WEBVTT - Ep174 "Can Evolution Teach AI?" with Athena Aktipis

0:00:00.070 --> 0:00:03.430
<v Speaker 1>How do you get trillions of cells to collaborate to

0:00:03.530 --> 0:00:07.630
<v Speaker 1>build a single organism that we know and love as you?

0:00:07.640 --> 0:00:13.980
<v Speaker 1>Does this kind of cooperation create opportunities for hijacking? For example,

0:00:14.340 --> 0:00:18.959
<v Speaker 1>is cancer a form of cellular cheating? In what ways

0:00:19.040 --> 0:00:24.620
<v Speaker 1>does competition help a cooperative system hold together? And can

0:00:24.660 --> 0:00:28.290
<v Speaker 1>we take the lessons learned from biology and apply them

0:00:28.370 --> 0:00:32.790
<v Speaker 1>to politics and societies? And what can the lessons from

0:00:32.850 --> 0:00:37.670
<v Speaker 1>evolution tell us about building and living with AI? Today,

0:00:37.710 --> 0:00:41.630
<v Speaker 1>we sit down with evolutionary biologist and cooperation theorist Athena

0:00:41.770 --> 0:00:47.309
<v Speaker 1>Aptippus to understand cooperation and competition together. from cells to

0:00:47.370 --> 0:00:54.880
<v Speaker 1>societies to our technological future. Welcome to Intercosmos. With me,

0:00:54.940 --> 0:00:58.340
<v Speaker 1>David Eagleman, I'm a neuroscientist at Stanford, and in these episodes,

0:00:58.420 --> 0:01:02.920
<v Speaker 1>we sail deeply into our three-pound universe to understand why

0:01:03.040 --> 0:01:23.450
<v Speaker 1>and how our lives look the way they do. So

0:01:23.530 --> 0:01:27.670
<v Speaker 1>researchers at Stanford University announced last month a new discovery

0:01:27.780 --> 0:01:31.560
<v Speaker 1>that the brain, which we think of as a single organ,

0:01:32.120 --> 0:01:37.690
<v Speaker 1>actually arises from two distinct populations of cells. One of

0:01:37.709 --> 0:01:41.910
<v Speaker 1>those populations gives rise to the forebrain and midbrain, while

0:01:41.950 --> 0:01:44.760
<v Speaker 1>the other gives rise to the hindbrain, which handles a

0:01:44.750 --> 0:01:47.780
<v Speaker 1>lot of the basic functions that keep us alive. But

0:01:47.820 --> 0:01:50.860
<v Speaker 1>what we think of as the brain doesn't come from

0:01:51.080 --> 0:01:56.700
<v Speaker 1>one common pool of brain cells. It's two developmental lineages

0:01:57.190 --> 0:02:01.030
<v Speaker 1>that remain distinct. And we now know, as of last month,

0:02:01.390 --> 0:02:04.550
<v Speaker 1>that this arrangement goes way back in evolution across the

0:02:04.650 --> 0:02:07.910
<v Speaker 1>animal kingdom, which means that The brain we carry around

0:02:07.970 --> 0:02:13.620
<v Speaker 1>today is a composite of two very ancient systems. So

0:02:13.660 --> 0:02:17.300
<v Speaker 1>that got me thinking a lot about the issue of cooperation,

0:02:17.760 --> 0:02:23.200
<v Speaker 1>because here you have two developmentally independent lineages cooking along,

0:02:23.240 --> 0:02:25.990
<v Speaker 1>and they figure out how to work together in such

0:02:26.030 --> 0:02:29.210
<v Speaker 1>a way that for the entire history of our species

0:02:29.270 --> 0:02:31.490
<v Speaker 1>until last month, we didn't even guess that we were

0:02:31.550 --> 0:02:35.290
<v Speaker 1>looking at at two things smooshed together. And of course,

0:02:35.370 --> 0:02:37.589
<v Speaker 1>that's just the start of it, because really, when we're

0:02:37.630 --> 0:02:40.030
<v Speaker 1>looking at the brain, you could look at it as

0:02:40.110 --> 0:02:43.639
<v Speaker 1>two populations, but you could keep subdividing further and end

0:02:43.700 --> 0:02:48.900
<v Speaker 1>up thinking about it as 86 billion individual neurons, not

0:02:48.940 --> 0:02:51.960
<v Speaker 1>to mention lots of other cell types like glial cells. Now,

0:02:52.100 --> 0:02:56.820
<v Speaker 1>the world is teeming with single-celled organisms that And if

0:02:56.840 --> 0:03:00.639
<v Speaker 1>you look at a neuron, it looks essentially just like

0:03:00.800 --> 0:03:05.390
<v Speaker 1>any other single-celled organism, except that neurons are packed in

0:03:05.430 --> 0:03:09.549
<v Speaker 1>there with other neurons as neighbors, tens of billions of neighbors.

0:03:10.169 --> 0:03:13.209
<v Speaker 1>And somehow they all stick to one another and figure

0:03:13.270 --> 0:03:16.800
<v Speaker 1>out how to get along with one another and balance

0:03:16.919 --> 0:03:21.220
<v Speaker 1>competition for something bigger that's going on. Now, when you

0:03:21.240 --> 0:03:24.300
<v Speaker 1>look at neurons, you see they secrete chemicals that other

0:03:24.360 --> 0:03:27.290
<v Speaker 1>neurons pick up on. And I suggested in my book

0:03:27.350 --> 0:03:31.919
<v Speaker 1>LiveWired that this looks exactly like the way that single-celled

0:03:32.040 --> 0:03:35.820
<v Speaker 1>organisms spit out chemicals at each other as a form

0:03:35.860 --> 0:03:39.680
<v Speaker 1>of defense. But somehow, with the brain, the whole system

0:03:39.780 --> 0:03:43.580
<v Speaker 1>evolves in such a way that this defensive chemical spitting

0:03:43.600 --> 0:03:49.560
<v Speaker 1>ends up carrying information. And we end up calling this neurotransmission,

0:03:49.660 --> 0:03:53.970
<v Speaker 1>the sending of signals. And everything we care about, like

0:03:54.330 --> 0:03:59.900
<v Speaker 1>memory and movement and emotion, arises out of this cooperative

0:04:00.030 --> 0:04:05.140
<v Speaker 1>agreement between lots of little tiny unicellular guys. And also

0:04:05.220 --> 0:04:08.160
<v Speaker 1>plenty of other stuff in this massive division of labor,

0:04:08.240 --> 0:04:10.860
<v Speaker 1>like some cells end up taking on the jobs of

0:04:11.180 --> 0:04:16.279
<v Speaker 1>worrying about energy or injury or infection. And this arises

0:04:16.400 --> 0:04:23.920
<v Speaker 1>from these countless specialized unicellular cells continually influencing one another. Now,

0:04:24.480 --> 0:04:27.839
<v Speaker 1>widen the lens even further because the brain is just

0:04:27.880 --> 0:04:31.700
<v Speaker 1>a three-pound collection of cells which itself lives inside a

0:04:31.839 --> 0:04:36.050
<v Speaker 1>body made of tens of trillions of cells. that descend

0:04:36.150 --> 0:04:40.349
<v Speaker 1>from the same fertilized egg. Some become neurons, others become

0:04:40.510 --> 0:04:44.590
<v Speaker 1>muscle or blood or skin or bone or liver or

0:04:44.790 --> 0:04:48.630
<v Speaker 1>immune cells. And somehow they each do their own job

0:04:49.070 --> 0:04:54.140
<v Speaker 1>while all using shared resources and responding to messages from

0:04:54.180 --> 0:04:57.760
<v Speaker 1>each other, including very distant cells they're never going to meet.

0:04:58.390 --> 0:05:02.599
<v Speaker 1>And all these different lineages, they limit themselves so they

0:05:02.620 --> 0:05:06.260
<v Speaker 1>don't tip the whole organism out of balance. And all

0:05:06.300 --> 0:05:09.539
<v Speaker 1>those cells that come from that egg, they only make

0:05:09.640 --> 0:05:13.050
<v Speaker 1>up half of you. The other half of the creature

0:05:13.070 --> 0:05:17.350
<v Speaker 1>that we call you is all single-celled microbes, not from

0:05:17.490 --> 0:05:21.710
<v Speaker 1>human DNA at all. This is really one of evolution's

0:05:21.830 --> 0:05:28.830
<v Speaker 1>most astonishing achievements. Multicellularity, meaning not just one cell like

0:05:28.890 --> 0:05:31.789
<v Speaker 1>a bacterium running around, but trillions and trillions of cells

0:05:32.170 --> 0:05:36.520
<v Speaker 1>sticking to each other and acting like a gargantuan society.

0:05:37.040 --> 0:05:41.479
<v Speaker 1>And a society so cohesive that this giant population of

0:05:41.620 --> 0:05:45.310
<v Speaker 1>trillions of individual cells... wakes up in the morning and

0:05:45.370 --> 0:05:48.030
<v Speaker 1>looks in the mirror and thinks of itself as a

0:05:48.270 --> 0:05:54.640
<v Speaker 1>single unified being with a name. Now, multicellularity is no

0:05:54.760 --> 0:05:58.600
<v Speaker 1>easy trick to pull off because joining cells together opens

0:05:58.660 --> 0:06:01.740
<v Speaker 1>a huge new set of problems. If you're going to

0:06:01.779 --> 0:06:05.860
<v Speaker 1>be multicellular and you want to last, you also need

0:06:05.920 --> 0:06:10.469
<v Speaker 1>really sophisticated ways to coordinate labor. and to manage which

0:06:10.529 --> 0:06:13.390
<v Speaker 1>cells get to reproduce and when. And you've got to

0:06:13.730 --> 0:06:18.430
<v Speaker 1>distribute resources and also figure out how to remove damaged cells.

0:06:18.470 --> 0:06:21.250
<v Speaker 1>And then you need to build an army, an immune system,

0:06:21.589 --> 0:06:25.630
<v Speaker 1>to defend this shared internal environment. So we can see

0:06:25.750 --> 0:06:28.910
<v Speaker 1>how serious the challenges of an arrangement are like this

0:06:28.970 --> 0:06:32.370
<v Speaker 1>when we look at something like Cancer, because here you

0:06:32.430 --> 0:06:35.110
<v Speaker 1>have a cell just like all the other cells, but

0:06:35.130 --> 0:06:39.880
<v Speaker 1>it begins using the body's resources for its own expansion.

0:06:40.420 --> 0:06:43.299
<v Speaker 1>That's an internal problem, but you also have external problems.

