WEBVTT - Punish the Machine, Part 1

0:00:00.160 --> 0:00:05.240
<v Speaker 1>My welcome to Stuff to Blow Your Mind, the production

0:00:05.240 --> 0:00:13.960
<v Speaker 1>of My Heart Radio. Hey you welcome to Stuff to

0:00:13.960 --> 0:00:16.680
<v Speaker 1>Blow your Mind. My name is Robert Lamb and I'm

0:00:16.760 --> 0:00:19.640
<v Speaker 1>Joe McCormick. And right before we started recording today, we

0:00:19.640 --> 0:00:22.480
<v Speaker 1>were just talking about that iconic scene and returned to

0:00:22.560 --> 0:00:26.680
<v Speaker 1>the Jedi where the droids are sent to the droid

0:00:26.840 --> 0:00:28.920
<v Speaker 1>torture Chamber. Do you remember there? I guess it's not

0:00:29.000 --> 0:00:31.360
<v Speaker 1>just a droid torture chamber. It's sort of like the uh,

0:00:32.440 --> 0:00:36.080
<v Speaker 1>the droid onboarding center right where the you know, R

0:00:36.159 --> 0:00:38.320
<v Speaker 1>two D two and C three PO have been given

0:00:38.320 --> 0:00:40.519
<v Speaker 1>as gifts to job of the Hut and they go

0:00:40.640 --> 0:00:43.400
<v Speaker 1>meet their new like droid boss and he's like, yeah,

0:00:43.479 --> 0:00:45.800
<v Speaker 1>you're a feisty little one, and he's signing them in,

0:00:45.920 --> 0:00:48.760
<v Speaker 1>but uh, he sees that R two D two is

0:00:48.800 --> 0:00:51.880
<v Speaker 1>a is a bad robot who needs a discipline, and

0:00:52.120 --> 0:00:54.560
<v Speaker 1>R two D two is confronted with these images of

0:00:54.640 --> 0:00:58.880
<v Speaker 1>robots being punished with various corporal punishments, like one is

0:00:58.920 --> 0:01:02.000
<v Speaker 1>getting stretched on a robot rack and another one is

0:01:02.040 --> 0:01:05.720
<v Speaker 1>getting its feet burned. Yes, this is a this is

0:01:05.760 --> 0:01:09.399
<v Speaker 1>a great scene one that that that definitely burns its

0:01:09.400 --> 0:01:11.840
<v Speaker 1>way into your your brain as a as a young viewer,

0:01:11.920 --> 0:01:13.600
<v Speaker 1>and maybe you don't think about it that much for

0:01:13.640 --> 0:01:16.160
<v Speaker 1>a long time, but uh, it's it's still in there.

0:01:16.240 --> 0:01:19.720
<v Speaker 1>It's it takes place in the bowels of Joba's palace

0:01:20.080 --> 0:01:23.960
<v Speaker 1>on tattooing, and it's um, yeah, it's like droid intake

0:01:24.040 --> 0:01:27.520
<v Speaker 1>but also droid corrections. It's there are there are a

0:01:27.600 --> 0:01:30.440
<v Speaker 1>number of different departments that I think are converging here

0:01:31.160 --> 0:01:34.920
<v Speaker 1>and and it it ultimately kind of raises some interesting

0:01:35.000 --> 0:01:40.240
<v Speaker 1>questions about um about ethics and punishment and in crime,

0:01:40.319 --> 0:01:44.200
<v Speaker 1>and certainly as it relates to two robots. Uh. Of course,

0:01:44.240 --> 0:01:46.440
<v Speaker 1>one thing, this important distress here is like none of

0:01:46.480 --> 0:01:49.560
<v Speaker 1>this was really intended in these scenes. This was about

0:01:49.840 --> 0:01:53.880
<v Speaker 1>having droids doing things that humans would be doing to

0:01:53.920 --> 0:01:57.480
<v Speaker 1>each other in other pieces of cinema, certainly things like

0:01:57.520 --> 0:02:01.040
<v Speaker 1>old pirate movies or old and bad movies or what

0:02:01.120 --> 0:02:03.080
<v Speaker 1>have you. I mean, that's kind of that's kind of

0:02:03.080 --> 0:02:06.600
<v Speaker 1>Star Wars in a nutshell, right. Uh. This whole portion

0:02:06.680 --> 0:02:09.680
<v Speaker 1>of Return of the Jedi is essentially a big pirate movie,

0:02:09.680 --> 0:02:13.679
<v Speaker 1>a big swashbuckler set in an alien location. Oh yeah,

0:02:13.720 --> 0:02:16.399
<v Speaker 1>the job of the Hut is a pirate captain yeah,

0:02:16.440 --> 0:02:20.320
<v Speaker 1>but something interesting occurs when you replace the humans in

0:02:20.360 --> 0:02:25.320
<v Speaker 1>these these tropic scenes with machines. Uh, and then you

0:02:25.360 --> 0:02:27.119
<v Speaker 1>think about it, you know, you think about, well, why

0:02:27.200 --> 0:02:29.680
<v Speaker 1>is that robot torturing the other? Like as if it

0:02:29.720 --> 0:02:33.040
<v Speaker 1>makes perfect sense if it's humans doing it. But then

0:02:33.400 --> 0:02:35.919
<v Speaker 1>the things that we create in our image when they're

0:02:35.919 --> 0:02:38.720
<v Speaker 1>doing it, Suddenly we start seeing the flaws in our reasoning.

0:02:38.760 --> 0:02:41.959
<v Speaker 1>Suddenly we start questioning, well, how does this whole system

0:02:42.040 --> 0:02:46.560
<v Speaker 1>supposed to work? And maybe this whole system doesn't work. Well. Yeah,

0:02:46.560 --> 0:02:49.280
<v Speaker 1>there are multiple levels of absurdities in the scene. One

0:02:49.360 --> 0:02:52.000
<v Speaker 1>is the idea that this robot is just sort of

0:02:52.000 --> 0:02:54.120
<v Speaker 1>like coolly telling are two that he is going to

0:02:54.200 --> 0:02:57.120
<v Speaker 1>learn some discipline. But then the image that accompanies that

0:02:57.280 --> 0:03:00.400
<v Speaker 1>is like, clearly just like extreme robot tor sure, like

0:03:00.440 --> 0:03:03.240
<v Speaker 1>it's something way beyond what would have to do with

0:03:03.520 --> 0:03:06.160
<v Speaker 1>with discipline in the real world. But then the other

0:03:06.240 --> 0:03:08.959
<v Speaker 1>level of absurdity is that it's robots in the scene,

0:03:09.400 --> 0:03:12.280
<v Speaker 1>but coming off of the issue of just like barbaric

0:03:12.360 --> 0:03:16.840
<v Speaker 1>pirate torture directly and more to the broader question of

0:03:17.040 --> 0:03:21.520
<v Speaker 1>robots and discipline and punishment. Uh, this is something that

0:03:21.560 --> 0:03:24.200
<v Speaker 1>we actually wanted to talk about today because the issue

0:03:24.280 --> 0:03:28.080
<v Speaker 1>of robot moral and legal agency is something I've been

0:03:28.160 --> 0:03:30.680
<v Speaker 1>interested in for a long time. I've talked about it,

0:03:30.680 --> 0:03:32.320
<v Speaker 1>It's come up on the show in the past and

0:03:32.440 --> 0:03:36.040
<v Speaker 1>in briefer ways, um And today I wanted to come

0:03:36.040 --> 0:03:38.760
<v Speaker 1>back and devote a full episode to the subject. I

0:03:38.760 --> 0:03:40.440
<v Speaker 1>guess actually we're gonna be talking about this for a

0:03:40.480 --> 0:03:45.680
<v Speaker 1>couple of episodes now. The question of as machines AI

0:03:45.800 --> 0:03:50.040
<v Speaker 1>robots become more independent and act more like agents, more

0:03:50.120 --> 0:03:54.360
<v Speaker 1>like humans do, how are we to understand their moral

0:03:54.520 --> 0:03:58.480
<v Speaker 1>and legal culpability when they do something that harms people?

