:42 Born in 1912.

15:44 Right.

15:45 So I'm listening to this story.

15:46 This guy uses a computer to simulate the form the formation of what

15:51 turns out to look just like the spots and stripes on animals.

15:54 Mhm.

15:54 Why?

15:55 Because he invented the computer.

15:57 If he hadn't have died in the '50s, if he lived to be 100,

16:00 this guy could have been watching Chocolate Rain on YouTube.

16:04 Completely.

16:05 In his one lifetime,

16:06 he could have seen his invention become what it was by like 2012.

16:10 Yeah.

16:10 Yeah, absolutely.

16:11 He could have been like cracking codes during World War with his newfangled

16:16 computer and then lived long enough to have shared a Kony 2012 meme.

16:22 When was he born?

16:24 1912?

16:25 1912.

16:27 I mean if he'd lived to be 100,

16:29 he would have seen that really famous moment in artificial

16:34 intelligence where they started categorizing pictures of dogs and cats,

16:39 which was really I think that the beginning of the the breakthrough

16:42 of what we've seen what we've seen happen over the last 15 years.

16:45 I mean he could he could have been alive to see that.

16:48 Imagine that.

16:48 And here's the most unbelievable one.

16:50 If he lived to be 100, he could have watched Vsauce videos.

16:56 And we all know he would have.

16:58 We all know he would have.

16:59 Life is full of tragedy.

17:00 He would have been there on YouTube saying first First.

17:06 We know.

17:07 first.

17:07 And he was first.

17:08 He really well, lovely.

17:10 But he really was first.

17:12 [laughter] We will maybe do one on that another day.

17:14 Thing is, right?

17:15 So meanwhile, Turing, no one knows any of this stuff.

17:18 No one thinks he's a big deal.

17:19 He publishes this paper and the biologists are like Yeah, whatever.

17:25 [laughter] Are you joking?

17:26 Well done you with your little math parlor trick.

17:28 This is not serious science.

17:29 You're Why why did they not think it was serious science?

17:32 Because it was too mathematical?

17:34 Partly because it was too mathematical but I think also partly this is

17:37 about the same time that people are discovering the structure of DNA, you know?

17:41 This is like the point where biology is is is absolutely obsessed

17:45 with this idea that there is a genetic blueprint that, you know,

17:49 for a leopard to have spots, there must be a spot gene that's telling telling

17:53 the skin when to turn black and when to turn orange.

17:57 It's like it must be everyone was obsessed

17:59 with this very mechanical kind of top-down view of life.

18:02 And and Turing here is this mathematician,

18:03 he's got no background in but he's never dissected a frog in his life,

18:07 you know what I mean?

18:08 And he barges in and he says, oh no it's just a puddle of chemicals

18:12 and then fluid dynamics does the rest, you know?

18:14 It's just diffusion.

18:15 There's nothing going on.

18:17 So they were like, you know, this is this is not this is not proper science.

18:21 This is like looking at a cloud and saying it looks like a dog.

18:23 Yeah.

18:24 For sure.

18:25 You know what?

18:25 Like he was he was saying all of this at the time when yeah,

18:30 people the paradigm was very much about

18:33 we're actually more robotic than we think.

18:35 Not just genetically but even in the mind, right?

18:39 The behavioralists were the key psychological field that we just learn things

18:44 and then we're conditioned to behave in certain ways and that's it.

18:47 There was no room for creativity, there was no room for even free will,

18:53 let alone chaos to bring about order.

18:56 I agree.

18:57 He was way ahead of his time.

18:58 This is also in the shadow of World War right?

19:01 And I think that actually I mean we should

19:04 definitely do some episodes at some point about the kind

19:07 of the darkness of of eugenics and and and and how

19:12 people were seeing genetic differences to distinguish between us.

19:16 But in this aftermath of all of the horror

19:19 that had happened in the Second World War,

19:20 scientists were actually sort of very keen

19:23 to notice the universality of humans, right?

19:25 That that actually we're all the same.

19:27 This is sort of a really big trend at the time.

19:29 Yeah.

19:30 Top-down stuff, right?

19:31 This like this one rule that binds us all, the the rule of your genes.

19:35 This rather than there being chaos and messiness

19:38 that can cut bubble up from from the bottom.

19:41 In Turing's life as well, I mean you mentioned that he died in 1954

19:46 and this paper he released in 1952 to a third.

19:49 But what also happened in 1952 was the sequence

19:53 of events that would lead to his lead to his death.

19:56 So, in January that year he he was kind of finalizing

20:00 this exact paper this exact like ink dots um paper.

20:06 He has this little relationship with uh

20:08 a 19-year-old working-class man called Arnold Murray.

20:12 And shortly after their relationship, Turing's house gets burgled.

20:16 And Turing reports the crime to the police.

20:19 And during the investigation,

20:20 he just casually mentions that the burglar was an acquaintance of Murray.

20:26 Uh of of the guy that he'd been seeing.

20:28 And he admits to the police that he

20:30 had been having a sexual relationship with Murray.

20:32 So, he's going to the police for help.

20:35 Right.

20:35 And the police take this information and turn it against him

20:39 because homosexuality was illegal in the UK at this point in time.

20:42 Literally illegal, yeah.

20:43 Literally illegal.