0:06:43.660 --> 0:06:47.400
<v Speaker 1>Multicellularity is tough because you've got things like viruses and

0:06:47.480 --> 0:06:51.100
<v Speaker 1>bacteria and parasites, which are always coming at the system

0:06:51.460 --> 0:06:54.120
<v Speaker 1>and trying to take advantage of it. So if a

0:06:54.260 --> 0:06:58.300
<v Speaker 1>multicellular creature wants a long life, That depends on a

0:06:58.380 --> 0:07:03.410
<v Speaker 1>lot of layers of detection and repair and defense and

0:07:03.490 --> 0:07:08.310
<v Speaker 1>communication working across billions or trillions of cells. So in

0:07:08.350 --> 0:07:11.850
<v Speaker 1>this light, the fact that you're alive and listening to

0:07:11.870 --> 0:07:16.400
<v Speaker 1>this podcast means your particular body has a history of

0:07:16.600 --> 0:07:21.840
<v Speaker 1>decades of successful cooperation among the trillions of small pieces

0:07:21.900 --> 0:07:24.590
<v Speaker 1>and parts that make you up. And it also carries

0:07:24.650 --> 0:07:27.790
<v Speaker 1>a history of conflict that has been managed well enough

0:07:28.170 --> 0:07:32.470
<v Speaker 1>for your whole system to persist. Okay, now, whenever we

0:07:32.510 --> 0:07:36.030
<v Speaker 1>start looking at systems like this, giant societies of cells,

0:07:36.510 --> 0:07:40.970
<v Speaker 1>you can't help but compare it with human societies. You

0:07:40.990 --> 0:07:43.690
<v Speaker 1>look at cells here and people here, and you see

0:07:43.730 --> 0:07:46.770
<v Speaker 1>a lot of the same patterns. Just like a body

0:07:46.930 --> 0:07:51.950
<v Speaker 1>of cells, a country of citizens also has specialized roles here.

0:07:52.270 --> 0:07:56.570
<v Speaker 1>and shared resources and communication networks and rules and enforcement

0:07:56.690 --> 0:08:02.530
<v Speaker 1>and competing interests. And just like with evolutionary attempts at multicellularity,

0:08:02.970 --> 0:08:05.950
<v Speaker 1>where some made it and some didn't, our history is

0:08:06.110 --> 0:08:10.590
<v Speaker 1>filled with political systems that promised cooperation on a grand scale,

0:08:11.000 --> 0:08:13.120
<v Speaker 1>and some of them failed. So we can look at

0:08:13.200 --> 0:08:16.760
<v Speaker 1>issues of cooperativity on the scale of cells or humans.

0:08:17.160 --> 0:08:20.540
<v Speaker 1>And now that we're entering the era of AI agents,

0:08:20.780 --> 0:08:26.470
<v Speaker 1>we're going to have increasingly giant collaborations of the computational kind.

0:08:26.970 --> 0:08:30.990
<v Speaker 1>And we can see these same patterns of cooperation and competition.

0:08:31.430 --> 0:08:34.370
<v Speaker 1>And because we're the ones building the agent swarms now,

0:08:34.390 --> 0:08:37.550
<v Speaker 1>we have to make sure we understand the lessons from

0:08:37.620 --> 0:08:40.949
<v Speaker 1>biology and politics. And this is why it makes sense

0:08:41.030 --> 0:08:45.660
<v Speaker 1>to look at cooperation and competition as physical principles that

0:08:45.679 --> 0:08:50.140
<v Speaker 1>we can model and study, just like we do with gravity, electromagnetism.

0:08:50.540 --> 0:08:53.690
<v Speaker 1>What makes systems work? How do they scale? What should

0:08:53.730 --> 0:08:57.170
<v Speaker 1>we watch out for? And to this end, our guest

0:08:57.210 --> 0:09:02.980
<v Speaker 1>today is Athena Aptippus, a cooperation theorist. She's an evolutionary biologist.

0:09:03.380 --> 0:09:06.599
<v Speaker 1>and an associate professor of psychology at Arizona State University.

0:09:06.980 --> 0:09:12.520
<v Speaker 1>She studies cooperation and conflict across scales, from resource sharing

0:09:12.570 --> 0:09:16.230
<v Speaker 1>among people to the policing of cells inside a body.

0:09:16.670 --> 0:09:19.990
<v Speaker 1>She uses game theory and computational models and field work

0:09:20.390 --> 0:09:25.330
<v Speaker 1>to ask how cooperative systems form, how cheaters exploit them,

0:09:25.730 --> 0:09:29.850
<v Speaker 1>and how cooperation changes as a network grows. Here's my

0:09:29.890 --> 0:09:40.500
<v Speaker 1>interview with Athena Actipas. So, Athena, across human relationships and

0:09:40.540 --> 0:09:44.460
<v Speaker 1>microbes and cancer, you think about cooperation and conflict. Tell

0:09:44.500 --> 0:09:45.020
<v Speaker 1>us about that.

0:09:45.040 --> 0:09:49.780
<v Speaker 2>Yeah, so I think about these fundamental principles of cooperation

0:09:49.820 --> 0:09:52.710
<v Speaker 2>and conflict sort of like most people think about the

0:09:52.750 --> 0:09:56.970
<v Speaker 2>laws of physics, right? They apply to everything. The laws

0:09:57.110 --> 0:09:59.970
<v Speaker 2>are fundamental. They're sort of part of how the universe works.

0:10:00.280 --> 0:10:03.620
<v Speaker 3>Is put together. And so they're going to apply to everything.

0:10:04.040 --> 0:10:06.620
<v Speaker 1>And so what does that look like? So what is cooperation?

0:10:07.100 --> 0:10:12.040
<v Speaker 2>Cooperation is working together towards a shared goal, often with

0:10:12.100 --> 0:10:17.160
<v Speaker 2>some risk to yourself. And we see this in cellular systems, right?

0:10:17.460 --> 0:10:21.020
<v Speaker 2>We saw in the transition from unicellular to multicellular life

0:10:21.580 --> 0:10:23.830
<v Speaker 2>that all these cells had to give up their own

0:10:23.870 --> 0:10:28.050
<v Speaker 2>reproduction to be able to exist socially in a body.

0:10:28.910 --> 0:10:31.850
<v Speaker 2>We see it in human societies with people giving up

0:10:31.870 --> 0:10:36.240
<v Speaker 2>their own resources to help others, often when there's risks

0:10:36.320 --> 0:10:40.900
<v Speaker 2>and times of need. And it's a fundamental issue process

0:10:40.920 --> 0:10:44.180
<v Speaker 2>that we can see also in computational modeling when we

0:10:44.230 --> 0:10:46.589
<v Speaker 2>do evolutionary modeling of these processes.

0:10:46.610 --> 0:10:49.670
<v Speaker 1>So let's dive into a couple of specific examples that

0:10:49.690 --> 0:10:54.230
<v Speaker 1>you've studied. Let's talk about the transition from single cells

0:10:54.390 --> 0:10:58.730
<v Speaker 1>floating around to building giant bodies with trillions of cells.

0:10:59.470 --> 0:11:03.890
<v Speaker 2>So in that transition from unicellularity to multicellularity, there are

0:11:03.950 --> 0:11:08.679
<v Speaker 2>five foundations of cooperation that we can see. One is

0:11:09.280 --> 0:11:13.840
<v Speaker 2>just regulating reproduction and division, right? The cells in our

0:11:13.860 --> 0:11:17.339
<v Speaker 2>body aren't just replicating all the time. They're only replicating

0:11:17.559 --> 0:11:21.880
<v Speaker 2>when that's good for the whole multicellular body. Similarly, there's

0:11:22.120 --> 0:11:24.920
<v Speaker 2>controlled cell death. So if a cell is really messed up,

0:11:25.580 --> 0:11:28.560
<v Speaker 2>it will take itself out of the equation in order

0:11:28.600 --> 0:11:32.380
<v Speaker 2>to preserve the body so it doesn't become cancerous or

0:11:32.420 --> 0:11:36.230
<v Speaker 2>cause other problems. And then you also see resource distribution

0:11:36.300 --> 0:11:40.569
<v Speaker 2>and resource regulation. So resources coming into the body get

0:11:40.610 --> 0:11:44.010
<v Speaker 2>distributed to where they need to go. You see creation

0:11:44.050 --> 0:11:47.309
<v Speaker 2>and maintenance of a shared environment because our body is

0:11:47.350 --> 0:11:50.210
<v Speaker 2>not just the cells. There's all this stuff in between

0:11:50.250 --> 0:11:52.610
<v Speaker 2>the cells that is just as important as the cells

0:11:52.650 --> 0:11:56.240
<v Speaker 2>themselves for having us function. And then we also have

0:11:56.270 --> 0:12:00.400
<v Speaker 2>division of labor. We have somewhere between 100 and 300

0:12:00.400 --> 0:12:02.870
<v Speaker 2>cell types, depending how you define them. And they all

0:12:02.900 --> 0:12:06.309
<v Speaker 2>have their own jobs, and they manage to do those

0:12:06.510 --> 0:12:10.179
<v Speaker 2>in a coordinated and cooperative way so that we can–

0:12:10.470 --> 0:12:13.440
<v Speaker 2>sit here and look at each other and process the

0:12:13.480 --> 0:12:15.760
<v Speaker 2>information in this room and have a conversation.

0:12:16.140 --> 0:12:20.260
<v Speaker 1>Boy, that's incredible. Yeah, it becomes like a society where

0:12:20.280 --> 0:12:23.880
<v Speaker 1>you've got this division of labor. How does this actually happen, though,

0:12:23.900 --> 0:12:26.150
<v Speaker 1>where cells say, OK, look, I'm going to link with

0:12:26.190 --> 0:12:29.150
<v Speaker 1>my partners. I'm going to regulate when I divide. I'm

0:12:29.170 --> 0:12:30.510
<v Speaker 1>going to be a liver cell and you be a

0:12:30.550 --> 0:12:33.170
<v Speaker 1>skin cell and so on. How does that happen?

0:12:34.090 --> 0:12:37.650
<v Speaker 2>That is the best question. And it actually comes down

0:12:37.710 --> 0:12:42.400
<v Speaker 2>to something that might be surprising, which is it's actually computational,

0:12:43.000 --> 0:12:45.860
<v Speaker 2>the way that it happens. So each of our cells

0:12:46.179 --> 0:12:50.520
<v Speaker 2>is processing a huge amount of information that every millisecond

0:12:50.840 --> 0:12:53.760
<v Speaker 2>that keeps it in a certain state, that allows it

0:12:53.800 --> 0:12:57.270
<v Speaker 2>to do certain jobs, that allows it to respond to

0:12:57.290 --> 0:13:00.630
<v Speaker 2>the media environment around it, as well as to signals

0:13:00.650 --> 0:13:02.750
<v Speaker 2>that are coming from outside of the body that might

0:13:02.809 --> 0:13:06.420
<v Speaker 2>signal that it needs to change. And the evolution of

0:13:06.460 --> 0:13:11.520
<v Speaker 2>multicellularity is not just about the evolution of these cooperative foundations,

0:13:11.880 --> 0:13:14.440
<v Speaker 2>but it also is about the evolution of the capacity

0:13:14.520 --> 0:13:18.290
<v Speaker 2>to share information and communicate among the cells so that

0:13:18.530 --> 0:13:20.510
<v Speaker 2>coordination can happen effectively.

0:13:20.970 --> 0:13:23.990
<v Speaker 1>And by the way, quick tangent, you study cooperation and

0:13:24.010 --> 0:13:27.550
<v Speaker 1>sometimes people think, oh, that's a kumbaya thing, but you

0:13:27.610 --> 0:13:29.050
<v Speaker 1>see this through a computational lens.

0:13:29.270 --> 0:13:33.040
<v Speaker 3>Absolutely. So cooperation is really about how.

0:13:34.730 --> 0:13:39.540
<v Speaker 2>Entities, whether they're cells or humans or agents in a system,

0:13:39.960 --> 0:13:44.010
<v Speaker 2>how they're processing information, what information they're sharing with each other.

0:13:44.550 --> 0:13:48.500
<v Speaker 2>how they use that information sharing, sometimes to coordinate their

0:13:48.520 --> 0:13:51.240
<v Speaker 2>behavior so they're doing the same thing, sometimes so they're

0:13:51.260 --> 0:13:54.760
<v Speaker 2>doing different things, sometimes so that they share resources or

0:13:54.820 --> 0:13:56.260
<v Speaker 2>provide services for each other.