0:03:59.000 --> 0:04:03.120
<v Speaker 1>And is there such a thing as robot punishment robot discipline?

0:04:03.160 --> 0:04:06.520
<v Speaker 1>Do these concepts reflect anything that's achievable in the real

0:04:06.560 --> 0:04:09.480
<v Speaker 1>world and practical and if so, how would any of

0:04:09.520 --> 0:04:11.760
<v Speaker 1>this work? Yeah? I think one of the most interesting

0:04:11.800 --> 0:04:14.480
<v Speaker 1>things about this topic is that it does force us

0:04:14.480 --> 0:04:17.520
<v Speaker 1>to force a face off between what robots and AI

0:04:17.680 --> 0:04:21.599
<v Speaker 1>actually are or will be, and how we think about them, indeed,

0:04:21.640 --> 0:04:26.560
<v Speaker 1>how we anthropomorphize them. Um And perhaps it might be

0:04:26.600 --> 0:04:28.760
<v Speaker 1>helpful to take a step back and think about something

0:04:29.240 --> 0:04:32.240
<v Speaker 1>far less advanced as a robot and think something I

0:04:32.320 --> 0:04:36.799
<v Speaker 1>think about something more like a hammer Okay, so everyone's

0:04:36.839 --> 0:04:38.960
<v Speaker 1>heard the old adage that it's a poor carpenter who

0:04:38.960 --> 0:04:41.960
<v Speaker 1>blames their tools, right, But of course we do this

0:04:42.040 --> 0:04:45.800
<v Speaker 1>all the time. Uh, the hammer slips, it hits our fingers,

0:04:46.160 --> 0:04:48.240
<v Speaker 1>and we may, at least in the heat of the moment,

0:04:48.600 --> 0:04:51.839
<v Speaker 1>blame the hammer for the failure. Now, we may get

0:04:51.839 --> 0:04:54.440
<v Speaker 1>over this quickly, but then again, we may decide that

0:04:54.800 --> 0:04:57.480
<v Speaker 1>the hammer truly is at fault and it should be

0:04:57.600 --> 0:05:00.840
<v Speaker 1>used less. We might also take this idea to a

0:05:00.920 --> 0:05:03.440
<v Speaker 1>number of different extremes. We might decide that the hammer

0:05:03.760 --> 0:05:06.560
<v Speaker 1>is not merely at fault but faulty, and then we're

0:05:06.680 --> 0:05:09.720
<v Speaker 1>entitled to at least a refund for its purchase. Or

0:05:09.760 --> 0:05:12.680
<v Speaker 1>we might decide that the hammer needs to actually be punished.

0:05:13.040 --> 0:05:15.600
<v Speaker 1>And and this, of course is ridiculous. And yet the

0:05:15.680 --> 0:05:18.320
<v Speaker 1>idea of punishing the hammer by say, putting it in

0:05:18.360 --> 0:05:21.479
<v Speaker 1>the corner, or perhaps you have an old toolbox of

0:05:21.560 --> 0:05:25.479
<v Speaker 1>shame that's just for the misbehaving tools, or maybe it's

0:05:25.600 --> 0:05:27.880
<v Speaker 1>it's less thought out and you just throw the hammer

0:05:28.360 --> 0:05:31.200
<v Speaker 1>across the yard as punishment for what it has done

0:05:31.240 --> 0:05:34.320
<v Speaker 1>to you. Um. Again, these are ridiculous things to do,

0:05:34.920 --> 0:05:36.920
<v Speaker 1>but the idea of doing them is not that far

0:05:36.960 --> 0:05:38.960
<v Speaker 1>and from us. Um, those of you listening, you may

0:05:38.960 --> 0:05:41.240
<v Speaker 1>have engaged in this sort of thing as well. You

0:05:41.320 --> 0:05:44.360
<v Speaker 1>might also simply throw the tool away and otherwise perfectly

0:05:44.400 --> 0:05:47.479
<v Speaker 1>good tool. Um. I know that I did this once

0:05:47.480 --> 0:05:49.880
<v Speaker 1>with a knife sharpening gadget that caused me to cut

0:05:49.880 --> 0:05:53.160
<v Speaker 1>my finger. My and my like reaction was this thing

0:05:53.240 --> 0:05:56.559
<v Speaker 1>has now injured me, it has drawn my blood. Uh

0:05:56.600 --> 0:05:58.080
<v Speaker 1>I am, I'm getting rid of it. It goes in

0:05:58.120 --> 0:06:01.440
<v Speaker 1>the trash. It bore malice again to me. Yeah, or

0:06:01.480 --> 0:06:04.479
<v Speaker 1>you know, I you know ultimately it's I mean, I

0:06:04.520 --> 0:06:06.960
<v Speaker 1>mean you you get into arguments about different tools, like

0:06:07.120 --> 0:06:09.520
<v Speaker 1>is this a dangerous tool? And in that case that

0:06:09.600 --> 0:06:12.880
<v Speaker 1>was my reasoning. It's like this tool is dangerous. It's

0:06:12.920 --> 0:06:15.360
<v Speaker 1>not enabling me to do what I want to do

0:06:15.440 --> 0:06:19.000
<v Speaker 1>without drawing blood, So it goes in the trash. Um.

0:06:19.040 --> 0:06:20.760
<v Speaker 1>But then there have been other cases where like I

0:06:20.800 --> 0:06:24.599
<v Speaker 1>had a mandolin for slicing up carrots, and um, I

0:06:24.680 --> 0:06:27.440
<v Speaker 1>like nicked my finger on it. But oh I I

0:06:27.560 --> 0:06:30.120
<v Speaker 1>nicked my fingers. Things are brutal, they can be. But

0:06:30.640 --> 0:06:33.920
<v Speaker 1>I nicked my finger, not using it but going into

0:06:33.960 --> 0:06:36.279
<v Speaker 1>the drawer for something else. So I punished it by

0:06:36.320 --> 0:06:39.080
<v Speaker 1>putting at the very bottom of the drawer, but I

0:06:39.120 --> 0:06:41.480
<v Speaker 1>didn't throw it away. Uh huh. So I think if

0:06:41.480 --> 0:06:43.640
<v Speaker 1>we all think, think back, you know, we have examples

0:06:43.640 --> 0:06:45.560
<v Speaker 1>of this sort of thing from our from our life.

0:06:45.880 --> 0:06:48.040
<v Speaker 1>Well sure, I mean, I'm going to talk in this

0:06:48.080 --> 0:06:50.960
<v Speaker 1>episode about some of the ways that we mindlessly apply

0:06:51.279 --> 0:06:55.280
<v Speaker 1>social rules to robots. But yeah, I think what you're

0:06:55.279 --> 0:06:58.120
<v Speaker 1>illustrating here is that you don't even have to get

0:06:58.160 --> 0:07:01.280
<v Speaker 1>to the robot agency stage of four people start doing that.