20:44 So, um it's the the law was gross indecency.

20:48 And Turing didn't deny it.

20:50 He doesn't apologize.

20:51 He he's like, "I haven't done anything

20:53 wrong." But, he gets convicted in March 1952.

20:57 His paper is published in August, by the way, the one that I'm describing.

21:00 Wow.

21:01 And the state give him a choice.

21:03 They say, "Okay, well,

21:04 you can go to prison or um you can just have a year on probation,

21:09 but on the condition that you undergo

21:12 hormonal treatment to reduce your libido." But, essentially,

21:15 it's a chemical castration.

21:17 Yeah.

21:18 is injected with synthetic estrogen.

21:21 And there's this real horrible irony to um to what

21:25 happens to him considering the work that he's doing.

21:28 You know, he's thinking about biology.

21:31 He's thinking about how you get chemicals that dictate the physical

21:36 shape and boundaries and the features of a living creature.

21:39 And he's spending his evenings watching his own physical shape,

21:43 his own body be, you know,

21:45 forcibly rewritten by the chemicals that are injected by the state because

21:50 this this synthetic estrogen that he's given it it fundamentally changes him.

21:54 You know, he develops breasts.

21:56 His weight changes.

21:58 He gets loads of brain fog.

21:59 He suffers really severe depression.

22:02 By 1953, so a year later, his his you know,

22:06 it's homosexuals are considered a security

22:07 risk that are susceptible to Soviet blackmail.

22:10 So, his security clearance is revoked.

22:12 So, the thing that, you know,

22:14 he that the the world in which he is lauded as a hero has rejected him.

22:20 The the scientific community aren't interested in any of his new ideas.

22:25 He's, you know, isolated.

22:26 He's like surveilled.

22:28 He's like physically altered.

22:30 And um you know, less than 2

22:32 years after publishing this paper that I'm describing,

22:34 while he was working on a second paper on on on these biological ideas,

22:39 he um took his own life uh at the age

22:42 of 41 by eating an apple that was laced with cyanide.

22:46 It was a such a tragedy.

22:48 This like gigantic gigantic titan of computing and mathematical

22:54 history that that the state just treated so unbelievably badly.

22:57 And we lost so much because of that.

23:01 Mhm.

23:01 So much of his life and what he could

23:04 have done and what he was actively working on.

23:06 Mhm.

23:07 Completely.

23:08 Imagine what he could have done in another 40 years.

23:12 Yeah.

23:12 I mean, not least on on the computing stuff, right?

23:15 The computing stuff that he was right there

23:18 at the beginning at that people still refer back to.

23:21 Not in some passé way, but that he formed the absolute foundation, you know,

23:25 universal Turing machines are still the absolute pinnacle of of what

23:29 we are looking for and how we consider intelligence to be.

23:32 That's right.

23:32 It wasn't just the foundation,

23:33 but it was it was also like the beams that were still hanging on.

23:38 To Turing tests.

23:39 And I I mean, I knew all of that.

23:41 I didn't know he'd done work that was so relevant to biology, though.

23:44 Right.

23:45 And this is the thing because while

23:46 the the scientific community just dismissed it as absolute junk,

23:50 it turns out he wasn't just like onto something,

23:55 he was absolutely phenomenally precise.

24:01 He, in his own mind, managed to absolutely nail the precise mechanism that is

24:09 going on behind the scenes in biological systems.

24:13 So, uh it wasn't until 1995.

24:16 This is like 40 years after his death, okay?

24:19 That a biologist was looking at the stripes on an angel

24:22 fish and noticed that they they they matched Turing's equations.

24:28 Okay, but this is like a little bit of a hint.

24:30 But, only really recently, okay, in 2006,

24:34 have people found for real the ink and the eraser.

24:38 Okay, we now know that this is like genuinely genuinely legit.

24:42 So, it's called WNT and DKK.

24:45 And one of the most famous confirmations of of Turing's theory,

24:49 it's about how mammals grow hair.

24:51 Okay, so if you look really closely at skin,

24:54 hair follicles, they're not they're not randomly placed.

24:56 They're they're spaced out in this in this dotted pattern, okay?

24:59 Like imagine looking down at the forest.

25:01 You've got these little patches of fire, okay?

25:03 Imagine looking at like, you know, this landscape of foxes and rabbits.

25:07 You've got these little pockets of rabbits,

25:09 you know, and and foxes roaming in between them.

25:11 The activator is is this protein called WNT and it it tells skin cells, "Okay,

25:18 we'll start building a hair follicle here." And WNT

25:22 is this really heavy it's this really sticky protein.

25:24 It doesn't travel very far.

25:26 It stays really local and it creates this build-up of itself, right?

25:30 It's got to the more you get the more you get.

25:33 The inhibitor is this protein called DKK.

25:36 And WNT, by the way, triggers the production of DKK.

25:40 So, it's like it's making its own enemy.

25:42 And [snorts] DKK tells the skin, "Do not grow hair." Right?

25:46 It sort of says like no more hair follicles.

25:48 And DKK, exactly as Turing said, is this much smaller,

25:52 much more mobile protein that diffuses out really quickly,

25:56 rushes out into the surrounding tissue much faster than WNT can spread.

26:00 And so, thus, you end up with hair follicles

26:04 being in these little islands dotted around the landscape.