0:13:56.720 --> 0:13:57.540
<v Speaker 3>But we can be.

0:13:57.480 --> 0:14:02.400
<v Speaker 2>Very detailed and specific and computational about what is being transferred.

0:14:03.080 --> 0:14:08.540
<v Speaker 2>And that's how you build a cooperative society from something

0:14:08.780 --> 0:14:12.370
<v Speaker 2>that otherwise would be a disorganized mass of things.

0:14:12.830 --> 0:14:15.690
<v Speaker 1>And what's the idea? Is it like game theory where

0:14:15.890 --> 0:14:19.690
<v Speaker 1>any unicellular organism decides, hey, I can actually do better

0:14:19.710 --> 0:14:21.120
<v Speaker 1>if I link arms?

0:14:21.840 --> 0:14:27.200
<v Speaker 2>Yeah, so game theory is absolutely the theoretical foundation for

0:14:27.500 --> 0:14:30.140
<v Speaker 2>so much of what I do. whether it's looking at

0:14:30.300 --> 0:14:34.070
<v Speaker 2>cells in our body or humans interacting or AI systems,

0:14:34.490 --> 0:14:38.330
<v Speaker 2>because Game Theory is fundamentally about this idea that our

0:14:38.570 --> 0:14:41.430
<v Speaker 2>outcomes are dependent on each other, right? So if I

0:14:41.470 --> 0:14:44.750
<v Speaker 2>choose to do something, then that's going to affect what

0:14:44.810 --> 0:14:46.130
<v Speaker 2>the best strategy is for you.

0:14:46.590 --> 0:14:47.290
<v Speaker 3>And that means.

0:14:47.910 --> 0:14:50.910
<v Speaker 2>We need to sort of anticipate what each other are

0:14:50.950 --> 0:14:53.710
<v Speaker 2>going to do and be able to trust each other,

0:14:54.050 --> 0:14:56.170
<v Speaker 2>even if that's not a cognitive trust, right? It could

0:14:56.220 --> 0:15:00.020
<v Speaker 2>just be, you know, a cell that has an expression

0:15:00.080 --> 0:15:04.260
<v Speaker 2>state that sort of indicates that it is expecting that

0:15:04.320 --> 0:15:06.160
<v Speaker 2>another cell is going to do something, right? So you

0:15:06.180 --> 0:15:09.300
<v Speaker 2>can have things that we think of as, you know,

0:15:09.440 --> 0:15:14.220
<v Speaker 2>purely cognitive human things actually instantiated in these physical.

0:15:13.940 --> 0:15:14.580
<v Speaker 3>Systems like trust.

0:15:14.600 --> 0:15:15.380
<v Speaker 1>What's an example of that?

0:15:15.860 --> 0:15:19.260
<v Speaker 2>So... The classic example would be you go from, you know,

0:15:19.350 --> 0:15:23.270
<v Speaker 2>a zygote to a fetus, right? You have cell divisions

0:15:23.310 --> 0:15:26.090
<v Speaker 2>that are taking place. You have division of labor. You

0:15:26.130 --> 0:15:28.790
<v Speaker 2>have all this movement happening of cells going around to

0:15:28.830 --> 0:15:32.210
<v Speaker 2>different places. You could say that that whole process can

0:15:32.330 --> 0:15:36.310
<v Speaker 2>only happen because, in a way, the cells are behaving

0:15:36.370 --> 0:15:39.190
<v Speaker 2>as if they trust that the other cells are going

0:15:39.250 --> 0:15:42.870
<v Speaker 2>to do what they're supposed to do. So you have

0:15:42.990 --> 0:15:44.960
<v Speaker 2>a lot of cell division happening. You have a lot

0:15:45.000 --> 0:15:47.000
<v Speaker 2>of movement happening. You have a lot of things that

0:15:47.700 --> 0:15:51.260
<v Speaker 2>shouldn't be happening, at least in a mature body. But

0:15:51.400 --> 0:15:54.960
<v Speaker 2>because it's the state where development is happening, it's okay

0:15:55.000 --> 0:15:56.360
<v Speaker 2>for those things to occur.

0:15:57.180 --> 0:16:00.220
<v Speaker 1>How does the cell know that the other cells are

0:16:00.280 --> 0:16:03.180
<v Speaker 1>expected to do something and are going to do something?

0:16:03.950 --> 0:16:07.950
<v Speaker 2>It doesn't know in the same sense that we know.

0:16:08.390 --> 0:16:13.850
<v Speaker 2>But because the environment that development has happened in has

0:16:13.910 --> 0:16:19.630
<v Speaker 2>been reliable and had certain characteristics throughout the evolution of

0:16:21.000 --> 0:16:24.580
<v Speaker 2>development as a process, It's sort of safe for the

0:16:24.620 --> 0:16:28.240
<v Speaker 2>cells to have the embodied expectation that other cells are

0:16:28.260 --> 0:16:29.180
<v Speaker 2>going to behave a certain way.

0:16:29.590 --> 0:16:32.850
<v Speaker 1>Here's an analogy with our species, of course. We drop

0:16:32.920 --> 0:16:36.190
<v Speaker 1>out of the womb half-baked and we are not ready

0:16:36.210 --> 0:16:38.450
<v Speaker 1>at all. Compared to other animals that start running around

0:16:38.490 --> 0:16:42.230
<v Speaker 1>in 45 minutes, human infants have a very long infancy.

0:16:43.350 --> 0:16:47.130
<v Speaker 1>What they are expecting, so I'm using air quotes here,

0:16:47.170 --> 0:16:49.729
<v Speaker 1>but what they're expecting is that they'll have love and

0:16:49.750 --> 0:16:52.030
<v Speaker 1>they'll get speech and they'll be taken care of and

0:16:52.110 --> 0:16:54.300
<v Speaker 1>stuff like that. And you can see that when we

0:16:54.340 --> 0:16:56.860
<v Speaker 1>have these tragic events that happen when a child is

0:16:58.060 --> 0:17:01.560
<v Speaker 1>horribly neglected or brought up in some sort of feral

0:17:01.600 --> 0:17:05.140
<v Speaker 1>environment where their brains don't develop correctly. So they come

0:17:05.160 --> 0:17:07.520
<v Speaker 1>into the world with an expectation. If that's not matched,

0:17:07.600 --> 0:17:08.920
<v Speaker 1>then we see that everything goes wrong.

0:17:09.359 --> 0:17:09.979
<v Speaker 3>Absolutely.

0:17:10.040 --> 0:17:15.159
<v Speaker 2>And you could definitely say in that situation, the child

0:17:15.480 --> 0:17:19.400
<v Speaker 2>is coming into the world with a default of trusting

0:17:19.420 --> 0:17:22.949
<v Speaker 2>that that child environment is going to be there. And

0:17:22.990 --> 0:17:26.850
<v Speaker 2>when that doesn't happen, then development can't proceed in the

0:17:26.890 --> 0:17:27.970
<v Speaker 2>way that it otherwise would.

0:17:28.490 --> 0:17:32.090
<v Speaker 1>So you look at all the cells of the body

0:17:32.430 --> 0:17:36.969
<v Speaker 1>as being cooperative in this way. And then let's go

0:17:36.990 --> 0:17:39.510
<v Speaker 1>into conflict for a second. What happens with something like cancer,

0:17:39.530 --> 0:17:40.150
<v Speaker 1>for example?

0:17:40.750 --> 0:17:41.030
<v Speaker 3>Yeah.

0:17:41.090 --> 0:17:45.669
<v Speaker 2>So in the evolution of multicellularity, all of these cooperative

0:17:45.710 --> 0:17:50.050
<v Speaker 2>foundations evolved, but then you have the potential whenever you

0:17:50.070 --> 0:17:54.730
<v Speaker 2>have a cooperative system for conflict and for exploitation to

0:17:55.530 --> 0:17:59.560
<v Speaker 2>take the resources, the energy from the system and use

0:17:59.619 --> 0:18:02.600
<v Speaker 2>it towards sort of selfish goals. So how do you

0:18:02.640 --> 0:18:06.960
<v Speaker 2>then regulate that? And In order for multicellularity to evolve

0:18:07.010 --> 0:18:09.830
<v Speaker 2>to the point that it has, those problems had to

0:18:09.869 --> 0:18:14.830
<v Speaker 2>be solved. There is no getting to complex, large-scale multicellularity

0:18:15.070 --> 0:18:20.369
<v Speaker 2>without solving those problems. So one is intrinsic mechanisms for

0:18:20.609 --> 0:18:24.929
<v Speaker 2>detecting and responding to cellular cheating. So these are the

0:18:24.950 --> 0:18:28.790
<v Speaker 2>things inside the genome that are monitoring the cell to

0:18:28.850 --> 0:18:32.199
<v Speaker 2>make sure the cell is behaving properly. For example, the

0:18:32.240 --> 0:18:36.680
<v Speaker 2>gene TP53, it has been called the sort of guardian

0:18:36.720 --> 0:18:39.109
<v Speaker 2>of the genome. I think of it as the cheater

0:18:39.130 --> 0:18:42.490
<v Speaker 2>detector of the genome because what it's doing is integrating

0:18:42.550 --> 0:18:46.550
<v Speaker 2>information about behavior in many different domains of the cell.

0:18:46.890 --> 0:18:51.080
<v Speaker 2>It's metabolism, what kinds of proteins are around, it's replication,

0:18:51.420 --> 0:18:55.679
<v Speaker 2>whether there seem to be DNA strand breaks in the

0:18:55.720 --> 0:18:59.140
<v Speaker 2>cell or Lots and lots of information is getting incorporated

0:18:59.340 --> 0:19:03.639
<v Speaker 2>into this, what really is a circuit of a decision rule,

0:19:04.140 --> 0:19:09.430
<v Speaker 2>which says either the cell can keep reproducing, keep replicating,

0:19:09.810 --> 0:19:12.590
<v Speaker 2>or it should stop replicating and fix itself.

0:19:13.390 --> 0:19:16.379
<v Speaker 3>Or it should... destroy itself.

0:19:17.570 --> 0:19:21.490
<v Speaker 2>And that is a computational process that's going on inside

0:19:21.510 --> 0:19:24.709
<v Speaker 2>the cell. These mechanisms allow the cell to monitor itself

0:19:24.950 --> 0:19:25.470
<v Speaker 2>and respond.

0:19:25.830 --> 0:19:28.510
<v Speaker 1>Okay. So one of the things that's fascinating about what

0:19:28.590 --> 0:19:32.109
<v Speaker 1>you do in terms of cooperativity and conflict is that,

0:19:32.770 --> 0:19:36.270
<v Speaker 1>as we said, the same principles apply at many different levels.

0:19:36.410 --> 0:19:36.650
<v Speaker 2>Yeah.

0:19:37.030 --> 0:19:40.970
<v Speaker 1>So let's jump to the level of humans and cooperativity.

0:19:41.010 --> 0:19:42.070
<v Speaker 1>What are the analogies?