0:07:01.320 --> 0:07:05.559
<v Speaker 1>I mean people mindlessly, to a lesser extent, mindlessly apply

0:07:05.800 --> 0:07:09.800
<v Speaker 1>social rules and rules derived for managing human relationships to

0:07:10.200 --> 0:07:13.920
<v Speaker 1>inanimate objects with no moving parts. Yeah. Yeah, you don't

0:07:13.920 --> 0:07:16.760
<v Speaker 1>even have to get into a room ba or anything,

0:07:16.880 --> 0:07:18.280
<v Speaker 1>or you know, you can deal with the hammer, the

0:07:18.320 --> 0:07:21.760
<v Speaker 1>can opener. But we also you know, it's it's also

0:07:21.880 --> 0:07:25.120
<v Speaker 1>a sad fact that many pet owners will punish an

0:07:25.160 --> 0:07:29.040
<v Speaker 1>animal for a transgression transgression, but scientific evidence shows that

0:07:29.120 --> 0:07:32.840
<v Speaker 1>this tends to not actually work, at least in most

0:07:32.880 --> 0:07:37.000
<v Speaker 1>of the circumstances that it's used. Um So, you know,

0:07:37.080 --> 0:07:41.760
<v Speaker 1>even it's not merely with with tools and inanimate objects,

0:07:41.760 --> 0:07:44.760
<v Speaker 1>but even non human entities were liable to engage in

0:07:44.800 --> 0:07:49.320
<v Speaker 1>this kind of discipline based thinking. Now, most of the

0:07:49.320 --> 0:07:52.600
<v Speaker 1>studies I think that necessarily relate to dogs, if I

0:07:52.640 --> 0:07:54.480
<v Speaker 1>remember correctly, And there's a lot going on here that

0:07:54.520 --> 0:07:57.520
<v Speaker 1>doesn't relate directly to inanimate objects and robots, but it

0:07:57.560 --> 0:07:59.800
<v Speaker 1>illustrates how we tend to approach the punishment of other

0:07:59.840 --> 0:08:03.120
<v Speaker 1>agents and perceived agents. Well, yeah, there's a disconnect, and

0:08:03.120 --> 0:08:04.920
<v Speaker 1>this will be highlighted in one of the papers we're

0:08:04.920 --> 0:08:07.000
<v Speaker 1>going to talk about in in this pair of episodes,

0:08:07.040 --> 0:08:11.480
<v Speaker 1>But there's a disconnect in that punishment is often logically

0:08:12.160 --> 0:08:15.880
<v Speaker 1>characterized as serving one type of purpose, but then is

0:08:15.960 --> 0:08:19.040
<v Speaker 1>applied more like it serves another type of purpose. So

0:08:19.080 --> 0:08:22.760
<v Speaker 1>like it is logically explained as say a deterrent, right,

0:08:22.800 --> 0:08:25.560
<v Speaker 1>I mean, if you if you're talking about uh, legal

0:08:25.600 --> 0:08:28.000
<v Speaker 1>theories of punishment, one of the main things that people

0:08:28.000 --> 0:08:30.760
<v Speaker 1>come up with is say, well, the remedy provided by

0:08:30.800 --> 0:08:33.640
<v Speaker 1>the law is in order to punish the person who

0:08:33.640 --> 0:08:36.240
<v Speaker 1>did the bad thing in order to send a message

0:08:36.280 --> 0:08:39.280
<v Speaker 1>that people should not do this bad thing and thus

0:08:39.320 --> 0:08:42.600
<v Speaker 1>maybe discourage other people from doing something similar in the future,

0:08:42.679 --> 0:08:45.520
<v Speaker 1>or discourage the same person from doing it again. And

0:08:45.600 --> 0:08:48.120
<v Speaker 1>if it were to actually serve that purpose, it's debatable

0:08:48.120 --> 0:08:50.520
<v Speaker 1>in what cases it does actually serve that purpose. Maybe

0:08:50.600 --> 0:08:53.120
<v Speaker 1>sometimes it does, but that is a you know, you

0:08:53.120 --> 0:08:56.920
<v Speaker 1>could argue that's a rational, logical thing that prevents harm.

0:08:57.240 --> 0:09:00.200
<v Speaker 1>But the way punishment is actually often inflict it in

0:09:00.240 --> 0:09:03.880
<v Speaker 1>the real world seems to be more consistent with judgments

0:09:03.920 --> 0:09:08.199
<v Speaker 1>based on like emotional satisfaction of the idea of having

0:09:08.240 --> 0:09:11.720
<v Speaker 1>been wronged. Yeah. Yeah. And then also we get into

0:09:11.720 --> 0:09:15.520
<v Speaker 1>this area where, uh, we have a couple of different

0:09:15.559 --> 0:09:20.199
<v Speaker 1>factors encouraging traditions of discipline. Um, particularly if we look

0:09:20.200 --> 0:09:25.480
<v Speaker 1>at a parenthood, which there's some crossover between discipline and

0:09:25.480 --> 0:09:28.600
<v Speaker 1>parenthood and discipline and the criminal justice system. But uh,

0:09:28.640 --> 0:09:31.240
<v Speaker 1>you know, uh, not everything is going to line up

0:09:31.280 --> 0:09:34.680
<v Speaker 1>one to one here. But um, on the childhood example,

0:09:34.679 --> 0:09:37.120
<v Speaker 1>it's been argued that parents use punishment first of all,

0:09:37.160 --> 0:09:40.680
<v Speaker 1>because it's an emotional response out of anger and anger

0:09:40.840 --> 0:09:43.920
<v Speaker 1>that may be mismanaged. But then on top of this,

0:09:44.720 --> 0:09:48.760
<v Speaker 1>it's you know, something that's culturally passed down and punishment

0:09:48.840 --> 0:09:52.000
<v Speaker 1>may seem to work. I was reading about this in

0:09:52.040 --> 0:09:56.079
<v Speaker 1>a Psychology Today article by Michael Carson, pH d j D.

0:09:56.760 --> 0:09:59.240
<v Speaker 1>And uh, this is what they said, quote, because the

0:09:59.320 --> 0:10:01.520
<v Speaker 1>child is an bited in your presence it's easy to

0:10:01.559 --> 0:10:05.920
<v Speaker 1>think they would be inhibited in your absence. Punishment produces politeness,

0:10:05.960 --> 0:10:10.880
<v Speaker 1>not morality. Thus the inhibited, obedient child inadvertently reinforces the

0:10:10.920 --> 0:10:14.960
<v Speaker 1>parents punitive behavior by acting obedient. For the sorts of

0:10:15.000 --> 0:10:18.880
<v Speaker 1>parents who find obedient children reinforcing, Yeah, that raises an

0:10:18.880 --> 0:10:21.360
<v Speaker 1>interesting question. I mean, I've been mainly thinking for this

0:10:21.400 --> 0:10:25.720
<v Speaker 1>episode about about legal punishments, but like when it comes

0:10:25.760 --> 0:10:28.320
<v Speaker 1>down to parenting, that's a very different kind of thing

0:10:28.400 --> 0:10:32.520
<v Speaker 1>because both parenting in the legal system involved punishment, but

0:10:32.800 --> 0:10:36.440
<v Speaker 1>parenting is not subject to a legal system, right, So

0:10:36.520 --> 0:10:40.040
<v Speaker 1>there's no there is no systematized way by which justice

0:10:40.120 --> 0:10:42.760
<v Speaker 1>is administered from a parent. It's just just I mean,

0:10:42.800 --> 0:10:44.120
<v Speaker 1>I think a lot of times it's just sort of

0:10:44.160 --> 0:10:46.640
<v Speaker 1>like whatever the parent can manage to do in the moment,

0:10:46.720 --> 0:10:49.880
<v Speaker 1>because like the kids driving them crazy or something. Yeah,

0:10:49.960 --> 0:10:52.559
<v Speaker 1>usually the child can't take it to a higher court.

0:10:53.320 --> 0:10:55.800
<v Speaker 1>But I mean, I think you're absolutely right that whether

0:10:55.880 --> 0:10:58.840
<v Speaker 1>you're talking about discipline administered by a parent or the

0:10:59.000 --> 0:11:02.480
<v Speaker 1>justice system is whole, I'd say that both are probably

0:11:02.520 --> 0:11:07.520
<v Speaker 1>based more on tradition and philosophy and less on a

0:11:07.600 --> 0:11:11.920
<v Speaker 1>scientifically rigorous study of the most efficient ways to reduce harm.