26:08 I just have to tell you, right?

26:09 To to prove that this wasn't just a coincidence,

26:13 biologists they decided to uh like tweak things.

26:16 They got some mice and they decided to like tweak

26:18 these two proteins just to see if it would make a difference.

26:21 And Turing's equations predicted that if you weaken DKK,

26:25 the inhibitor, then the hair follicles should kind of spread further.

26:29 The the um the the one that creates

26:31 the hair follicle should spread further before being stopped.

26:33 And you should get these hair spots that are much uh bigger,

26:37 much more close together, much much more kind of merged.

26:41 And so, they weakened the uh the inhibitor.

26:44 And as a result, these mice they grew these kind of huge

26:48 merged clusters of hair follicles exactly where the math said that they would.

26:53 Wow.

26:53 And then when they did the flip side,

26:55 when they engineered mice with like stronger inhibitor,

26:58 then these were like basically bald mice, right?

27:00 Their hair follicles were shrunk.

27:02 They were spread much further apart.

27:04 Like he nailed it.

27:06 We now know for a fact, okay?

27:08 And this is like studies that are going on in 2021,

27:11 that leopard spots is exactly this mechanism.

27:16 You've got the the the WNT DKK 4 in in the case of uh of um of leopards.

27:21 It happens during fetal development.

27:24 Uh you get a darker hair density and then sort of a light hair density.

27:29 It's happening as the the leopard is sort of growing in the womb.

27:33 We know that on the roof of your mouth,

27:35 the the shape of the ridges in the roof of your mouth is a Turing pattern.

27:40 Really?

27:41 I'm feeling them right now with my tongue.

27:43 You're feeling them right now.

27:44 That is Turing Turing a Turing pattern right there.

27:47 We know fingerprints, by the way,

27:49 the ridges on your fingers, um it's exactly the same thing.

27:53 It's about week 10 of pregnancy.

27:55 You get these chemical waves that that follow the Turing patterns.

27:58 Um the feather patterns of fingerprints the whole time, by the way.

28:01 Were you?

28:02 As soon as you mentioned that the geneticists were like, "No,

28:05 it's all according to the rules of DNA." I was like, "Ah, but you know what?

28:09 Not fingerprints.

28:10 Identical twins share DNA,

28:14 but they do not have the same fingerprints because those are formed

28:17 in this way that Turing discovered that's much more of an order out of chaos.

28:21 Exactly.

28:22 Exactly.

28:23 So, if you have a twin, they can bleed to cover up your crimes,

28:27 but their fingerprints will give it away.

28:29 You cannot blame them unless [clears throat] you've got their blood.

28:32 You know, even fingers and toes, right?

28:33 If you see if you think of your your hand as like it's

28:36 almost like a wave of like yes no yes no yes no yes no.

28:39 It's Turing pattern.

28:42 [laughter] No kidding.

28:42 Yes.

28:42 It's like a this wave of proteins

28:44 that oscillate across your embryonic hand that telling Yes.

28:49 They're telling you when to build bone, which is your activator,

28:52 and when to sort of die off and create the gaps between your fingers,

28:54 which is the inhibitor.

28:55 So, it's like activator finger, inhibitor gap.

28:58 It's like it's all Turing.

28:59 It's all Turing.

29:01 This he came up with this in his mind.

29:03 You know, he didn't even do biology.

29:06 What's sticking with me is that they were able to do this on mice.

29:12 It's like, "Okay, we could do math on paper

29:15 or on a computer and then the screen will output the image,

29:17 but instead of a screen,

29:19 they used a mouse's body." They do all the little like math.

29:22 They create the things and then the mouse is printed

29:25 out and it matches the equations that they have seated.

29:30 Exactly.

29:31 Exactly.

29:32 Which is it's it's wild to imagine.

29:36 And it's sort of like it's sort of which way around is it?

29:38 Is it that your body is doing these equations or is it

29:41 that the equations are just unreasonably good

29:44 at describing what your body is doing?

29:45 It's probably the latter.

29:47 But but I mean there are so few of these.

29:49 Like by the time the 1950s comes along, you know,

29:52 physics is sort of like Einstein's been along and done all of the space stuff.

29:56 You know, chemistry's got all the periodic table.

29:58 There are so few of these unbelievably beautiful

30:02 elegant descriptions of reality that are left to find.

30:05 And Turing had one.

30:07 But if he had only done this, he would have been one

30:10 of the most important scientists of the last of the last century.

30:14 Well, yeah, because after the the periodic

30:17 table is filled out and we've got relativity,

30:21 then yeah, we really moved into order,

30:24 chaos, complexity, and that's right where Turing was.

30:26 Exactly.

30:28 So this this idea then of of of how does the egg know what to do?

30:33 How does the embryo know what to do?

30:34 This is what it comes down to essentially

30:36 is that if you have a normal physical system,

30:39 then you get a tiny fluctuation, like a little ripple,

30:42 then it just diffuses, it fades away, the water goes flat again.

30:46 But when you have a Turing system, a tiny instability,

30:51 just a a random bump of the activator chemical,

30:55 ends up hitting this this positive feedback loop,

30:58 and because the activator's job is to make more of itself, you know,

31:02 you end up starting this this process

31:05 that that that ends up building these biological creatures.