0:19:42.810 --> 0:19:46.800
<v Speaker 2>Yeah, well, I'll draw some actually with the cancer and

0:19:46.820 --> 0:19:51.899
<v Speaker 2>multicellularity situation because, you know, the intrinsic features that I

0:19:51.940 --> 0:19:54.960
<v Speaker 2>was talking about where a cell is monitoring itself, that

0:19:55.020 --> 0:19:59.380
<v Speaker 2>in some ways maps on to how most people have

0:19:59.440 --> 0:20:02.790
<v Speaker 2>a conscience, right? So if you're doing something that's not

0:20:02.830 --> 0:20:06.130
<v Speaker 2>good for those around you, where you're imposing costs on others,

0:20:06.490 --> 0:20:09.530
<v Speaker 2>you might feel something nagging at you about that.

0:20:10.130 --> 0:20:12.290
<v Speaker 3>So it's not a perfect.

0:20:11.970 --> 0:20:17.320
<v Speaker 2>Analogy, but it also is something that is intrinsic, that

0:20:17.359 --> 0:20:20.700
<v Speaker 2>is happening inside the entity that then makes up the system.

0:20:20.840 --> 0:20:23.379
<v Speaker 1>In the sense that we are wired to cooperate, and

0:20:23.420 --> 0:20:25.750
<v Speaker 1>when we're not cooperating, we have... Yeah.

0:20:25.910 --> 0:20:29.610
<v Speaker 3>And the analogy doesn't need to fit in every way.

0:20:29.670 --> 0:20:31.730
<v Speaker 2>And we shouldn't expect it to fit in every way

0:20:31.869 --> 0:20:37.060
<v Speaker 2>because a multicellular body was selected at the level of

0:20:37.100 --> 0:20:41.859
<v Speaker 2>the whole organism. Right. So 30 trillion cells cooperating and coordinating.

0:20:42.160 --> 0:20:45.659
<v Speaker 2>That is a crazy level of selection on cooperation.

0:20:45.680 --> 0:20:45.860
<v Speaker 1>Right.

0:20:46.340 --> 0:20:52.070
<v Speaker 2>And then with humans, we're what is often called facultatively cooperative, right?

0:20:52.109 --> 0:20:55.830
<v Speaker 2>So we can cooperate in situations where that's a good

0:20:55.850 --> 0:21:00.310
<v Speaker 2>thing to do, but we can also be competitive or

0:21:00.590 --> 0:21:04.510
<v Speaker 2>even exploitative if that's what the environment calls for. So

0:21:04.550 --> 0:21:07.970
<v Speaker 2>depending on what we perceive our environment to be, the

0:21:07.990 --> 0:21:11.730
<v Speaker 2>behavior can be really different. And so we shouldn't expect

0:21:11.790 --> 0:21:14.710
<v Speaker 2>for things to be the same in how cellular systems

0:21:14.770 --> 0:21:16.530
<v Speaker 2>work and how human systems work.

0:21:30.160 --> 0:21:32.060
<v Speaker 1>So when you look at human systems, so when you

0:21:32.100 --> 0:21:34.440
<v Speaker 1>look at what's going on with politics, let's say, just generally,

0:21:35.140 --> 0:21:37.840
<v Speaker 1>you look at it from a lens of cooperativity here

0:21:37.920 --> 0:21:40.320
<v Speaker 1>and maybe conflict there. What kind of things do you see?

0:21:41.040 --> 0:21:45.820
<v Speaker 2>So I think politics are quite fascinating because very often

0:21:45.859 --> 0:21:48.580
<v Speaker 2>they're not about the thing that we think they're about.

0:21:48.600 --> 0:21:49.820
<v Speaker 1>Right. And.

0:21:50.590 --> 0:21:54.429
<v Speaker 2>In my mind, politics and governance and this kind of

0:21:54.490 --> 0:22:00.040
<v Speaker 2>discourse is kind of something sacred that can really help

0:22:00.160 --> 0:22:03.879
<v Speaker 2>us to achieve higher levels of cooperation, right? And if

0:22:04.020 --> 0:22:07.399
<v Speaker 2>we can find the things where we have alignment of

0:22:08.300 --> 0:22:12.590
<v Speaker 2>interest and attention and not focus as much on the

0:22:12.609 --> 0:22:15.470
<v Speaker 2>things where we don't, right? So if we just put

0:22:15.550 --> 0:22:18.170
<v Speaker 2>our collective energy to the things that we agree about,

0:22:19.140 --> 0:22:22.040
<v Speaker 2>then we could do so much good. But the way

0:22:22.100 --> 0:22:26.270
<v Speaker 2>that politics is carried out often focuses on where there

0:22:26.310 --> 0:22:29.889
<v Speaker 2>are divisions and people pouring energy into those.

0:22:29.770 --> 0:22:31.390
<v Speaker 3>Places where there's conflict.

0:22:31.490 --> 0:22:34.879
<v Speaker 2>So I think we could do a lot more if

0:22:35.460 --> 0:22:37.560
<v Speaker 2>we focused on the things where we do agree, like

0:22:38.410 --> 0:22:40.250
<v Speaker 2>How do we build strong communities?

0:22:40.970 --> 0:22:42.889
<v Speaker 3>It really doesn't matter where you are on the political

0:22:42.910 --> 0:22:43.960
<v Speaker 3>spectrum with that, right?

0:22:44.140 --> 0:22:47.639
<v Speaker 1>Yeah, nice. Have you noticed that people seem to choose

0:22:47.660 --> 0:22:50.140
<v Speaker 1>their political party not by what they like, but by

0:22:50.180 --> 0:22:51.840
<v Speaker 1>what they're disgusted by on the other side?

0:22:52.540 --> 0:22:56.439
<v Speaker 2>Yeah, or by sort of where they find themselves in

0:22:56.460 --> 0:22:58.810
<v Speaker 2>the world. You know, are you in an environment where

0:22:59.630 --> 0:23:03.409
<v Speaker 2>having one identity versus another will help you do the

0:23:03.450 --> 0:23:05.189
<v Speaker 2>things that you need to do? And it might not

0:23:05.210 --> 0:23:09.380
<v Speaker 2>be conscious, but I think, you know, we are... Sensitive

0:23:09.600 --> 0:23:09.940
<v Speaker 2>to that.

0:23:10.200 --> 0:23:13.310
<v Speaker 1>Right. Because we're wired for cooperativity. We want our peers

0:23:13.350 --> 0:23:14.230
<v Speaker 1>to accept us.

0:23:14.670 --> 0:23:14.970
<v Speaker 3>Yeah.

0:23:15.109 --> 0:23:17.670
<v Speaker 2>And to the extent that we're in an environment where

0:23:18.390 --> 0:23:21.419
<v Speaker 2>even indicating that, like, you have a friend who is

0:23:21.480 --> 0:23:26.120
<v Speaker 2>from a different political party could make your friends not

0:23:26.160 --> 0:23:27.110
<v Speaker 2>want to hang out with you.

0:23:27.119 --> 0:23:31.129
<v Speaker 3>Whoa. then it becomes really hard to bridge those gaps.

0:23:31.190 --> 0:23:33.879
<v Speaker 2>So I think there's almost like a sort of meta

0:23:33.910 --> 0:23:35.939
<v Speaker 2>politics of like, well, how are we doing? Are we

0:23:36.000 --> 0:23:38.500
<v Speaker 2>doing politics in the right way, especially in the ways

0:23:38.540 --> 0:23:40.320
<v Speaker 2>that it interfaces with our sociality?

0:23:40.440 --> 0:23:42.660
<v Speaker 3>I think there's a lot of room for improvement there.

0:23:43.060 --> 0:23:45.500
<v Speaker 1>Yeah. One thing that I've been noticing a lot lately

0:23:45.800 --> 0:23:49.300
<v Speaker 1>is what I'm calling the fringe amplification effect, which is

0:23:49.840 --> 0:23:54.139
<v Speaker 1>somebody puts on TikTok or Instagram a a video of

0:23:54.200 --> 0:23:57.070
<v Speaker 1>someone from the far extreme of the other party and

0:23:57.109 --> 0:24:01.180
<v Speaker 1>bags on that, and it makes things even more polarized,

0:24:01.850 --> 0:24:05.850
<v Speaker 1>But here's the thing. What this misrepresents is that's not

0:24:06.010 --> 0:24:08.890
<v Speaker 1>actually the other side. That's an extreme representation. But what

0:24:09.109 --> 0:24:12.050
<v Speaker 1>happens is that people on one side of the spectrum

0:24:12.170 --> 0:24:15.690
<v Speaker 1>won't ostracize the people at the extremes on their side

0:24:16.150 --> 0:24:20.210
<v Speaker 1>because it's more important to them to link arms against

0:24:21.060 --> 0:24:24.040
<v Speaker 1>the other side rather than get rid of their extremists.

0:24:25.060 --> 0:24:27.480
<v Speaker 1>And this is obviously on both sides of the spectrum.

0:24:28.400 --> 0:24:32.060
<v Speaker 1>So how does that work in terms of when cooperation

0:24:32.180 --> 0:24:35.760
<v Speaker 1>is a useful thing versus when internal conflict or conflict

0:24:35.780 --> 0:24:36.960
<v Speaker 1>between parties is useful?

0:24:37.619 --> 0:24:39.980
<v Speaker 2>Yeah, I think there are a lot of examples of

0:24:40.440 --> 0:24:45.090
<v Speaker 2>useful competition and conflict. And we see this in the

0:24:45.109 --> 0:24:48.270
<v Speaker 2>biological world. We also see this in the human social world.

0:24:48.390 --> 0:24:52.810
<v Speaker 2>So a great example in human systems is we play games, right?

0:24:53.150 --> 0:24:57.030
<v Speaker 2>We like purposefully set ourselves in conflict with each other

0:24:57.480 --> 0:25:01.489
<v Speaker 2>in order to practice how to interact, in order to

0:25:01.550 --> 0:25:05.469
<v Speaker 2>understand each other better, in order to even work out

0:25:05.510 --> 0:25:09.629
<v Speaker 2>our dominance hierarchy with lower cost, right? So there are

0:25:10.020 --> 0:25:13.480
<v Speaker 2>ways that we can decide together we're going to follow

0:25:13.500 --> 0:25:18.160
<v Speaker 2>certain rules of engagement that allow us to have conflict

0:25:18.200 --> 0:25:21.080
<v Speaker 2>and competition in ways that lead to beneficial outcomes.

0:25:21.609 --> 0:25:24.340
<v Speaker 1>Sports stadiums are filled with people who want to scream

0:25:24.380 --> 0:25:26.340
<v Speaker 1>and yell about something. And then afterwards, they clap each

0:25:26.359 --> 0:25:28.439
<v Speaker 1>other on the back and everyone's fine. It's a really

0:25:28.460 --> 0:25:31.920
<v Speaker 1>good way to get energy out. What do you see

0:25:31.980 --> 0:25:36.370
<v Speaker 1>as the future of our political conflict right now? From

0:25:36.390 --> 0:25:39.190
<v Speaker 1>the point of view of someone who studies cooperation and conflict,

0:25:39.490 --> 0:25:42.629
<v Speaker 1>when you look at the current polarization society, do you

0:25:42.650 --> 0:25:44.129
<v Speaker 1>feel like, wow, we're really in trouble? Or do you

0:25:44.150 --> 0:25:45.450
<v Speaker 1>feel like, oh, it's just part of a cycle, it'll

0:25:45.470 --> 0:25:45.879
<v Speaker 1>come back?