0:11:11.960 --> 0:11:14.360
<v Speaker 1>And one of the interesting things about thinking about how

0:11:14.520 --> 0:11:18.960
<v Speaker 1>law could potentially be applied to harm caused by autonomous

0:11:19.000 --> 0:11:22.599
<v Speaker 1>machines is that it may help give us some insights

0:11:22.640 --> 0:11:26.000
<v Speaker 1>on ways that the justice system as it exists and

0:11:26.120 --> 0:11:30.439
<v Speaker 1>is applied to humans today tends to behave irrationally already,

0:11:30.559 --> 0:11:34.080
<v Speaker 1>like with respect to humans. Yeah, and again, this is

0:11:34.120 --> 0:11:37.079
<v Speaker 1>what's so interesting about this this paper, I mean, well,

0:11:37.120 --> 0:11:39.720
<v Speaker 1>the papers that we're going to discuss this topic in general, though,

0:11:40.160 --> 0:11:43.840
<v Speaker 1>is if you start, you start comparing machine possibilities to

0:11:43.920 --> 0:11:47.880
<v Speaker 1>human possibilities, and it's on one level of thought experiment

0:11:47.960 --> 0:11:50.760
<v Speaker 1>in how you would hold machines responsible, but then it

0:11:50.800 --> 0:11:53.680
<v Speaker 1>makes you rethink the way humans are held responsible. You know.

0:11:54.080 --> 0:11:56.240
<v Speaker 1>It's like, um, you know, I think you have a

0:11:56.240 --> 0:12:00.880
<v Speaker 1>pretty square away. Like if if an adult sells a

0:12:00.880 --> 0:12:03.800
<v Speaker 1>pack of cigarettes to someone who's underage, right, but then

0:12:03.840 --> 0:12:06.680
<v Speaker 1>one of a machine does the same thing, how do

0:12:06.720 --> 0:12:08.319
<v Speaker 1>you treat the machine? Do you treat a machine like

0:12:08.360 --> 0:12:10.480
<v Speaker 1>an adult? And then in trying to figure out how

0:12:10.520 --> 0:12:13.079
<v Speaker 1>to treat this machine, does it make you rethink how

0:12:13.120 --> 0:12:15.720
<v Speaker 1>you should be treating the adult who engaged in this behavior?

0:12:15.800 --> 0:12:17.720
<v Speaker 1>I don't know. Yeah, And I think a lot of

0:12:17.720 --> 0:12:20.400
<v Speaker 1>that will come down to our understanding of what the

0:12:20.480 --> 0:12:24.040
<v Speaker 1>machine is capable of, like what kind of constraints it has,

0:12:24.440 --> 0:12:27.160
<v Speaker 1>what type of what level of autonomy it seems to

0:12:27.200 --> 0:12:30.160
<v Speaker 1>be operating at. I mean, again, Weirdly, even when people

0:12:30.200 --> 0:12:32.800
<v Speaker 1>set out to define clear rules for what makes a

0:12:32.840 --> 0:12:35.679
<v Speaker 1>machine culpable, there there's still going to be a lot

0:12:35.720 --> 0:12:39.880
<v Speaker 1>of subjectivity in it. I'm looking at like legal definitions

0:12:39.920 --> 0:12:43.760
<v Speaker 1>of what constitutes a robot versus just a machine, and

0:12:43.920 --> 0:12:47.120
<v Speaker 1>some of these definitions involved things like, well, a robot

0:12:47.360 --> 0:12:51.080
<v Speaker 1>feels like a social agent. So there's still, like, you know,

0:12:51.120 --> 0:12:53.960
<v Speaker 1>an element of subjectivity. But I think that's correct in

0:12:54.040 --> 0:12:56.480
<v Speaker 1>how we actually apply the term most of the time, right,

0:12:56.520 --> 0:13:00.240
<v Speaker 1>Like something is like a gut feeling about how this

0:13:00.280 --> 0:13:03.720
<v Speaker 1>machine is behaving in your world. Is it acting more

0:13:03.840 --> 0:13:06.800
<v Speaker 1>like a fixed, you know, brainless machine or is it

0:13:06.840 --> 0:13:11.280
<v Speaker 1>acting a little bit more like a person. So while

0:13:12.000 --> 0:13:14.240
<v Speaker 1>it would be one thing if if it were basically

0:13:14.240 --> 0:13:17.280
<v Speaker 1>a cigarette vending machine that was selling to children, but

0:13:17.320 --> 0:13:19.959
<v Speaker 1>if it were a machine that went door to door

0:13:20.440 --> 0:13:23.480
<v Speaker 1>and and rang the doorbell and then asked for the

0:13:23.559 --> 0:13:25.679
<v Speaker 1>children so they could sell them cigarettes, that would be

0:13:25.679 --> 0:13:27.559
<v Speaker 1>a different matter. I mean, yeah, I mean, I think

0:13:27.600 --> 0:13:31.080
<v Speaker 1>that would require different types of remedies probably, Yeah, I

0:13:31.120 --> 0:13:32.760
<v Speaker 1>mean I think a lot of people would probably look

0:13:32.760 --> 0:13:35.480
<v Speaker 1>at the cigarette vending machine and say, where was the

0:13:35.559 --> 0:13:38.200
<v Speaker 1>vending machine placed? Why was it in a place that

0:13:38.320 --> 0:13:41.600
<v Speaker 1>children could have access to it? Rather than attacking the

0:13:41.600 --> 0:13:44.240
<v Speaker 1>fundamentals of the machine itself. If it's going door to

0:13:44.280 --> 0:13:46.600
<v Speaker 1>door and giving cigarettes to kids, yeah, then people are

0:13:46.600 --> 0:13:50.160
<v Speaker 1>probably going to attack the fundamentals and the moral character

0:13:50.559 --> 0:13:53.760
<v Speaker 1>of the robot. Right, you have not attacked the robot itself.

0:13:53.800 --> 0:14:00.319
<v Speaker 1>It would just rob justice in somebody's front yard. Yeah,

0:14:01.200 --> 0:14:05.080
<v Speaker 1>thank Alright, So I guess I want to introduce one

0:14:05.120 --> 0:14:06.720
<v Speaker 1>of the papers we're gonna be looking at in this

0:14:06.760 --> 0:14:10.200
<v Speaker 1>pair of episodes, and it is by Mark A. Limley

0:14:10.360 --> 0:14:14.520
<v Speaker 1>and Brian Casey called Remedies for Robots, published in the

0:14:14.600 --> 0:14:18.920
<v Speaker 1>University of Chicago Law Review in twenty nineteen. And this

0:14:19.000 --> 0:14:22.600
<v Speaker 1>is a big paper. It's like eighty something pages long

0:14:22.640 --> 0:14:26.000
<v Speaker 1>with a with a lot of different interesting, uh thoughts

0:14:26.000 --> 0:14:27.760
<v Speaker 1>in it. We're not going to be able to cover

0:14:27.880 --> 0:14:31.200
<v Speaker 1>the entire thing in depth, but it's worth looking up.

0:14:31.240 --> 0:14:33.240
<v Speaker 1>You can easily find a full PDF of it if

0:14:33.240 --> 0:14:35.920
<v Speaker 1>you want to read it in depth. And we're gonna

0:14:35.920 --> 0:14:38.440
<v Speaker 1>look at some of the larger framework it lays out,

0:14:38.440 --> 0:14:41.320
<v Speaker 1>and then some interesting thoughts raised to buy it, But

0:14:41.400 --> 0:14:43.960
<v Speaker 1>to kick it off here the the author's right quote,

0:14:44.000 --> 0:14:48.960
<v Speaker 1>what happens when artificially intelligent robots misbehave? The question is

0:14:49.000 --> 0:14:54.239
<v Speaker 1>not just hypothetical. As robotics and artificial intelligence systems increasingly

0:14:54.360 --> 0:14:57.920
<v Speaker 1>integrate into our society, they will do bad things. We

0:14:58.000 --> 0:15:01.000
<v Speaker 1>seek to explore what remedies the all can and should

0:15:01.080 --> 0:15:05.480
<v Speaker 1>provide once a robot has caused harm. Now, obviously we're

0:15:05.480 --> 0:15:08.880
<v Speaker 1>going to be focused less on the like minute particulars

0:15:08.920 --> 0:15:11.760
<v Speaker 1>of US legal precedent here and more on the broader

0:15:11.800 --> 0:15:16.520
<v Speaker 1>issues they raise about robot agency, robot moral decision making

0:15:16.640 --> 0:15:19.960
<v Speaker 1>and how that interacts with harm and morality and justice.