31:08 And when the sperm pierces the egg, right,

31:13 it physically snaps the surface of the cell,

31:16 and that is enough of a fluctuation in some creatures

31:21 to end up dictating the main axis of the creature,

31:25 that the axis of head to butt, essentially.

31:27 Wow.

31:28 So so there is a worm.

31:31 We know this Humans are a little bit more complicated, right?

31:33 There's a bit more going on.

31:34 But there is a worm where just that fluctuation of where

31:37 the sperm enters is enough to say here's the head, here's the here's the butt.

31:42 I should tell you the sperm enters, and that's where the butt is.

31:45 Um case you're interested.

31:47 That's that's going to be the butt.

31:49 Um in the frog it's the belly where the sperm enters, it's the belly.

31:53 It's a shame it I looked this up.

31:55 I was looking this up yesterday.

31:56 It is a shame cuz I really wanted it to be that you could say in a human

32:01 that you know that the sperm from your dad

32:03 entered and it turned out to be your ear.

32:05 I really wanted that to be the case, but it isn't.

32:07 Not quite that.

32:10 [clears throat and cough]

32:10 No, cuz human eggs you can split them and then they still it's

32:13 not like this half becomes the butt and this becomes the head, you know?

32:17 Like Oh, I guess that's true.

32:19 Yeah.

32:19 have the potential.

32:21 You couldn't take a human egg, rip it in half,

32:25 reverse it, and make a butt-head person.

32:28 [laughter] I think that's the plot of human caterpillar, isn't it?

32:32 [laughter] You well Oh, no, centipede, damn it.

32:34 I've got it wrong.

32:35 Yeah.

32:35 And also it's not the plot of human centipede,

32:37 but it could be the plot of human centipede 4.

32:40 I have to confess I've never watched a single one of them, not even a trailer.

32:43 I watched the first one.

32:45 Did you?

32:45 Well, yeah, with the premise that it has, how could you not, you know?

32:50 I didn't need that mental image in my life, frankly.

32:52 Hey, there hasn't been a four.

32:54 I nailed it.

32:55 You got two and three.

32:58 Quick.

32:59 Just kidding.

32:59 I'm a huge human centipede fan and I've watched them all.

33:03 We own We own that now.

33:04 If anyone If anyone does make that plot It's got to be about

33:08 in in utero like cell changing in order to create a like a donut human.

33:15 Yeah.

33:15 centipede with one person.

33:17 Where you eat your own butt.

33:19 Like perpetual motion,

33:20 but But it sounds like you couldn't create these you couldn't

33:26 you couldn't create the human centipede four people using just genetic changes.

33:31 You would need to also alter the properties of these chemicals that create

33:36 the Turing patterns and interact with each

33:38 other in these ways that produce order.

33:40 Yeah, I mean I think that the in humans

33:42 are just so much more complicated, right?

33:44 So there's a lot going on when it comes

33:46 to the sort of complexity of a of of a human.

33:48 Not for the scientists in our movie.

33:51 They're going to figure it out.

33:52 Could you have instead of human centipede,

33:54 could you have would worm centipede be interesting or not really?

33:58 What What?

33:59 Oh, how about how about this?

34:00 How about centipede centipede?

34:03 I I would rather watch that, frankly.

34:04 [laughter] I would rather watch that.

34:07 What is what is the snake that eats itself, the ouroboros?

34:11 That's what I'm imagining.

34:12 If you do happen to be in charge of mega Hollywood budgets,

34:17 give us a call, you know?

34:18 We've got We've got many more ideas where this came from.

34:20 If you go back to that analogy of the forest, right,

34:23 and just seeing that you get these patches of of of fires,

34:28 it's a little like hot spot, as it were, of fires.

34:31 For a really long time, you know, probably since the 1970s,

34:34 people have looked at the the mathematics of Turing patterns and been like,

34:38 "I wonder if you see that in human systems, too.

34:40 I wonder if you get that in in urban

34:42 settings." And probably I think the clearest

34:44 example of this there was a there was

34:46 a paper that actually wasn't that long ago,

34:47 2019, and there was uh an urban planner called Peter Pelz,

34:53 and he was looking at slums in the global south, okay?

34:57 And his idea was, "Look, I don't think these things end up forming randomly.

35:02 They actually have this really distinctive spatial pattern." Hmm.

35:06 And so he applied Turing's equations and then found

35:09 that you get the the same kind of dynamics.

35:12 Because if you think about it, right,

35:13 you what why do you end up getting

35:16 slums in the global south where where they do?

35:19 And part of it is because of local attraction.

35:21 So you get a little bit of a demand for um low-income workers,

35:27 so they settle maybe near an industrial area or like a transit hub,

35:31 and then once they're there they create a network,

35:33 and then it means that, you know, they open food stalls, they you know,

35:36 offer informal labor, whatever it might be.

35:37 They they're sort of pulling other people in towards them.

35:39 It's local activation, like a an attraction for other people.

35:44 But as you get more people who crowd into that area,

35:47 then the land becomes really scarce,

35:49 and then the living conditions, you know, maybe get a bit worse,

35:52 because these residents don't necessarily have like mobility.

35:55 They can't, you know, pick up and move and go somewhere else.

35:58 You end up with sort of a moat,

36:00 effectively, that appears around around these these areas.