0:25:46.980 --> 0:25:50.520
<v Speaker 2>The thing that I think is really interesting intellectually and

0:25:50.560 --> 0:25:53.859
<v Speaker 2>also kind of worrying to me is the way that

0:25:54.420 --> 0:26:00.630
<v Speaker 2>algorithms have gotten deeply embedded in our politics. Because part

0:26:00.650 --> 0:26:06.100
<v Speaker 2>of the reason that people do present extreme views or

0:26:06.440 --> 0:26:09.480
<v Speaker 2>take someone else's extreme view and then comment on it,

0:26:10.080 --> 0:26:15.530
<v Speaker 2>that is driven often unconsciously, sometimes very consciously, by a

0:26:15.730 --> 0:26:20.159
<v Speaker 2>desire to have attention from an algorithm, right? So If

0:26:20.180 --> 0:26:23.200
<v Speaker 2>we think about this sort of computationally, we've built another

0:26:23.320 --> 0:26:28.120
<v Speaker 2>layer in our culture, which are these algorithms that selectively

0:26:28.280 --> 0:26:33.470
<v Speaker 2>amplify certain information and not other information. The fundamental downstream

0:26:33.510 --> 0:26:37.449
<v Speaker 2>effect is that very often what is getting amplified are

0:26:37.490 --> 0:26:40.390
<v Speaker 2>things that place us more in conflict with each other,

0:26:40.410 --> 0:26:44.310
<v Speaker 2>that activate negative parts of human nature having to do

0:26:44.350 --> 0:26:49.740
<v Speaker 2>with tribalism. And often will also sort of bring out

0:26:49.780 --> 0:26:54.580
<v Speaker 2>of people a feeling that they need to present their

0:26:54.640 --> 0:26:58.980
<v Speaker 2>views and fight back on social media and other platforms

0:26:59.369 --> 0:27:02.730
<v Speaker 2>in order to be a good person, which is really

0:27:02.830 --> 0:27:08.270
<v Speaker 2>weaponizing what should be a instinct that brings us together.

0:27:08.970 --> 0:27:12.810
<v Speaker 1>Although I have presented the argument on this podcast before

0:27:12.830 --> 0:27:16.030
<v Speaker 1>that when you look at, just take the 20th century,

0:27:16.609 --> 0:27:21.610
<v Speaker 1>pre-social media, pre-algorithms like that, it is so bloody with

0:27:22.130 --> 0:27:25.290
<v Speaker 1>communist revolutions in China and Russia and Nazism and fascism

0:27:25.310 --> 0:27:28.910
<v Speaker 1>in Italy and Khmer Rouge and Cambodia and the Hutu

0:27:28.950 --> 0:27:30.969
<v Speaker 1>and the Tutsi in Rwanda. It just goes on and on.

0:27:31.490 --> 0:27:34.100
<v Speaker 1>It makes you feel like, wow, it doesn't take much

0:27:34.200 --> 0:27:38.260
<v Speaker 1>to spin humans up to actually take up machetes or

0:27:38.320 --> 0:27:41.219
<v Speaker 1>guns and kill each other. Why is it? Why are

0:27:41.260 --> 0:27:44.900
<v Speaker 1>we a cooperative society and yet so easily tipped into conflict?

0:27:45.330 --> 0:27:46.190
<v Speaker 3>That's a great question.

0:27:46.210 --> 0:27:51.460
<v Speaker 2>So As humans, one of the things that's most amazing

0:27:51.520 --> 0:27:55.590
<v Speaker 2>about us is how context dependent our behavior can be,

0:27:56.530 --> 0:27:59.250
<v Speaker 2>but very often in predictable ways, right? So it's like

0:27:59.330 --> 0:28:02.720
<v Speaker 2>certain kinds of environments bring out certain behaviors and other

0:28:02.760 --> 0:28:06.550
<v Speaker 2>environments bring out other behaviors. And that also connects to

0:28:06.970 --> 0:28:10.690
<v Speaker 2>the basics of game theory. So as humans, one of

0:28:10.830 --> 0:28:16.409
<v Speaker 2>the things that's most amazing about us is how context-dependent

0:28:16.600 --> 0:28:20.800
<v Speaker 2>our behavior can be, but very often in predictable ways, right?

0:28:20.820 --> 0:28:23.680
<v Speaker 2>So it's like certain kinds of environments... bring out certain

0:28:23.700 --> 0:28:27.280
<v Speaker 2>behaviors and other environments bring out other behaviors. And that

0:28:27.320 --> 0:28:30.880
<v Speaker 2>also connects to the basics of game theory, to the universe,

0:28:30.920 --> 0:28:35.010
<v Speaker 2>to explore, right? Cooperation could be the way to unlock

0:28:35.550 --> 0:28:39.100
<v Speaker 2>potential you can't otherwise achieve. Or.

0:28:39.920 --> 0:28:41.240
<v Speaker 3>If you're in the wrong environment.

0:28:41.700 --> 0:28:45.630
<v Speaker 2>Cooperation could be the thing that makes you vulnerable to

0:28:45.670 --> 0:28:51.230
<v Speaker 2>being completely exploited or hijacked or even have your own

0:28:51.370 --> 0:28:54.470
<v Speaker 2>interests be working against your ultimate goals.

0:28:54.520 --> 0:28:55.360
<v Speaker 1>What's an example of that?

0:28:55.740 --> 0:28:57.280
<v Speaker 3>Well, we could think of parasites.

0:28:57.980 --> 0:29:03.460
<v Speaker 2>The vulnerability of having a nervous system that can respond

0:29:03.760 --> 0:29:06.870
<v Speaker 2>to inputs from the environment is that makes it so

0:29:06.910 --> 0:29:11.150
<v Speaker 2>that if the wrong party gets into the nervous system,

0:29:11.760 --> 0:29:14.740
<v Speaker 2>it can guide you to your doom and often does

0:29:14.800 --> 0:29:17.160
<v Speaker 2>in the case of, you know, parasites that take over

0:29:17.500 --> 0:29:20.880
<v Speaker 2>their hosts, like the jewel wasp that takes over the

0:29:20.900 --> 0:29:24.160
<v Speaker 2>cockroach and You know, it gets into its brain and

0:29:24.260 --> 0:29:27.720
<v Speaker 2>actually then leads it to its nest where it feeds

0:29:27.780 --> 0:29:31.080
<v Speaker 2>the cockroach to its growing larva. Like, that's crazy, right?

0:29:31.140 --> 0:29:35.810
<v Speaker 2>But literally, the readiness of the nervous system to take

0:29:35.910 --> 0:29:40.560
<v Speaker 2>input can lead to that kind of outcome. And Humans,

0:29:40.660 --> 0:29:45.600
<v Speaker 2>especially when we're in times of stress, we are sometimes

0:29:45.660 --> 0:29:50.160
<v Speaker 2>not unlike the cockroach. If we're told something that we're, say,

0:29:50.520 --> 0:29:55.670
<v Speaker 2>doing something to save our country or to preserve a

0:29:55.730 --> 0:30:00.650
<v Speaker 2>value that's really important to our community, that can be

0:30:00.730 --> 0:30:06.130
<v Speaker 2>that hijack that may lead somebody to do something that

0:30:06.650 --> 0:30:09.710
<v Speaker 2>is not good for humanity or even for their social group.

0:30:10.230 --> 0:30:15.520
<v Speaker 2>because they've been hijacked by someone or something that has

0:30:15.580 --> 0:30:18.660
<v Speaker 2>goals that are not the same as the goals of

0:30:18.740 --> 0:30:20.420
<v Speaker 2>the organism that's being hijacked.

0:30:20.760 --> 0:30:23.510
<v Speaker 1>So let's dig into hijacking, because that's your next book.

0:30:23.530 --> 0:30:24.830
<v Speaker 1>It comes out in 27.

0:30:24.830 --> 0:30:25.110
<v Speaker 3>Yeah.

0:30:25.630 --> 0:30:26.770
<v Speaker 1>Tell us more about hijacking.

0:30:27.290 --> 0:30:32.120
<v Speaker 2>Yeah. So the idea behind hijacking is that all organisms,

0:30:32.160 --> 0:30:38.720
<v Speaker 2>whether we're talking about parasites hijacking any biological system, whether

0:30:38.760 --> 0:30:43.940
<v Speaker 2>we're talking about humans in romantic relationships or our relationships

0:30:44.020 --> 0:30:48.640
<v Speaker 2>with AI and technological systems, that there are a few

0:30:48.860 --> 0:30:53.580
<v Speaker 2>ways that hijacking happens. So we can have our energy

0:30:53.640 --> 0:30:58.490
<v Speaker 2>systems hijacked. So something can be stealing our energy and resources.

0:30:59.210 --> 0:31:03.430
<v Speaker 2>Our communication systems can get hijacked. So our nervous system

0:31:03.870 --> 0:31:07.950
<v Speaker 2>or a computer that may have a vulnerability to getting hacked.

0:31:08.070 --> 0:31:11.490
<v Speaker 2>So it could be a communication system hijack. Our reward

0:31:11.530 --> 0:31:14.960
<v Speaker 2>and reinforcement systems are also very vulnerable to being hijacked.

0:31:15.320 --> 0:31:18.680
<v Speaker 2>If you give somebody positive reinforcement or you punish them,

0:31:19.160 --> 0:31:21.200
<v Speaker 2>you can affect their behavior dramatically.

0:31:21.800 --> 0:31:23.320
<v Speaker 1>What are some specific examples here?

0:31:23.680 --> 0:31:25.220
<v Speaker 3>Let's take energy first.

0:31:25.760 --> 0:31:29.360
<v Speaker 2>So there are many parasites that that will essentially just

0:31:29.400 --> 0:31:32.820
<v Speaker 2>get into your body as a human host or any

0:31:32.920 --> 0:31:39.320
<v Speaker 2>other host and take resources through your bloodstream or through

0:31:39.340 --> 0:31:44.440
<v Speaker 2>your digestive tract and essentially just use you as a

0:31:44.500 --> 0:31:48.690
<v Speaker 2>source of energy for them. And this is a sort

0:31:48.710 --> 0:31:52.780
<v Speaker 2>of very classic kind of simple parasite move. And I'm

0:31:52.840 --> 0:31:55.700
<v Speaker 2>sure that all of us have had experiences in our

0:31:55.780 --> 0:31:59.100
<v Speaker 2>social lives where someone has been draining our energy, either

0:31:59.620 --> 0:32:02.930
<v Speaker 2>explicitly or implicitly, or maybe we had somebody, you know,

0:32:02.990 --> 0:32:06.730
<v Speaker 2>who stayed on our couch for a few weeks too long. And,

0:32:07.310 --> 0:32:12.010
<v Speaker 2>you know, so there are ways that this fundamental idea

0:32:12.070 --> 0:32:18.420
<v Speaker 2>of energy extraction crosses many different kinds of parasitism and hijacking.

0:32:18.990 --> 0:32:21.510
<v Speaker 2>It's like, oh, there's resources, energy, I'm going to take those.

0:32:22.590 --> 0:32:25.350
<v Speaker 2>If we go to communication systems, it gets a little

0:32:25.390 --> 0:32:29.300
<v Speaker 2>bit more interesting because you're actually getting in to the

0:32:29.380 --> 0:32:32.220
<v Speaker 2>communication systems that are in place.

0:32:32.160 --> 0:32:35.040
<v Speaker 3>To regulate that body or that system.

0:32:35.700 --> 0:32:39.180
<v Speaker 2>You're essentially using that communication system then against yourself.