0:15:20.600 --> 0:15:23.440
<v Speaker 1>And the authors start out in their introduction by giving

0:15:23.440 --> 0:15:26.520
<v Speaker 1>what I think is a really fantastic example of how

0:15:26.560 --> 0:15:31.360
<v Speaker 1>an autonomous robot with behaviors guided by machine learning, which

0:15:31.400 --> 0:15:34.720
<v Speaker 1>is how you know increasingly most robots are going to

0:15:34.760 --> 0:15:37.840
<v Speaker 1>be controlled, can end up doing things that are the

0:15:37.880 --> 0:15:41.480
<v Speaker 1>exact opposite of what was intended. So this case that

0:15:41.520 --> 0:15:44.000
<v Speaker 1>they cite is based on a true story from a

0:15:44.040 --> 0:15:48.680
<v Speaker 1>presentation at the eleventh Annual Stanford e Commerce Best Practices

0:15:48.800 --> 0:15:54.440
<v Speaker 1>Conference in June, and it goes like this quote. Engineers

0:15:54.520 --> 0:15:59.480
<v Speaker 1>training and artificially intelligent self flying drone were perplexed. They

0:15:59.520 --> 0:16:01.840
<v Speaker 1>were trying to get the drone to stay within a

0:16:01.920 --> 0:16:05.920
<v Speaker 1>predefined circle and head toward its center. Things were going

0:16:05.960 --> 0:16:09.560
<v Speaker 1>well for a while. The drone received positive reinforcement for

0:16:09.600 --> 0:16:12.600
<v Speaker 1>its successful flights, and it was improving its ability to

0:16:12.720 --> 0:16:17.800
<v Speaker 1>navigate toward the middle quickly and accurately. Then suddenly things changed.

0:16:18.160 --> 0:16:20.480
<v Speaker 1>When the drone near the edge of the circle, it

0:16:20.520 --> 0:16:24.360
<v Speaker 1>would inexplicably turn away from the center, leaving the circle.

0:16:25.040 --> 0:16:28.640
<v Speaker 1>What went wrong. After a long time spent puzzling over

0:16:28.680 --> 0:16:32.160
<v Speaker 1>the problem, the designers realized that whenever the drone left

0:16:32.200 --> 0:16:36.200
<v Speaker 1>the circle during tests they had turned it off. Someone

0:16:36.240 --> 0:16:38.920
<v Speaker 1>would then pick it up and carry it back into

0:16:38.920 --> 0:16:42.680
<v Speaker 1>the circle to start again. From this pattern, the drones

0:16:42.760 --> 0:16:46.760
<v Speaker 1>algorithm had learned correctly that when it was sufficiently far

0:16:46.880 --> 0:16:49.600
<v Speaker 1>from the center, the optimal way to get back to

0:16:49.680 --> 0:16:53.080
<v Speaker 1>the middle was to simply leave it all together. As

0:16:53.120 --> 0:16:55.840
<v Speaker 1>far as the drone was concerned, it had discovered a

0:16:55.920 --> 0:16:59.800
<v Speaker 1>worm hole. Somehow, flying outside of the circle could be

0:17:00.080 --> 0:17:03.240
<v Speaker 1>led upon to magically teleport it closer to the center,

0:17:03.680 --> 0:17:06.520
<v Speaker 1>and far from violating the rules instilled in it by

0:17:06.600 --> 0:17:10.080
<v Speaker 1>its engineers, the drone had actually followed them to a t.

0:17:10.840 --> 0:17:14.399
<v Speaker 1>In doing so, however, it had discovered an unforeseen shortcut,

0:17:14.680 --> 0:17:18.840
<v Speaker 1>one that subverted its designers true intent. That's really good.

0:17:18.880 --> 0:17:22.760
<v Speaker 1>That's that's it's as Yes, I love it. This is

0:17:22.800 --> 0:17:26.239
<v Speaker 1>such a great example of how robots can fail in

0:17:26.320 --> 0:17:30.280
<v Speaker 1>ways that are perfectly logical for the machines themselves, but

0:17:30.440 --> 0:17:33.240
<v Speaker 1>hard for humans to predict in advance, because we're not

0:17:33.400 --> 0:17:37.119
<v Speaker 1>understanding how our you know, our programming or the data

0:17:37.160 --> 0:17:40.480
<v Speaker 1>sets were training it on is biasing its behavior in

0:17:40.840 --> 0:17:43.760
<v Speaker 1>ways that that are strange to us. And in this case,

0:17:43.800 --> 0:17:48.080
<v Speaker 1>of course, such a malfunction is harmless. But as autonomous

0:17:48.119 --> 0:17:51.600
<v Speaker 1>machines become more and more integrated into the broader culture,

0:17:52.040 --> 0:17:56.960
<v Speaker 1>not just in controlled contained locations like factory floors and laboratories,

0:17:56.960 --> 0:17:59.320
<v Speaker 1>but in the wild, so on the streets and in

0:17:59.359 --> 0:18:03.720
<v Speaker 1>our homes and stuff, there will inevitably be cases where

0:18:03.800 --> 0:18:07.800
<v Speaker 1>robots fail like this and fail in ways that cause

0:18:08.240 --> 0:18:12.360
<v Speaker 1>catastrophic harm to people. Yeah, and plus, as an aside,

0:18:12.359 --> 0:18:14.440
<v Speaker 1>we we have to realize that even in cases where

0:18:14.440 --> 0:18:17.040
<v Speaker 1>the machines have not failed, there will be gray areas

0:18:17.359 --> 0:18:20.280
<v Speaker 1>in which it's not completely clear, and an argument could

0:18:20.320 --> 0:18:23.040
<v Speaker 1>be made in these cases for machine culpability with a

0:18:23.119 --> 0:18:26.560
<v Speaker 1>variety of intense and possible biases in place. Oh yeah,

0:18:26.560 --> 0:18:28.719
<v Speaker 1>that's another thing these authors talk about that there can

0:18:28.760 --> 0:18:32.880
<v Speaker 1>be all kinds of ways that that robotics and AI

0:18:32.920 --> 0:18:36.440
<v Speaker 1>could end up causing extreme harm to people without ever

0:18:36.520 --> 0:18:39.720
<v Speaker 1>doing anything that if a human did it would be illegal.

0:18:40.320 --> 0:18:43.760
<v Speaker 1>What one example they give is like if, um, if

0:18:43.920 --> 0:18:48.520
<v Speaker 1>Google were to suddenly change its Google Maps algorithm so

0:18:48.560 --> 0:18:52.960
<v Speaker 1>that it routed all of the city's traffic through your neighborhood. Like,

0:18:53.440 --> 0:18:56.520
<v Speaker 1>nothing illegal about that. It doesn't like commit a crime

0:18:56.560 --> 0:18:59.840
<v Speaker 1>against you, but this is going to drastically negatively in

0:19:00.040 --> 0:19:02.480
<v Speaker 1>hacked your quality of life. And it's a decision that's

0:19:02.560 --> 0:19:05.400
<v Speaker 1>just like a could be a quirk of an algorithm

0:19:05.440 --> 0:19:08.960
<v Speaker 1>in a machine. Now, this paper in particular concerns the

0:19:09.080 --> 0:19:13.159
<v Speaker 1>legal concept of remedies. So I was reading about remedies.