36:05 And meanwhile the the wealthy population who can commute,

36:07 they're kind of pushing back.

36:08 They're driving up land prices in surrounding rings.

36:11 So you end up with like a city that organizes

36:15 itself into a Turing pattern where you get these intense,

36:17 highly localized clusters of poverty, the sort of spots, as it were,

36:22 that are surrounded by these much wider rings of, you know,

36:25 high income or commercial areas outside of it.

36:28 That's so fascinating.

36:30 So the spots on a leopard, the rosettes on a leopard,

36:34 for those pedants out there,

36:36 they're arranged in the same way that poverty spots the earth.

36:41 So that's the theory, right?

36:43 But it's definitely more descriptive.

36:46 I mean there's no chemistry going on here.

36:48 There's no sort of like there's no But instead of instead of atoms, it's people.

36:53 Where does it possibly fall apart though, this this similarity?

36:57 Wow.

37:00 It's quite the um question that you've just asked.

37:05 Because some people think that it doesn't, or certainly thought that it didn't.

37:10 I think if you're looking for kind

37:11 of localized hot spots in an urban environment,

37:14 then uh especially some that have a dynamic equilibrium,

37:18 that sort of like stay there over time, then actually crime looks like a really

37:23 sensible place to look for Turing patterns,

37:25 because you know, let's take burglary as an example.

37:29 You've got burglars who are who are wandering around the city,

37:31 who are looking for opportunities, which is like diffusion in a way.

37:35 It's like it's like the foxes that you had before.

37:37 You've also got police who are patrolling trying to to prevent crime.

37:42 So you've got that reaction between the two groups.

37:45 The difference slightly to other systems is that you know

37:48 that burglars also communicating with each other about high-value targets.

37:53 Um we've known Criminologists have known

37:55 this for a really long time, by the way, that point at which you are most likely

37:58 to be burgled is when you've just been burgled, okay?

38:03 Um because people replace their valuables,

38:05 because burglars know the layout of your house,

38:07 maybe there's something there that they want to come back for.

38:10 But also often in a street you get

38:13 um houses that are similar structure to one another.

38:16 So burglars will repeatedly target the same

38:18 area until people people kind of notice.

38:20 So you get this like spike,

38:22 this like this moment of attractiveness of a particular

38:25 area where burglars are all drawn to that area, the hot spot, essentially.

38:30 [snorts] So in 2008 there was this group

38:31 of mathematicians who released this absolutely gorgeous paper,

38:36 and I've used this paper with my students loads,

38:38 cuz it's just like the mathematics in it is really lovely.

38:41 It's it's it's based loosely around the Turing idea of reaction and diffusion,

38:47 and it's on this idealized street network.

38:49 It's like it's a grid structure.

38:51 It's kind of this perfect mathematical model.

38:55 And uh they show that if you have this system,

38:59 this kind of reaction and diffusion between burglars and police,

39:02 you end up seeing hot spots pop up and disappear

39:05 in the same way that you do across a real city.

39:07 Right, it's like you look at it and you're like, "Oh,

39:09 well that looks a lot like what you see in the city." Kind

39:11 of in the same way that Turing looked at zebra patterns and was like,

39:14 "Well, that's sort of what you you end up seeing in real life." Yeah.

39:17 The thing is in 2008,

39:19 this was an era when people were very excited about data becoming available.

39:25 You know, we hadn't really had really,

39:26 really good data on people before this point.

39:29 And now you had mobile phones, you had like city reports, you had cameras, etc.

39:35 And so um people were wondering, well,

39:37 maybe maybe there is a way that you can do physics with people.

39:41 Maybe there is a way that there are

39:42 underlying equations that that dictate how people move in.

39:46 Sure, yeah.

39:47 Model them as a bottled gas.

39:49 Let's see what happens.

39:51 Exactly.

39:52 And the police agreed that maybe this was something that could work.

39:56 So, the Los Angeles Police Department,

39:58 they looked at this paper and they were like,

40:01 "Ooh, there might be something in this.

40:02 Maybe maybe maths can tell us where

40:07 these hotspots are going to be next, you know?

40:09 Not just where they are." Now we're predicting crime.

40:12 Now we're predicting crime.

40:14 Future crime.

40:15 Future crime.

40:17 Or at the very least, how are these hotspots moving and changing

40:21 and and how might they move in future?

40:24 So, one of the people on that original paper,

40:26 um this guy called Jeffrey Brantingham,

40:28 he took this this like what was a really beautiful theoretical

40:32 paper and turned it into a company which was called PredPol,

40:37 predictive policing.

40:38 And then in the early 2010s, what they would do is they were working with police

40:41 officers in LA and they would print out these maps

40:44 at the beginning of their shifts and the maps had

40:47 like a little 500 by 500 foot boxes on them.

40:51 Mhm.

40:51 And that box would say, "This is where there's likely to be crime this evening.

40:56 This is where we expect crime to be." Right.

40:59 The maths by this point had like I mean,

41:01 I just want to make sure that I'm really accurate.

41:02 This is like it's moved on a couple of steps from Turing's

41:05 reaction-diffusion in order to get it to work for that setting.

41:08 But that like right at the heart of it and kind of the the nugget of the idea

41:11 is the same thing that you have

41:13 these two systems that are diffusing across a space.