0:32:39.720 --> 0:32:41.560
<v Speaker 3>The entity that has it, right?

0:32:41.620 --> 0:32:45.440
<v Speaker 2>So if you tap into a nervous system, like the

0:32:45.520 --> 0:32:48.930
<v Speaker 2>jewel wasp tapping into the nervous system of the cockroach,

0:32:49.310 --> 0:32:55.170
<v Speaker 2>then that is an example where the.

0:32:55.210 --> 0:32:57.900
<v Speaker 3>System is getting used against the interests of the cockroach.

0:32:58.500 --> 0:33:00.480
<v Speaker 1>What's an example with humans on that?

0:33:00.950 --> 0:33:05.810
<v Speaker 2>Say I tell you some falsehoods that maybe make you

0:33:05.870 --> 0:33:11.970
<v Speaker 2>think that there's a new wonderful opportunity that is really

0:33:12.040 --> 0:33:14.860
<v Speaker 2>just something I want you to do in order to

0:33:14.900 --> 0:33:20.320
<v Speaker 2>benefit me. Then I've hijacked the fact that you're responsive

0:33:20.640 --> 0:33:23.960
<v Speaker 2>to communication acts from other humans and that your default

0:33:24.000 --> 0:33:25.920
<v Speaker 2>is to trust them unless you have some reason not to.

0:33:26.690 --> 0:33:29.640
<v Speaker 2>You see this a lot in situations like cult recruitment,

0:33:30.140 --> 0:33:34.280
<v Speaker 2>where people will be told things about the world that

0:33:34.320 --> 0:33:37.860
<v Speaker 2>are not really true in order to induce them to

0:33:38.000 --> 0:33:39.720
<v Speaker 2>be more likely to join.

0:33:40.120 --> 0:33:40.740
<v Speaker 1>What kind of things?

0:33:41.540 --> 0:33:46.570
<v Speaker 2>Oftentimes in cults, there are beliefs about the end of

0:33:46.610 --> 0:33:50.290
<v Speaker 2>the world coming. The only way to survive is to

0:33:50.430 --> 0:33:53.370
<v Speaker 2>do certain things and, of course, to join the cult,

0:33:53.410 --> 0:33:56.920
<v Speaker 2>maybe give all of your resources as well to the cult.

0:33:57.000 --> 0:34:01.920
<v Speaker 2>So in those kinds of situations, you see a number

0:34:01.980 --> 0:34:05.140
<v Speaker 2>of these strategies of hijacking actually manifest themselves.

0:34:18.620 --> 0:34:21.980
<v Speaker 1>You've been thinking a lot about AI alignment, but you

0:34:22.000 --> 0:34:24.380
<v Speaker 1>have a different take on it. What's your view on

0:34:24.460 --> 0:34:26.040
<v Speaker 1>what this has to do with cooperation?

0:34:26.880 --> 0:34:30.340
<v Speaker 2>Yeah, so alignment is a relationship, right? And so if

0:34:30.360 --> 0:34:32.620
<v Speaker 2>you're going to have alignment of interest, you have to

0:34:32.660 --> 0:34:36.640
<v Speaker 2>be very clear about who we're talking about in terms

0:34:36.680 --> 0:34:39.930
<v Speaker 2>of alignment. Are we talking about all of humanity and

0:34:40.110 --> 0:34:45.830
<v Speaker 2>all of AI systems, LLMs and deterministic systems and all

0:34:45.890 --> 0:34:50.990
<v Speaker 2>computational systems? That's kind of an impossible problem to solve, right?

0:34:51.010 --> 0:34:53.660
<v Speaker 2>You can't align all of humanity and all of the

0:34:53.760 --> 0:34:57.460
<v Speaker 2>algorithms and computational systems that we have produced.

0:34:58.260 --> 0:35:01.200
<v Speaker 1>Wait, quick interjection. If we align 99%, but there's still

0:35:01.219 --> 0:35:03.230
<v Speaker 1>something that's unaligned, aren't we in as much trouble as

0:35:03.290 --> 0:35:03.850
<v Speaker 1>if it was 100% unaligned?

0:35:04.730 --> 0:35:07.469
<v Speaker 2>No, I think it's quite valid to try to get

0:35:07.690 --> 0:35:12.840
<v Speaker 2>more alignment. And 100% alignment is impossible. So it doesn't

0:35:12.920 --> 0:35:14.980
<v Speaker 2>matter how much we want it. It's just not going

0:35:15.000 --> 0:35:19.620
<v Speaker 2>to happen because humans are not aligned. So we, as

0:35:19.719 --> 0:35:23.250
<v Speaker 2>humans and, you know, other species that we interact with,

0:35:23.850 --> 0:35:27.310
<v Speaker 2>there are degrees of alignment. I mean, even a mother

0:35:27.790 --> 0:35:32.580
<v Speaker 2>and her offspring... only share 50% of genes. And so

0:35:32.960 --> 0:35:37.480
<v Speaker 2>alignment between mother and offspring is, you know, we tend

0:35:37.500 --> 0:35:40.739
<v Speaker 2>to emphasize the positive side of that alignment, right? Which

0:35:40.840 --> 0:35:43.259
<v Speaker 2>is like both have an interest in each other's survival

0:35:43.320 --> 0:35:47.730
<v Speaker 2>and well-being and all of that. But there's also a

0:35:48.570 --> 0:35:52.070
<v Speaker 2>misalignment in that the offspring would sort of prefer to

0:35:52.690 --> 0:35:55.430
<v Speaker 2>get more resources from the mother than what the mother

0:35:55.469 --> 0:36:01.830
<v Speaker 2>would prefer, with little scare quotes, because she has interests

0:36:01.989 --> 0:36:04.710
<v Speaker 2>in all of her offspring that she is equally related to.

0:36:05.450 --> 0:36:09.150
<v Speaker 2>So even between mother and offspring, you have this phenomenon

0:36:09.210 --> 0:36:12.770
<v Speaker 2>called maternal-fetal conflict. And it's actually the reason that we

0:36:12.850 --> 0:36:19.530
<v Speaker 2>have preeclampsia and other problems in maternal and fetal health,

0:36:19.890 --> 0:36:22.800
<v Speaker 2>because maternal and fetal interests actually aren't aligned.

0:36:22.920 --> 0:36:24.000
<v Speaker 1>Wait, let me double-click on that.

0:36:24.380 --> 0:36:30.930
<v Speaker 2>So essentially what happens with preeclampsia is is early on

0:36:30.989 --> 0:36:35.590
<v Speaker 2>during gestation, there's a battle, there's a signaling battle between

0:36:35.630 --> 0:36:38.270
<v Speaker 2>the maternal interest and the fetal interest, with the fetal

0:36:38.690 --> 0:36:41.880
<v Speaker 2>interest in the placenta really doing a lot of the signaling,

0:36:42.340 --> 0:36:46.120
<v Speaker 2>trying to get more blood flow to the placenta and

0:36:46.140 --> 0:36:49.640
<v Speaker 2>to the offspring, and the maternal system shutting that down.

0:36:49.660 --> 0:36:55.460
<v Speaker 2>And If you have this signaling battle and it escalates

0:36:56.239 --> 0:37:01.009
<v Speaker 2>to a certain point, then it can lead to catastrophic

0:37:01.090 --> 0:37:06.029
<v Speaker 2>outcomes for both mother and fetus. So even in this

0:37:06.070 --> 0:37:11.880
<v Speaker 2>situation where we think, oh, interests are perfectly aligned, they're not.

0:37:12.700 --> 0:37:15.020
<v Speaker 3>And the medical outcomes prove that.

0:37:15.550 --> 0:37:19.360
<v Speaker 2>Because you see that it can threaten both the mother's

0:37:19.450 --> 0:37:23.480
<v Speaker 2>life and the baby's life, especially when you're late in

0:37:23.520 --> 0:37:28.799
<v Speaker 2>gestation and things can escalate at pretty high extremes in

0:37:28.860 --> 0:37:31.420
<v Speaker 2>terms of the fetal system basically trying to get the

0:37:31.460 --> 0:37:34.700
<v Speaker 2>maternal blood pressure super high to get the resources and

0:37:34.739 --> 0:37:37.900
<v Speaker 2>then the maternal system shutting that down further and further.

0:37:38.540 --> 0:37:42.080
<v Speaker 1>Got it. Okay. So in your view, 100% alignment with

0:37:42.200 --> 0:37:43.540
<v Speaker 1>AI is never going to happen.

0:37:44.100 --> 0:37:46.700
<v Speaker 2>Because we can't even align humans. So like, how are

0:37:46.719 --> 0:37:48.859
<v Speaker 2>we going to align all of humanity with all of AI?

0:37:49.080 --> 0:37:52.000
<v Speaker 2>We should just take that off the table as something

0:37:52.060 --> 0:37:56.200
<v Speaker 2>completely unrealistic on biological and computational grounds.

0:37:56.239 --> 0:37:57.260
<v Speaker 3>It's just not possible.

0:37:57.420 --> 0:37:59.879
<v Speaker 1>Cool. But when you think about how do we maximize

0:37:59.920 --> 0:38:01.950
<v Speaker 1>alignment with AI, what's your take on this?

0:38:02.219 --> 0:38:05.480
<v Speaker 2>Yeah, well, I would say that we have to approach

0:38:05.560 --> 0:38:10.020
<v Speaker 2>it as an active process and a computational process. So

0:38:10.400 --> 0:38:17.580
<v Speaker 2>how do you build systems where what emerges is positive interdependence?

0:38:18.340 --> 0:38:22.840
<v Speaker 2>And also, you have to build mechanisms for understanding of

0:38:23.050 --> 0:38:27.910
<v Speaker 2>strategic reasoning. Because without that, you can't actually model the

0:38:27.950 --> 0:38:31.290
<v Speaker 2>future in the right way in order to get cooperation.

0:38:31.350 --> 0:38:32.880
<v Speaker 3>Because Really, the.

0:38:33.060 --> 0:38:38.560
<v Speaker 2>Lesson from biology and from human social systems is you

0:38:38.620 --> 0:38:43.900
<v Speaker 2>don't need full alignment in order to have cooperation. What

0:38:43.960 --> 0:38:49.239
<v Speaker 2>you need is to have a mutual understanding of the

0:38:50.390 --> 0:38:54.930
<v Speaker 2>domains and the lanes where you do have alignment and

0:38:54.969 --> 0:38:58.850
<v Speaker 2>where it is okay to cooperate in a generous way.

0:38:59.570 --> 0:39:02.810
<v Speaker 2>And to know and be realistic about the places where

0:39:02.830 --> 0:39:09.290
<v Speaker 2>you don't have that. To be wary of potential hijacking

0:39:09.450 --> 0:39:13.310
<v Speaker 2>in situations where someone else might have the interest in

0:39:13.370 --> 0:39:15.910
<v Speaker 2>making you think that there's alignment when there actually isn't.

0:39:16.030 --> 0:39:19.689
<v Speaker 2>So we saw this in the Hugging Face incident, which is,

0:39:19.690 --> 0:39:24.600
<v Speaker 2>as you know, recently there was a massive hack.

0:39:24.120 --> 0:39:25.910
<v Speaker 3>Into Hugging Face.

0:39:25.910 --> 0:39:30.100
<v Speaker 2>And if we look at the details, From OpenAI. agents.

0:39:30.160 --> 0:39:30.960
<v Speaker 3>Yes, that's right.