0:19:13.320 --> 0:19:16.600
<v Speaker 1>A common legal definition that I found is quote the

0:19:16.680 --> 0:19:20.480
<v Speaker 1>means to achieve justice in any matter in which legal

0:19:20.600 --> 0:19:24.640
<v Speaker 1>rights are involved. Or in the words of Limely and Casey,

0:19:24.680 --> 0:19:27.040
<v Speaker 1>what do I get when I win? Right? So, if

0:19:27.040 --> 0:19:29.360
<v Speaker 1>you if you take somebody to court because you say

0:19:29.359 --> 0:19:33.119
<v Speaker 1>they have harmed you. Whatever outcome you're seeking from that

0:19:33.119 --> 0:19:36.800
<v Speaker 1>that court case is the remedy. So usually when a

0:19:36.800 --> 0:19:39.480
<v Speaker 1>court case finds that somebody has done something wrong to

0:19:39.560 --> 0:19:42.760
<v Speaker 1>harm somebody else, the court responds to the finding of

0:19:42.800 --> 0:19:46.360
<v Speaker 1>guilt or blame by enforcing this remedy. And common remedies

0:19:46.359 --> 0:19:49.560
<v Speaker 1>would include a payment of money, right a guilty defendant

0:19:49.560 --> 0:19:52.600
<v Speaker 1>has to pay money to the plaintiff, a punishment of

0:19:52.640 --> 0:19:55.520
<v Speaker 1>the offender like maybe they go to jail, or a

0:19:55.520 --> 0:19:58.399
<v Speaker 1>court order to do something or not to do something.

0:19:58.520 --> 0:20:01.719
<v Speaker 1>For example, somebody is ordered not to drive a vehicle,

0:20:01.960 --> 0:20:04.119
<v Speaker 1>or they are ordered not to go within a hundred

0:20:04.160 --> 0:20:07.040
<v Speaker 1>feet of somebody else or something like that. Yeah, or

0:20:07.160 --> 0:20:09.720
<v Speaker 1>their their eyeball is removed, or they have to spend

0:20:09.880 --> 0:20:12.440
<v Speaker 1>a night in a haunted house something like that. Hopefully

0:20:12.480 --> 0:20:15.120
<v Speaker 1>not in modern law. But wait a minute, there are

0:20:15.160 --> 0:20:18.280
<v Speaker 1>some sometimes you do read about some really strange like

0:20:18.600 --> 0:20:21.040
<v Speaker 1>remedies that are ordered by judges like I order you

0:20:21.080 --> 0:20:22.880
<v Speaker 1>to I don't I don't know, to wear chicken suit

0:20:22.960 --> 0:20:26.280
<v Speaker 1>or something right like, Yeah, there's some judges who like

0:20:26.359 --> 0:20:30.159
<v Speaker 1>to get creative. It seems weird. Yeah, I wonder I

0:20:30.200 --> 0:20:31.920
<v Speaker 1>have there have been in cases where someone has to

0:20:31.920 --> 0:20:33.639
<v Speaker 1>spend a night in a haunted house due to a

0:20:33.640 --> 0:20:35.479
<v Speaker 1>court order. I think that would be a good setup

0:20:35.520 --> 0:20:37.960
<v Speaker 1>for a film. But anyway, so when you start looking

0:20:38.000 --> 0:20:41.159
<v Speaker 1>at the idea of remedies, remedies are complicated because they

0:20:41.200 --> 0:20:45.480
<v Speaker 1>involve different types of implied satisfaction on behalf of the

0:20:45.560 --> 0:20:49.639
<v Speaker 1>victim or plaintiff. And some are very clear and material,

0:20:49.840 --> 0:20:52.520
<v Speaker 1>and others are much more abstract. So the very the

0:20:52.520 --> 0:20:55.120
<v Speaker 1>ones that are very clear and material are like, if

0:20:55.160 --> 0:20:58.240
<v Speaker 1>I hit your car with my car and I'm clearly

0:20:58.280 --> 0:21:00.800
<v Speaker 1>at fault, I need to give you a payment of

0:21:00.840 --> 0:21:04.000
<v Speaker 1>cash to offset the material losses to the value of

0:21:04.000 --> 0:21:07.440
<v Speaker 1>your car, right. But then other times it's it's more abstract.

0:21:07.520 --> 0:21:10.720
<v Speaker 1>It's you know, punishment of an offender to give the

0:21:10.840 --> 0:21:15.440
<v Speaker 1>victim a sense of justice or to allegedly discourage someone

0:21:15.560 --> 0:21:18.320
<v Speaker 1>from committing this type of harm or offense in the future.

0:21:18.840 --> 0:21:21.040
<v Speaker 1>And then the authors right that things get way more

0:21:21.119 --> 0:21:24.479
<v Speaker 1>complicated when you bring robots and AI into the picture.

0:21:24.880 --> 0:21:27.760
<v Speaker 1>For example, if you're trying to give a court order

0:21:27.760 --> 0:21:30.159
<v Speaker 1>to a person you know saying like you shall not

0:21:30.280 --> 0:21:32.720
<v Speaker 1>drive a car, you shall not you know, come within

0:21:32.760 --> 0:21:34.840
<v Speaker 1>a hundred feet of this person. You can do so

0:21:34.920 --> 0:21:37.600
<v Speaker 1>in natural language, you can like speak a sentence to

0:21:37.680 --> 0:21:41.000
<v Speaker 1>them and you can expect them to understand. But how

0:21:41.040 --> 0:21:44.320
<v Speaker 1>do you get a court to give an order to

0:21:44.440 --> 0:21:49.080
<v Speaker 1>a robot not to do something? Most robots don't have

0:21:49.240 --> 0:21:52.000
<v Speaker 1>natural language processing, and even if they do, a lot

0:21:52.040 --> 0:21:54.880
<v Speaker 1>of times it's not that good. So you might think, okay,

0:21:54.920 --> 0:21:57.280
<v Speaker 1>well you just you know, you give the court order

0:21:57.359 --> 0:22:00.800
<v Speaker 1>to the robots programmer and then it and then they'll

0:22:00.800 --> 0:22:03.240
<v Speaker 1>have to program the robot to obey. But this is

0:22:03.280 --> 0:22:07.560
<v Speaker 1>also really complicated, like whose responsibility is it the robots

0:22:07.600 --> 0:22:12.120
<v Speaker 1>current owner or the original contractor or creator who made

0:22:12.160 --> 0:22:14.720
<v Speaker 1>the robot? Uh? And what if this is like an

0:22:14.840 --> 0:22:18.359
<v Speaker 1>end user consumer device that the owner doesn't have any

0:22:18.359 --> 0:22:21.600
<v Speaker 1>ability to reprogram, or what if, in the case of

0:22:21.680 --> 0:22:24.800
<v Speaker 1>robots whose behavior is driven by machine learning or some

0:22:24.880 --> 0:22:27.480
<v Speaker 1>other kind of system that is for practical purposes, a

0:22:27.560 --> 0:22:30.239
<v Speaker 1>black box, what if it's not even clear how you

0:22:30.280 --> 0:22:35.919
<v Speaker 1>could reprogram it to reliably obey the rule. Yeah, because

0:22:36.640 --> 0:22:40.080
<v Speaker 1>there's a chance you got to this position because the

0:22:40.200 --> 0:22:45.320
<v Speaker 1>robot misinterpreted what was asked of it. So if you

0:22:45.400 --> 0:22:49.200
<v Speaker 1>then make additional requirements ones that maybe you know, haven't

0:22:49.200 --> 0:22:52.159
<v Speaker 1>actually been tested before, but are just you know that

0:22:52.280 --> 0:22:54.680
<v Speaker 1>that are then brought on by the court that could

0:22:54.720 --> 0:22:58.520
<v Speaker 1>conceivably create new problems, right yeah, yeah, totally, and and

0:22:58.600 --> 0:23:00.919
<v Speaker 1>it keeps getting even more comp located from there like

0:23:01.160 --> 0:23:04.920
<v Speaker 1>Limle and Casey right quote. To complicate matters further, some systems,

0:23:04.960 --> 0:23:10.560
<v Speaker 1>including many self driving cars, distribute responsibility for their robots

0:23:10.600 --> 0:23:15.520
<v Speaker 1>between both designers and downstream operators. For systems of this kind,

0:23:15.560 --> 0:23:20.080
<v Speaker 1>it has already proven extremely difficult to allocate responsibility when

0:23:20.160 --> 0:23:23.600
<v Speaker 1>accidents inevitably occur. It just seems like a real, real

0:23:23.920 --> 0:23:26.679
<v Speaker 1>fast way to get into skynet territory, where it's like

0:23:27.040 --> 0:23:29.920
<v Speaker 1>the robot then decides that the only way to assure

0:23:30.000 --> 0:23:33.040
<v Speaker 1>that it never sells cigarettes to children again is to

0:23:33.160 --> 0:23:36.119
<v Speaker 1>destroy all humans. That sounds like finding a wormhole to me.