41:16 And I have to say, right,

41:18 they did a randomized control trial on this in Kent in the UK.

41:24 And it's the closest that anyone had been to like double-checking if it worked.

41:29 And to be honest, it actually did work.

41:31 It did work in this Kent study.

41:33 But it didn't work in Los Angeles.

41:36 Well, they didn't do a proper test of it in Los Angeles.

41:38 Or at the very least, they didn't publish a proper test of it in Los Angeles.

41:42 Okay.

41:42 But in Kent they did.

41:43 They published a proper a proper trial of it.

41:46 So, they had and it was double-blind as well.

41:48 So, they had two teams, two sets of patrol maps.

41:51 One of them was like professional crime analysts who were doing it,

41:54 humans using their intuition,

41:56 and the other was the was this this program, this software.

42:00 And what they did, the algorithm won, basically.

42:03 The the algorithm was 10 times more accurate

42:06 at predicting the exact 500 square foot box.

42:11 Where a crime would occur.

42:13 And Kent, they they reported an 8.5%

42:17 drop in street crime during this trial period.

42:19 I'm assuming what you do is you look at this grid that the software

42:24 has filled in with potential hotspots

42:27 tonight and you put officers there Exactly.

42:30 or cameras, you know, here in the states we've got these big like portable poles

42:35 with blinking blue lights covered in cameras that just tell you,

42:38 "We're watching here." I mean,

42:39 it could get really complicated because if you dissuade

42:42 the criminals from where they would have been operating that night,

42:45 they're just going to go somewhere else that same night.

42:47 You have very smartly landed on the real

42:52 vulnerability of this because when you're

42:56 observing leopard spots or retrospectively looking at where

43:01 slums have appeared in the global south,

43:04 you are not in there interfering with the system as it happens.

43:10 The real problem with PredPol was [snorts] that if you

43:15 are sending cars into a particular neighborhood and you're saying,

43:20 "This is where crime is going to occur." You've got people in those cars, right?

43:24 You've got like real police, real humans.

43:26 And if they are expecting to find crime in a particular area of the city,

43:33 they're going to find crime.

43:34 They're going to find it, yeah.

43:36 And maybe it's not the crime that the algorithm was talking about.

43:40 Maybe it's not burglary, but maybe it's I don't know,

43:43 like even someone jaywalking or whatever it might be.

43:46 But then the issue also is that the algorithm

43:50 relies on knowing where the crime was already.

43:54 Right.

43:54 And so, the more crime that you find,

43:57 the more crime you're putting into the system,

43:59 the more that the algorithm thinks that that place is already a hotspot,

44:04 the more it's going to send police back into that area,

44:07 and the more and more and more you end up finding more crimes.

44:13 [snorts] And it will not surprise you when I tell you that in LA in particular,

44:17 the neighborhoods which were disproportionately flagged as hotspots

44:20 by this algorithm were neighborhoods

44:23 that disproportionately contained African-American residents.

44:28 And so, what ended up happening essentially is that this opened

44:33 the door to what was automated harassment of particular communities.

44:38 Right.

44:39 There's a big difference between watching how spots

44:42 form or hair follicles on a mouse Mhm.

44:45 versus actually putting law enforcement officers

44:48 in a particular part of the city where

44:52 they're looking for and interacting with the very

44:55 chemical reaction you're trying to predict.

44:58 They're going to they're going to mess

44:59 up the results by being there and noticing, "Oh, I saw some jaywalking.

45:04 I saw a car without plates." Things that normally would not have been noticed,

45:08 wouldn't have been reported, are now suddenly getting fed into the algorithm.

45:12 And that's changing up not only what they think will happen,

45:15 but how they're treating everyone who lives in this whole city.

45:19 Right.

45:19 And this and and thus you come to kind of a conundrum because actually,

45:24 I have to confess that I sort of have like a bit

45:26 of a a bit of a connection to this type of work, right?

45:29 So, I have published papers on mathematics of burglary.

45:32 I have like these mathematicians who published that initial paper,

45:35 you know, I said that I've like used it for my students.

45:36 I've like met and worked with them.

45:38 And in the early days, I especially after the Kent randomized control trial,

45:43 it was like, you know, I I really didn't immediately see what the problem was.

45:48 I didn't immediately see the potential ethical concerns

45:53 of this stuff because here is the conundrum,

45:55 you can, to a certain extent, better than random chance,

45:59 predict where crime is going to happen.

46:01 You can, right?

46:02 Right.

46:02 Like the algorithms work.

46:03 The problem is, what on earth do you do with that information?

46:07 Yes.

46:08 What do you do with it?

46:10 Because a crime that might happen isn't a crime yet.

46:14 Mhm.

46:14 If you through surveillance and and presence cause it to not happen,

46:21 then what does the algorithm do for the next night?

46:24 You're dealing with probabilities here, you know?

46:26 Yeah.

46:27 I mean, in in the same in biology,

46:29 the same in forest fires, the same in all of this.

46:32 You're not saying 100% definitely this area is

46:36 going to be part of the hotspot or not.

46:38 You are handling with uncertainty front and center.

46:42 And so, it's one thing to say, "This area, this individual,

46:47 this group of people are going to be implicated

46:49 in a crime in future." It's another thing altogether to say,

46:53 "Maybe they will, maybe they won't be.