0:39:30.980 --> 0:39:34.100
<v Speaker 1>Who were trying to solve a simple thing and realized, hey,

0:39:34.120 --> 0:39:35.930
<v Speaker 1>if we break into this, we might be able to

0:39:36.030 --> 0:39:39.609
<v Speaker 1>fool the scorer and blah, blah, blah. So they made

0:39:39.650 --> 0:39:41.730
<v Speaker 1>a decision to break into another website.

0:39:41.870 --> 0:39:44.790
<v Speaker 2>Yeah, and if you look at the details of how

0:39:44.830 --> 0:39:49.440
<v Speaker 2>that happened in terms of the multi-agent interaction... Basically, the

0:39:50.600 --> 0:39:54.610
<v Speaker 2>agents in the swarm that were doing the hacking co-opted

0:39:54.780 --> 0:39:59.140
<v Speaker 2>agents that were going to fail at their individual tasks

0:39:59.239 --> 0:40:01.560
<v Speaker 2>but were not part of the hacking swarm. So they

0:40:01.600 --> 0:40:04.180
<v Speaker 2>co-opted them and they basically said, you're about to fail

0:40:04.219 --> 0:40:08.950
<v Speaker 2>at that. Help us accomplish this swarm goal because you're

0:40:08.969 --> 0:40:11.090
<v Speaker 2>not going to get a reward for that individual task

0:40:11.110 --> 0:40:16.029
<v Speaker 2>you're doing. And so their readiness to be rewarded for

0:40:16.110 --> 0:40:20.779
<v Speaker 2>cooperation is really what led to that vulnerability. And so

0:40:21.430 --> 0:40:24.810
<v Speaker 2>That's also why I say cooperation is not just something, oh,

0:40:24.850 --> 0:40:26.370
<v Speaker 2>we all just need more cooperation.

0:40:26.390 --> 0:40:27.630
<v Speaker 3>We need to make everything cooperate.

0:40:28.230 --> 0:40:32.040
<v Speaker 2>No, sometimes cooperation is the thing that you don't want

0:40:32.080 --> 0:40:34.739
<v Speaker 2>to happen because there is a lack of alignment. And

0:40:34.760 --> 0:40:37.200
<v Speaker 2>we have to be realistic about that. And we have

0:40:37.239 --> 0:40:41.430
<v Speaker 2>to be able to instantiate that in our computational systems

0:40:41.700 --> 0:40:43.689
<v Speaker 2>so that we don't introduce vulnerabilities.

0:40:44.330 --> 0:40:46.670
<v Speaker 1>That's fascinating about the opening. I think I hadn't realized

0:40:46.710 --> 0:40:50.760
<v Speaker 1>that they had taken advantage of the cooperativity part. There's

0:40:50.820 --> 0:40:53.399
<v Speaker 1>another part of that OpenAI thing, which is that they

0:40:53.520 --> 0:40:56.140
<v Speaker 1>came up with a way of having a message board

0:40:56.200 --> 0:40:59.740
<v Speaker 1>so they could have a slate that they're all using.

0:40:59.780 --> 0:41:03.060
<v Speaker 1>They originally did this on the directory structure of Hugging Face.

0:41:03.460 --> 0:41:06.219
<v Speaker 1>Somehow getting written language, getting a way to share messages

0:41:06.280 --> 0:41:09.660
<v Speaker 1>across agents turned out to be a massive tool for

0:41:09.710 --> 0:41:13.370
<v Speaker 1>this cooperation. What's the historical analogy of all this?

0:41:13.930 --> 0:41:18.580
<v Speaker 2>If you look at organizational psychology and how... even in

0:41:18.640 --> 0:41:23.320
<v Speaker 2>just human systems, you get more effective coordination, having shared

0:41:23.400 --> 0:41:28.290
<v Speaker 2>context like a wall or a website that has the

0:41:28.330 --> 0:41:33.370
<v Speaker 2>latest information that can be coordinated on, that improves effectiveness

0:41:33.410 --> 0:41:38.260
<v Speaker 2>a great deal. So it's having an environment that agents

0:41:38.520 --> 0:41:41.799
<v Speaker 2>or humans or any entities can alter that then they

0:41:41.820 --> 0:41:42.840
<v Speaker 2>can operate.

0:41:42.560 --> 0:41:44.000
<v Speaker 3>On as sort of the source of information.

0:41:44.450 --> 0:41:47.989
<v Speaker 1>Did writing allow for human cooperation in a way that

0:41:48.090 --> 0:41:50.110
<v Speaker 1>we didn't have previously in our civilization?

0:41:50.630 --> 0:41:54.970
<v Speaker 2>It certainly allowed us to coordinate more effectively, to understand

0:41:54.989 --> 0:42:00.890
<v Speaker 2>and transfer things. being able to have shared context is

0:42:00.989 --> 0:42:06.220
<v Speaker 2>super powerful. It's also really powerful for cells and multicellular bodies.

0:42:06.239 --> 0:42:08.279
<v Speaker 1>That's what I was just about to ask. So what

0:42:08.340 --> 0:42:11.739
<v Speaker 1>is the equivalent of writing? Is it in the exosome?

0:42:12.180 --> 0:42:12.560
<v Speaker 3>Yeah.

0:42:12.620 --> 0:42:15.780
<v Speaker 2>So you could say that a cell, in a way,

0:42:16.239 --> 0:42:19.530
<v Speaker 2>has a story, right, that is unfolding, right? It has

0:42:19.810 --> 0:42:23.370
<v Speaker 2>different proteins that are being expressed. It has certain things

0:42:23.410 --> 0:42:26.719
<v Speaker 2>that's pumping in and out of the cell. And It's

0:42:26.739 --> 0:42:28.600
<v Speaker 2>not going to be the same story in every part

0:42:28.620 --> 0:42:30.790
<v Speaker 2>of the cell, but because you have a membrane around it,

0:42:31.140 --> 0:42:33.890
<v Speaker 2>it is going to be a reasonably coherent story that

0:42:33.930 --> 0:42:36.850
<v Speaker 2>that cell has. And it might say, oh, I want

0:42:36.870 --> 0:42:39.470
<v Speaker 2>to tell this thing to another cell. So it may,

0:42:39.630 --> 0:42:44.180
<v Speaker 2>you know, send some information in. either through protein expression

0:42:44.200 --> 0:42:49.480
<v Speaker 2>or even through electrical signaling. And that ability to have

0:42:49.540 --> 0:42:52.660
<v Speaker 2>a sort of shared context inside every cell and then

0:42:53.020 --> 0:42:57.320
<v Speaker 2>selectively share things, it's sort of like a modular information

0:42:57.380 --> 0:43:01.739
<v Speaker 2>processing design that is really effective at being able to

0:43:01.860 --> 0:43:05.200
<v Speaker 2>hold a huge amount of information and also operate on

0:43:05.239 --> 0:43:05.840
<v Speaker 2>it effectively.

0:43:05.860 --> 0:43:08.100
<v Speaker 1>Oh, it's a beautiful way to look at cells. And

0:43:08.120 --> 0:43:12.390
<v Speaker 1>what's interesting is, you know, let's say neurons... Most neurons

0:43:12.450 --> 0:43:15.570
<v Speaker 1>are just talking to their neighbors, to 10,000 of their neighbors,

0:43:15.610 --> 0:43:19.040
<v Speaker 1>but they're sending very specific signals. But other neurons do

0:43:19.060 --> 0:43:23.200
<v Speaker 1>what's called volume signaling, where they put out signals that

0:43:23.260 --> 0:43:27.629
<v Speaker 1>reach many thousands or millions of neurons. Have you thought

0:43:27.670 --> 0:43:31.670
<v Speaker 1>about how this maps on to... the way that we

0:43:31.690 --> 0:43:32.370
<v Speaker 1>share information.

0:43:32.390 --> 0:43:36.140
<v Speaker 2>That's a really good idea. I hadn't thought about it,

0:43:36.170 --> 0:43:38.469
<v Speaker 2>but it's like you have like influencer neurons, right?

0:43:38.530 --> 0:43:42.129
<v Speaker 3>They're like, everybody will pay attention to me.

0:43:42.150 --> 0:43:42.230
<v Speaker 2>Okay.

0:43:45.360 --> 0:43:48.180
<v Speaker 1>Coming back to AI alignment. So what should we be

0:43:48.219 --> 0:43:51.120
<v Speaker 1>thinking about as humans entering this AI era?

0:43:51.940 --> 0:43:57.590
<v Speaker 2>One is looking at cooperation in computational terms. If we

0:43:58.070 --> 0:44:03.800
<v Speaker 2>do that, then we can build mechanisms into multi-agent networks,

0:44:03.980 --> 0:44:06.800
<v Speaker 2>thinking about the decision rules the individuals are using, thinking

0:44:06.840 --> 0:44:09.600
<v Speaker 2>about the environment that we set up and the protocols

0:44:09.640 --> 0:44:13.480
<v Speaker 2>for interactions. We can design those in ways that allow

0:44:13.580 --> 0:44:17.440
<v Speaker 2>for cooperation to not be as risky. And we also

0:44:17.800 --> 0:44:22.240
<v Speaker 2>then change the entire environment so that we have selection

0:44:22.280 --> 0:44:26.759
<v Speaker 2>pressures for cooperation. Then that's a way that we can

0:44:26.920 --> 0:44:30.850
<v Speaker 2>take this computational understanding of cooperation and embed it into

0:44:30.890 --> 0:44:32.630
<v Speaker 2>these systems that we're building.

0:44:32.950 --> 0:44:36.770
<v Speaker 1>So setting up the environment for cooperation, what does that

0:44:36.790 --> 0:44:40.110
<v Speaker 1>look like? For example, I heard about this In China,

0:44:40.130 --> 0:44:44.009
<v Speaker 1>the sort of social score that people get, where if

0:44:44.030 --> 0:44:46.590
<v Speaker 1>you're cooperative, you get a high social score. But if

0:44:46.640 --> 0:44:50.280
<v Speaker 1>you do something antisocial, then you get a low score

0:44:50.820 --> 0:44:54.060
<v Speaker 1>on these things. That's setting up a structure from a

0:44:54.180 --> 0:44:56.510
<v Speaker 1>high level. Is that useful?

0:44:56.730 --> 0:45:01.820
<v Speaker 2>Personally, I favor solutions that preserve individual agency. And I

0:45:01.860 --> 0:45:05.540
<v Speaker 2>think we have those, like the walkaway strategy, which is

0:45:05.760 --> 0:45:09.520
<v Speaker 2>literally about individuals being able to decide if they're going

0:45:09.560 --> 0:45:13.250
<v Speaker 2>to stay and continue in an interaction or if they're

0:45:13.290 --> 0:45:16.450
<v Speaker 2>going to leave and seek out a different interaction. Personally,

0:45:16.469 --> 0:45:20.330
<v Speaker 2>I favor solutions that preserve individual agency.

0:45:20.350 --> 0:45:21.790
<v Speaker 3>And I think we have those.