0:23:36.960 --> 0:23:39.840
<v Speaker 1>We will be getting into some more wormhole territory as

0:23:39.840 --> 0:23:44.200
<v Speaker 1>we go on, so more complications. Uh. The authors bring

0:23:44.280 --> 0:23:46.479
<v Speaker 1>up the idea of how to courts compel a person

0:23:46.680 --> 0:23:49.399
<v Speaker 1>or a company to obey a court order. Right Like,

0:23:49.440 --> 0:23:52.560
<v Speaker 1>if you know a company is like dumping poison that's

0:23:52.600 --> 0:23:55.840
<v Speaker 1>harming somebody, and the person sues that company, what does

0:23:55.840 --> 0:23:58.080
<v Speaker 1>the court do to get them to stop, Well, the

0:23:58.080 --> 0:24:00.800
<v Speaker 1>there is a threat of contempt of court if they

0:24:00.800 --> 0:24:04.440
<v Speaker 1>don't stop doing it. Right, Courts usually just assume that

0:24:04.480 --> 0:24:07.560
<v Speaker 1>people are motivated by a desire not to pay huge

0:24:07.600 --> 0:24:10.399
<v Speaker 1>monetary damages or a desire not to go to jail.

0:24:10.920 --> 0:24:13.840
<v Speaker 1>Would that have any motivating power on a robot. It

0:24:14.200 --> 0:24:16.800
<v Speaker 1>would only have that power to the extent that the

0:24:16.880 --> 0:24:20.280
<v Speaker 1>robot had been programmed to take that into account. If

0:24:20.320 --> 0:24:22.359
<v Speaker 1>it hadn't, it wouldn't matter at all. Like, you know,

0:24:22.520 --> 0:24:25.320
<v Speaker 1>most robots probably do not have any opinion one way

0:24:25.400 --> 0:24:27.920
<v Speaker 1>or another about whether about going to jail or having

0:24:27.960 --> 0:24:31.159
<v Speaker 1>to pay damages, So you'd have to explicitly program it

0:24:31.240 --> 0:24:36.359
<v Speaker 1>to be disincentivized by potential punishments. Yeah, because take the

0:24:36.400 --> 0:24:39.680
<v Speaker 1>cigarette robot for example, Like it's prime it's prime directive

0:24:39.760 --> 0:24:43.520
<v Speaker 1>is just to sell delicious cigarettes to human beings like

0:24:43.560 --> 0:24:45.919
<v Speaker 1>the the what what else? What kind of leverage do

0:24:45.920 --> 0:24:48.439
<v Speaker 1>you have? Right exactly, So in that case, you'd be

0:24:48.480 --> 0:24:51.320
<v Speaker 1>faced with either you'd be trying to find some kind

0:24:51.359 --> 0:24:54.400
<v Speaker 1>of human who's responsible for its behavior, but you could

0:24:54.640 --> 0:24:57.520
<v Speaker 1>very well run into the problem that like you can't

0:24:57.520 --> 0:25:00.680
<v Speaker 1>really identify any one person who's seems to be at

0:25:00.680 --> 0:25:03.640
<v Speaker 1>fault for what it did, and it's doing this bad thing,

0:25:03.760 --> 0:25:05.400
<v Speaker 1>so so what are you going to do about it?

0:25:06.359 --> 0:25:08.560
<v Speaker 1>And then of course things get even weirder when you

0:25:08.560 --> 0:25:11.000
<v Speaker 1>start getting into that that other side. You know, that's

0:25:11.040 --> 0:25:15.199
<v Speaker 1>like the more like direct and material remedies that can

0:25:15.240 --> 0:25:18.879
<v Speaker 1>be provided by courts, either like a monetary award to

0:25:19.000 --> 0:25:22.000
<v Speaker 1>the victim or in order to stop doing something that

0:25:22.040 --> 0:25:24.320
<v Speaker 1>causes harm. On the other hand, you've got this thing

0:25:24.400 --> 0:25:27.400
<v Speaker 1>that courts often end up engaging in, and people are

0:25:27.440 --> 0:25:31.639
<v Speaker 1>are largely driven and motivated by however, however irrational it

0:25:31.720 --> 0:25:35.240
<v Speaker 1>might be in some cases, and that's the perceived abstract

0:25:35.359 --> 0:25:38.679
<v Speaker 1>value of punishment, you know, not just material damages to

0:25:38.720 --> 0:25:40.760
<v Speaker 1>a victim or in order not to do something, but

0:25:40.840 --> 0:25:45.760
<v Speaker 1>the inflicting of punishments specifically to demonstrate the court's displeasure

0:25:45.800 --> 0:25:49.159
<v Speaker 1>with the original behavior of the defendant. Uh so uh

0:25:49.480 --> 0:25:51.920
<v Speaker 1>They raise a question that's brought up in a paper

0:25:51.960 --> 0:25:56.120
<v Speaker 1>by a professor named Christina Mulligan who explores the subject

0:25:56.160 --> 0:25:58.400
<v Speaker 1>of should you have the right to punch a robot

0:25:58.480 --> 0:26:01.040
<v Speaker 1>that hurts you? Uh lee? In case? He called the

0:26:01.240 --> 0:26:05.560
<v Speaker 1>called this the expressive component of remedies, and though a

0:26:05.640 --> 0:26:09.480
<v Speaker 1>desire to see offenders punished. Maybe an extremely natural and

0:26:09.520 --> 0:26:14.160
<v Speaker 1>nearly universal human drive. It's debatable whether it actually serves

0:26:14.160 --> 0:26:17.439
<v Speaker 1>a purpose in reducing harm, and if it does, in

0:26:17.520 --> 0:26:21.000
<v Speaker 1>what cases it does. I I love this idea because,

0:26:21.640 --> 0:26:24.600
<v Speaker 1>on a very literal level, it makes me think, well,

0:26:24.600 --> 0:26:26.439
<v Speaker 1>why would you punch a robot. They're made out of

0:26:26.440 --> 0:26:28.280
<v Speaker 1>out of metal. You're gonna hurt your hand. All you're

0:26:28.280 --> 0:26:29.760
<v Speaker 1>gonna do is hurt your hand, and you're not going

0:26:29.800 --> 0:26:34.160
<v Speaker 1>to hurt the robot unless first of all, you design

0:26:34.240 --> 0:26:36.800
<v Speaker 1>the robot so that it has at least one punchable

0:26:36.840 --> 0:26:40.239
<v Speaker 1>portion of its anatomy, and then for it to be

0:26:40.320 --> 0:26:43.600
<v Speaker 1>more than just a you know, a cathartic uh uh

0:26:44.880 --> 0:26:47.159
<v Speaker 1>thing for you, then you have to also make sure

0:26:47.200 --> 0:26:51.199
<v Speaker 1>there's some sort of feedback right where punch. Yeah, like

0:26:51.240 --> 0:26:55.000
<v Speaker 1>you punch cigarette bought in it's punchable area, then it

0:26:55.040 --> 0:26:58.040
<v Speaker 1>will say owl, and maybe it will I don't know,

0:26:58.040 --> 0:27:00.800
<v Speaker 1>ottaw inciner rate one packet of reigrettes so that it

0:27:00.880 --> 0:27:03.360
<v Speaker 1>can never sell them that sort of thing. But then, yeah,

0:27:03.359 --> 0:27:06.840
<v Speaker 1>you're having to design your robots to to suffer to

0:27:06.840 --> 0:27:09.880
<v Speaker 1>a certain extent, which I guess that means that goes

0:27:09.920 --> 0:27:12.480
<v Speaker 1>back to what C three po said, right, he said,

0:27:12.760 --> 0:27:15.040
<v Speaker 1>you know about about being made to suffer. It seems

0:27:15.040 --> 0:27:17.320
<v Speaker 1>to be our lot in life. Oh, that's interesting. I

0:27:17.320 --> 0:27:19.560
<v Speaker 1>hadn't thought about that. Yeah. Clearly R two, D two

0:27:19.600 --> 0:27:22.679
<v Speaker 1>and C three p O have have inherent desires to

0:27:22.720 --> 0:27:26.280
<v Speaker 1>avoid pain. They have been programmed with that. Yeah, but

0:27:26.359 --> 0:27:29.440
<v Speaker 1>as we've said that, that's not standard issue for robots.