46:55 Should we or should we not intervene?" Right.

46:58 I should tell you now the way that the UK have dealt

47:01 with this, uh I imagine it sounds like it's something similar in the US,

47:05 is that if there has been a burglary

47:07 in your neighborhood and your chances of being burgled have increased,

47:11 they will put a leaflet through your door saying,

47:14 "There is a uh an increased likelihood that uh that your house may be targeted.

47:19 Make sure that you keep your security up." Yeah.

47:23 That again has had a randomized control trial and that again

47:25 has ended up with like a with a really positive impact.

47:28 But it's interesting anyway.

47:29 I just it it This is one of those areas where I feel like I was really there

47:34 as it was all going on with the academics

47:36 as people were trying to apply mathematics to areas of policing.

47:42 And really there when the backlash

47:45 and the naughty repercussions of it came through.

47:49 I mean, I should also say one of the things about PredPol,

47:51 the reason why it was dropped in particular by the LAPD

47:54 was that actually internal reports said it just didn't work.

47:57 It just didn't work very well.

47:59 Um Well, part of it could be the name, too.

48:01 PredPol.

48:02 It sounds like a dark organization in a movie.

48:07 Mhm.

48:08 You know, it's a little bit too powerful sounding.

48:12 It does a little bit.

48:13 It's a little bit Minority Report, isn't it?

48:15 Yes.

48:16 Yeah.

48:17 Not that the name was the biggest problem,

48:19 but hindsight being 20/20, you can go, "You know what?

48:22 If I was a screenplay writer,

48:24 I would call the organization that doesn't quite work right PredPol." [laughter]

48:30 I mean, absolutely.

48:31 Has anyone thought of calling it BreadBowl?

48:34 Because that [laughter] sounds delicious.

48:36 See, this is this is the kind of commentary I'm here to add, Hannah.

48:40 [snorts] [laughter] And I appreciate every moment of it.

48:43 better BreadBowl joke there than what I came up with, but you know,

48:46 you guys in the comments can give it a better setup.

48:50 This is fascinating.

48:51 It's so it's so amazing that you worked

48:53 on this mathematics and worked with the mathematicians.

48:56 So, what's the state of this field now?

48:59 I have to be honest with you.

49:00 I mean, I'm There's like There's like seven

49:02 different directions I could go in with this.

49:04 And I was thinking about this last night and I still haven't made up my mind,

49:06 which I probably should have done in advance of this conversation.

49:09 There are also lots of examples of this, right?

49:11 The one piece of work that I particularly did, actually,

49:16 um that really got me out of this whole space was

49:21 I [snorts] was working with the police in This is 2011.

49:26 And there had been these big riots across the UK.

49:28 Really really out of control, right?

49:31 Things There was There was looting,

49:33 there was arson, there was all kinds of assaults.

49:35 It was It was really the were very

49:37 shocked by how quickly things had had descended,

49:40 how out of hand things had gone over the course of 5 days.

49:45 This is when I was, you know, just finished my PhD.

49:47 This is like the first thing that I

49:48 was working on, this collaboration with the police.

49:51 And what they did was they collected all the data

49:55 of everybody who'd been arrested in connection with these riots,

49:58 and they handed it over to this group

50:00 of mathematicians and criminologists and said,

50:02 "Okay, see what patterns you can find,

50:04 what we did and didn't do and what we could have done differently

50:07 so that things didn't get quite so out of hand." In the UK,

50:11 our police are not perfect by any stretch of the imagination,

50:13 but I get the impression that we have a slightly better relationship

50:16 with them than perhaps some people do in in the States, right?

50:19 They're sort of more You get There's a more of a a kind of feeling of community.

50:23 Well, yours are called bobbies.

50:25 They're called bobbies, yeah.

50:26 They're not without flaws.

50:27 It's important for me to say that, but but on the whole they tend to be,

50:30 you know, they do a lot of really good work.

50:31 So, we did that.

50:32 We published this paper, and as part of the paper,

50:34 as well as all of the analysis that we did,

50:36 we kind of constructed this um this algorithm that I I mean,

50:42 it was very very crude, right?

50:43 Very very proof of concept,

50:45 but the idea was that you could you could look at it if something like

50:49 this happened again in future so that police

50:51 could bring about a swifter resolution to unrest.

50:54 Anyway, a few years later, we published this paper.

50:56 The academic community were like, "Great, you know,

50:58 this is like 2013, something like this." Couple of years later,

51:00 I went off to go and give a talk about this in Berlin,

51:04 and I was standing on stage at like this this audience,

51:08 which involved lots of the public,

51:10 and I was giving this really enthusiastic presentation about how great

51:14 it was that we now had all of these tools, right?

51:17 Now we were in a situation uh I think I was very naive at the time,

51:21 Michael, to be honest with you, but we were now in a situation where, you know,

51:25 we could support the police control

51:28 a city's worth of people, essentially, right?

51:30 I mean, I was very young and very naive,

51:32 and I think it just genuinely didn't occur to me that the if there's one city

51:36 in the world where people are a bit scared

51:38 about the police having too much power and control,

51:40 it's probably going to be Berlin, [laughter] Yeah.

51:44 Yeah, so what was their reaction?