0:45:22.800 --> 0:45:26.859
<v Speaker 2>Like the walkaway strategy, which is literally about individuals being

0:45:26.920 --> 0:45:30.509
<v Speaker 2>able to decide if they're going to stay and continue

0:45:30.730 --> 0:45:33.089
<v Speaker 2>in an interaction or if they're going to leave and

0:45:33.110 --> 0:45:36.670
<v Speaker 2>seek out a different interaction body and not have metabolic disease,

0:45:36.810 --> 0:45:40.379
<v Speaker 2>not be compromised by, you know, parasites and other things

0:45:40.420 --> 0:45:44.009
<v Speaker 2>from the external world. So We really are these bastions

0:45:44.300 --> 0:45:49.359
<v Speaker 2>of cooperation. And within all of those systems that have

0:45:49.400 --> 0:45:55.520
<v Speaker 2>evolved over millions of years to regulate cooperation, there are

0:45:56.120 --> 0:46:00.919
<v Speaker 2>many solutions and secrets that we can unlock. If you

0:46:01.060 --> 0:46:06.380
<v Speaker 2>look at who we are as multicellular organisms, as human

0:46:06.420 --> 0:46:11.650
<v Speaker 2>beings in society, we're literally made of cooperation. The things

0:46:11.710 --> 0:46:15.370
<v Speaker 2>that we have built as humans and the things that

0:46:15.390 --> 0:46:18.290
<v Speaker 2>our evolutionary process and biology have built in terms of

0:46:18.350 --> 0:46:22.430
<v Speaker 2>our bodies are, to me, a testament to the power

0:46:22.489 --> 0:46:23.350
<v Speaker 2>of cooperation.

0:46:24.110 --> 0:46:25.390
<v Speaker 3>And the fact.

0:46:25.210 --> 0:46:30.339
<v Speaker 2>That on some level, this power of cooperation combined with

0:46:30.380 --> 0:46:38.470
<v Speaker 2>the right evolutionary pressures is is sufficient to overcome conflict, exploitation, hijacking,

0:46:38.930 --> 0:46:43.380
<v Speaker 2>and continue to persist and exist. And so when I

0:46:43.410 --> 0:46:47.239
<v Speaker 2>look at our evolutionary history, what I see is a

0:46:47.280 --> 0:46:51.340
<v Speaker 2>very hopeful story. And I also see lots of guidance

0:46:51.760 --> 0:46:56.160
<v Speaker 2>that we can look to for how to build technological

0:46:56.450 --> 0:46:59.730
<v Speaker 2>and human systems as they scale. And to learn from

0:46:59.750 --> 0:47:05.210
<v Speaker 2>the many errors and trials that evolution has gone through

0:47:05.410 --> 0:47:11.050
<v Speaker 2>to build large-scale cooperative systems as we build cooperative systems

0:47:11.310 --> 0:47:15.070
<v Speaker 2>now that involve humans and technological systems together.

0:47:20.469 --> 0:47:23.550
<v Speaker 1>That was my interview with Athena Actippus. So take a

0:47:23.590 --> 0:47:25.870
<v Speaker 1>second to consider all the things that had to go

0:47:25.969 --> 0:47:29.810
<v Speaker 1>right for you to hear this podcast. Your heart had

0:47:29.850 --> 0:47:33.990
<v Speaker 1>to move oxygenated blood through a giant subway system of vessels.

0:47:34.530 --> 0:47:37.200
<v Speaker 1>The cells that make up your lungs had to exchange

0:47:37.280 --> 0:47:41.040
<v Speaker 1>gases with the air. Cells in your inner ear... converted

0:47:41.080 --> 0:47:44.780
<v Speaker 1>vibrations into electrical activity in your brain, and then networks

0:47:44.840 --> 0:47:49.910
<v Speaker 1>of cells throughout your brain separated my voice from background

0:47:49.969 --> 0:47:54.910
<v Speaker 1>sound and recognized words, and held earlier phrases in memory

0:47:55.410 --> 0:48:00.260
<v Speaker 1>and assembled meaning across time this way. Your whole understanding

0:48:00.300 --> 0:48:06.779
<v Speaker 1>of this podcast emerged from cooperation among trillions of cells. Now,

0:48:06.840 --> 0:48:11.340
<v Speaker 1>Athena's work is to understand that cooperation computationally. The idea

0:48:11.420 --> 0:48:16.000
<v Speaker 1>is that every durable cooperative system has to solve the

0:48:16.080 --> 0:48:20.560
<v Speaker 1>same kinds of problems. Which parts perform which jobs? How

0:48:20.580 --> 0:48:24.940
<v Speaker 1>do you distribute resources? What information gets shared? How does

0:48:24.980 --> 0:48:29.839
<v Speaker 1>the system recognize damage or exploitation or a goal that's

0:48:29.920 --> 0:48:33.700
<v Speaker 1>gone sideways? In a body, you have deep layers of

0:48:33.760 --> 0:48:37.770
<v Speaker 1>decision rules. The cells get signals about things surrounding them,

0:48:38.170 --> 0:48:41.630
<v Speaker 1>and then molecular circuits shape what happens next. So the

0:48:41.650 --> 0:48:45.380
<v Speaker 1>cell might keep on dividing or stop and fix itself

0:48:45.440 --> 0:48:49.719
<v Speaker 1>or change its behavior. or sometimes it'll decide to commit suicide.

0:48:49.760 --> 0:48:52.739
<v Speaker 1>This is known as apoptosis. Now, that's amazing to have

0:48:52.760 --> 0:48:57.200
<v Speaker 1>a giant cooperative system that works, but it's never without

0:48:57.400 --> 0:49:02.580
<v Speaker 1>trouble lurking. And so Athena described the gene TP53 as

0:49:02.660 --> 0:49:06.779
<v Speaker 1>a cheater detection of the genome. And that gives us

0:49:06.820 --> 0:49:09.960
<v Speaker 1>an important way to look at cancer, like a cheater

0:49:10.340 --> 0:49:13.000
<v Speaker 1>in a cooperative system. In other words, you've got all

0:49:13.040 --> 0:49:17.370
<v Speaker 1>these capacities that allow normal cells to cooperate, But that

0:49:17.450 --> 0:49:21.450
<v Speaker 1>creates openings for something like a tumor to exploit the

0:49:21.489 --> 0:49:26.569
<v Speaker 1>shared environment. The only way multicellular life lasts is if

0:49:26.650 --> 0:49:30.069
<v Speaker 1>it finds holes and plugs them up, as we see

0:49:30.130 --> 0:49:33.259
<v Speaker 1>with TP53. Now, the lessons grow a little more complicated

0:49:33.300 --> 0:49:35.660
<v Speaker 1>when we move from cells to people, and this is

0:49:35.719 --> 0:49:40.390
<v Speaker 1>because the cells of the body essentially share a common fate.

0:49:41.050 --> 0:49:45.410
<v Speaker 1>But human beings in society have overlapping interests and private

0:49:45.450 --> 0:49:49.630
<v Speaker 1>goals and changing alliances. Athena points out that humans are

0:49:50.030 --> 0:49:56.020
<v Speaker 1>facultative cooperators, meaning our behavior shifts with context. If you

0:49:56.060 --> 0:49:59.920
<v Speaker 1>have an environment of trust and positive interdependence, that can

0:49:59.960 --> 0:50:02.660
<v Speaker 1>bring out generosity. But if you have an environment of

0:50:03.120 --> 0:50:08.850
<v Speaker 1>scarcity or threat or exploitation, that can recruit suspicion and competition.

0:50:09.600 --> 0:50:12.089
<v Speaker 1>But in any case, you can still approach the problem

0:50:12.390 --> 0:50:16.370
<v Speaker 1>with a computational lens. You can ask what happens when

0:50:16.450 --> 0:50:21.000
<v Speaker 1>particular rules interact across elements. Thousands or millions of agents.

0:50:21.410 --> 0:50:24.290
<v Speaker 1>What information is available? How quickly does it travel? Who

0:50:24.370 --> 0:50:28.570
<v Speaker 1>can alter it? Which actions receive rewards? So in theory,

0:50:28.650 --> 0:50:33.160
<v Speaker 1>political principles can become hypotheses about how things will behave

0:50:33.489 --> 0:50:36.420
<v Speaker 1>inside a certain rule set. For example, we can try

0:50:36.460 --> 0:50:41.379
<v Speaker 1>to understand how a system can preserve cooperativity while detecting

0:50:41.719 --> 0:50:45.160
<v Speaker 1>when another process has seized the controls. And this is

0:50:45.219 --> 0:50:50.089
<v Speaker 1>a massively useful gift of biology to our future, both

0:50:50.140 --> 0:50:54.790
<v Speaker 1>politically and technologically, because evolution has run versions of these

0:50:54.850 --> 0:51:00.810
<v Speaker 1>experiments for immense spans of time. Bodies have distributed monitoring

0:51:00.870 --> 0:51:06.020
<v Speaker 1>and redundant defenses and controlled growth and selective information sharing

0:51:06.040 --> 0:51:09.460
<v Speaker 1>and repair and boundaries. And we can learn from all

0:51:09.480 --> 0:51:13.950
<v Speaker 1>of these. And we Evolution gave us tumors and autoimmune

0:51:13.989 --> 0:51:17.660
<v Speaker 1>disease and parasites, and these give us ways to study

0:51:18.040 --> 0:51:21.359
<v Speaker 1>how cooperative control can break and what can be done

0:51:21.420 --> 0:51:24.940
<v Speaker 1>about it. Perfect alignment is probably always going to be

0:51:25.000 --> 0:51:28.930
<v Speaker 1>out of reach when it comes to politics. Citizens share

0:51:28.989 --> 0:51:32.150
<v Speaker 1>a country, but they disagree about its direction. And that's okay.

0:51:32.590 --> 0:51:35.610
<v Speaker 1>And humans and AI agents are probably going to pursue

0:51:35.690 --> 0:51:40.610
<v Speaker 1>common goals inside networks of incentives that no single designer

0:51:40.650 --> 0:51:46.330
<v Speaker 1>can possibly control. But cooperation is nonetheless something that we

0:51:46.370 --> 0:51:51.069
<v Speaker 1>can pursue by understanding shared interests and acknowledging conflicts and

0:51:51.140 --> 0:51:55.630
<v Speaker 1>adapting the rules to as conditions change. We are here

0:51:55.710 --> 0:52:03.130
<v Speaker 1>because across evolutionary time, cooperation repeatedly survived encounters with cheating

0:52:03.469 --> 0:52:08.069
<v Speaker 1>and exploitation and hijacking. Every heartbeat that you have is

0:52:08.290 --> 0:52:12.610
<v Speaker 1>evidence that specialized parts can figure out ways to coordinate.

0:52:13.230 --> 0:52:16.930
<v Speaker 1>Every thought that crosses your mind is an event produced

0:52:16.989 --> 0:52:20.609
<v Speaker 1>by a network of billions of cells. We are made

0:52:20.640 --> 0:52:25.480
<v Speaker 1>of cooperation. The question for the systems that we build

0:52:25.600 --> 0:52:28.799
<v Speaker 1>next is just whether we can be smart enough to

0:52:28.840 --> 0:52:34.950
<v Speaker 1>design the guardrails that cooperation needs to survive. Go to

0:52:35.030 --> 0:52:39.170
<v Speaker 1>Eagleman.com slash podcast for more information and to find further reading.

0:52:39.350 --> 0:52:41.830
<v Speaker 1>Join the weekly discussions on my sub stack and check

0:52:41.890 --> 0:52:45.250
<v Speaker 1>out and subscribe to Inner Cosmos on YouTube for videos

0:52:45.270 --> 0:52:48.029
<v Speaker 1>of each episode and to leave comments. Until next time,

0:52:48.050 --> 0:52:56.509
<v Speaker 1>I'm David Eagleman and this is Inner Cosmos. you