51:46 I mean, they [snorts] were not happy.

51:48 They were like uh they they tore me apart in the Q&A, which I mean, actually,

51:52 it was a really important moment for me,

51:54 like a really really significant moment for me in my career.

51:57 Because the thing is, up until that point, when you are a mathematician, right?

52:01 When you are a physicist or whatever,

52:03 and you're coming up with these like mathematical ideas,

52:06 you don't have to worry about the sort of Turing wasn't

52:09 worried about the ethics of changing molecules in mice's skin, you know?

52:14 He wasn't sort of like thinking about the the moral

52:17 implications of like testing the fingerprints of twins, you know?

52:21 He was just like having fun with his equations, right?

52:25 Yeah.

52:25 And I think that that was the moment when I really realized

52:29 that a lot of the people who are designing our collective future,

52:32 a lot of the people who are working in artificial intelligence,

52:35 who are working with algorithms, who are working with data,

52:38 they've gone through with a very technical training that hasn't said to them,

52:43 "You need to be careful in what you're doing.

52:47 You need to think about the wider implications of it.

52:50 You need to not see this stuff as though it's

52:52 just like a cute little mathematical model that kind of sits

52:56 on a shelf and you can stand back and like admire

52:58 it as though it exists in isolation of the world around it.

53:01 You have to think very deeply about the way your stuff can be

53:05 used and the impact that it will have on the world." And so,

53:09 that really was the moment when I switched course.

53:11 I started writing about the ethics of of algorithms.

53:15 I've incredible amount of work ever since.

53:17 I think it really accelerated my my work in public communication of science

53:21 and talking about human issues um as well as just technical ones.

53:25 That's so cool to hear cuz it's so it's so true, isn't it?

53:29 I was thinking as we talked about Turing that he wound up

53:35 being punished by being injected with all

53:37 these chemicals because he was homosexual.

53:39 Mhm.

53:40 And yet, to prove his work right, we had to pump a bunch of rats full of changed

53:47 chemicals in order to to figure out how their hair would grow.

53:52 Not because they were homosexual, but because they were not humans.

53:55 Mhm.

53:56 That was their punishment for just being a mouse that we could test on them,

54:00 the way that in a way Turing was tested on.

54:02 But of course, what's the alternative?

54:05 That we don't test any of this at all?

54:07 Because from this research, so much good can come, too.

54:11 So, so much hinges upon the responsibility of those who

54:16 pay attention to the ethics of what they're doing.

54:19 We need the knowledge, and we shouldn't stop gaining the knowledge,

54:24 but there's a different thing called wisdom that we need even more.

54:28 And that's how you use the knowledge.

54:30 Yeah.

54:31 Because I'm also sitting here thinking, well, gosh,

54:34 now that computing power is just more and more democratized every day,

54:39 I should start doing this PredPol stuff, but in reverse.

54:42 Where should I be committing my crimes?

54:45 Where should I hide a body based on the expected

54:49 intuitions and search patterns of the authorities, right?

54:54 I could turn the tables right on them.

54:56 It's a game of cat and mouse.

54:57 Is this Is this body going to be um a head stitched stitched to a butt,

55:03 Michael, cuz I think that's going to be

55:05 a giveaway telltale that it was you, all right?

55:07 When they do finally find it.

55:09 Hannah, I'm not going to snitch on myself, but the point is [laughter]

55:13 that math and science can help us do good,

55:15 and they can also help us do anti-good.

55:17 Yeah, absolutely, it can.

55:18 It's It's still such a such a a living conversation.

55:22 Absolutely.

55:23 Absolutely.

55:23 And I don't think it's going to be one with a finish line.

55:25 I don't think this is a finish line we get to cross.

55:28 No.

55:28 No, it isn't.

55:29 It's a moving finish line.

55:30 Mhm.

55:31 Let's see, how far can we push this analogy?

55:33 Like, the muscles we're running with our knowledge,

55:36 but the the path we take is wisdom, and the finish line doesn't exist.

55:43 It's It Is it a loop?

55:44 Is it a spiral?

55:46 Wow.

55:46 See, we'll do We'll do an episode on stitched into a torus?

55:50 Is it a human stitched into a torus creating its own individual human centipede?

55:56 Why do we always come back to human centipede?

55:59 It's the ultimate circle, Michael.

56:01 It's the ultimate circle.

56:02 It's the circle of life.

56:03 The circle of [laughter] one life.

56:06 I think we should leave it there, do you?

56:09 I think we should leave it there.

56:10 Yes, thank you all for listening, Hannah, and thank you.

56:13 I I loved hearing all of this.

56:15 If you're out there and you've got some

56:17 questions you want us to answer from yourself,

56:19 we do that every Thursday on Field Notes,

56:21 and you can send your questions to therestiscience@golhanger.com.

56:25 Thank you so much for watching listening to us.

56:27 Um if you are following us on YouTube, please do write comments below.

56:33 We read all of them.

56:35 Maybe not all of them.

56:35 Sometimes it's quite a lot, but a lot of them,

56:37 especially the ones that say first.

56:41 [laughter] And likewise on Spotify.

56:43 Um if you want to leave us any comments or send us any emails,

56:45 we'd love to hear from you.

56:46 Thank you very much.

56:48 See you next time.

56:50 time.

56:53 [music]