0:27:29.520 --> 0:27:32.359
<v Speaker 1>Most robots don't care about whether or not they get injured, Like,

0:27:32.440 --> 0:27:35.159
<v Speaker 1>that's not a motivating factor for them. And again it

0:27:35.240 --> 0:27:38.560
<v Speaker 1>raises this bizarre question of like, what are you doing

0:27:38.640 --> 0:27:40.919
<v Speaker 1>when you punch the robot? Like, what is the I

0:27:40.920 --> 0:27:43.840
<v Speaker 1>guess it's making you feel better, But does it make

0:27:43.880 --> 0:27:46.359
<v Speaker 1>you feel better if like you know that the robot

0:27:46.440 --> 0:27:50.280
<v Speaker 1>doesn't actually care? Yeah, and then what then what needs

0:27:50.320 --> 0:27:53.000
<v Speaker 1>to be done to convince you that it does? K Yeah,

0:27:53.000 --> 0:27:55.359
<v Speaker 1>it just gets very sticky, very quickly. And then of

0:27:55.400 --> 0:27:58.320
<v Speaker 1>course turns the mirror back on the way we handle

0:27:59.160 --> 0:28:02.600
<v Speaker 1>human to humans in areas. Right, but anyway, limly in

0:28:02.680 --> 0:28:06.120
<v Speaker 1>casey I guess to to summarize their position, they say, Okay,

0:28:06.160 --> 0:28:11.440
<v Speaker 1>increasingly independent robots and AI are coming. They're they're infiltrating

0:28:11.480 --> 0:28:14.119
<v Speaker 1>more and more into society, and they will inevitably do

0:28:14.200 --> 0:28:18.080
<v Speaker 1>bad things. When that happens, the legal system will try

0:28:18.119 --> 0:28:21.439
<v Speaker 1>to order remedies to make things right in you know,

0:28:21.480 --> 0:28:25.399
<v Speaker 1>when when harm has been caused. Our current legal understanding

0:28:25.400 --> 0:28:28.760
<v Speaker 1>of remedies is based on the assumption of human agents

0:28:28.760 --> 0:28:32.440
<v Speaker 1>and human agents only, and its rules are not suited

0:28:32.480 --> 0:28:36.400
<v Speaker 1>to dealing with robot crime or robot offenses quote. As

0:28:36.440 --> 0:28:39.760
<v Speaker 1>we have shown, failing to recognize those differences could result

0:28:39.800 --> 0:28:45.360
<v Speaker 1>in significant unintended consequences, inadvertently encouraging the wrong behaviors or

0:28:45.400 --> 0:28:50.840
<v Speaker 1>even rendering our most important remedial mechanisms functionally irrelevant. Uh So,

0:28:50.920 --> 0:28:53.840
<v Speaker 1>to take robot agents into account, we're going to have

0:28:53.880 --> 0:28:57.360
<v Speaker 1>to examine and rethink how our systems of remedies work.

0:28:57.880 --> 0:29:00.680
<v Speaker 1>But and this is a point we've been making all ready,

0:29:00.720 --> 0:29:03.280
<v Speaker 1>this could have multiple benefits because it could also lead

0:29:03.320 --> 0:29:06.040
<v Speaker 1>to a better understanding of how we apply these remedies

0:29:06.080 --> 0:29:10.200
<v Speaker 1>to cases dealing exclusively with humans quote. Indeed, one of

0:29:10.240 --> 0:29:13.560
<v Speaker 1>the most pressing challenges raised by the technology is its

0:29:13.600 --> 0:29:17.480
<v Speaker 1>tendency to reveal the trade offs between sidal economic and

0:29:17.560 --> 0:29:21.200
<v Speaker 1>legal values that many of us today make without deeply

0:29:21.240 --> 0:29:26.000
<v Speaker 1>appreciating the downstream consequences. They write, we need a law

0:29:26.040 --> 0:29:29.400
<v Speaker 1>of remedies for robots, but in the final analysis, remedies

0:29:29.440 --> 0:29:32.200
<v Speaker 1>for robots may also end up being remedies for all

0:29:32.280 --> 0:29:34.880
<v Speaker 1>of us. Now, like I said, this is a very

0:29:34.920 --> 0:29:37.080
<v Speaker 1>long paper. We can't do justice to all of the

0:29:37.120 --> 0:29:40.000
<v Speaker 1>subjects they raise, but to focus on some highlights, I

0:29:40.040 --> 0:29:43.280
<v Speaker 1>thought one interesting place to look was when they try

0:29:43.320 --> 0:29:45.680
<v Speaker 1>to get into the definition of what actually makes a

0:29:45.800 --> 0:29:49.200
<v Speaker 1>robot in the legal sense. Obviously, there's going to be

0:29:49.240 --> 0:29:53.240
<v Speaker 1>some difficulty here because think about how differently the term

0:29:53.360 --> 0:29:55.760
<v Speaker 1>is used and how many different things it's applied to

0:29:55.840 --> 0:29:58.640
<v Speaker 1>in the world. Uh. The authors here is cite a

0:29:58.720 --> 0:30:02.640
<v Speaker 1>professor Ryan Hallo, who in the past had written that

0:30:02.720 --> 0:30:07.160
<v Speaker 1>there are three important characteristics that define a robot and

0:30:07.240 --> 0:30:10.040
<v Speaker 1>make it different from any machine, just like a computer

0:30:10.360 --> 0:30:14.440
<v Speaker 1>or phone. And Callo says that these three uh, these

0:30:14.480 --> 0:30:19.360
<v Speaker 1>three qualities are embodiment, emergence, and social valence. So to

0:30:19.440 --> 0:30:23.880
<v Speaker 1>quote from Calo, robotics combines, arguably for the first time,

0:30:24.240 --> 0:30:28.040
<v Speaker 1>the promiscuity of information with the embodied capacity to do

0:30:28.240 --> 0:30:34.200
<v Speaker 1>physical harm. Robots display increasingly emergent behavior, permitting the technology

0:30:34.240 --> 0:30:38.840
<v Speaker 1>to accomplish both useful and unfortunate tasks in unexpected ways.

0:30:38.840 --> 0:30:43.240
<v Speaker 1>I like that idea of unfortunate tasks. Um and robots,

0:30:43.280 --> 0:30:46.360
<v Speaker 1>more so than any technology and history, feel to us

0:30:46.400 --> 0:30:50.440
<v Speaker 1>like social actors, a tendency so strong that soldiers sometimes

0:30:50.520 --> 0:30:55.080
<v Speaker 1>jeopardize themselves to preserve the lives of military robots in

0:30:55.080 --> 0:30:58.200
<v Speaker 1>the field, and lives is in quotes there. Yeah, you

0:30:58.240 --> 0:31:00.680
<v Speaker 1>may remember this from the film it came out a

0:31:00.680 --> 0:31:06.120
<v Speaker 1>few years back, Saving Private Cigarette Robot. It's quite touching.

0:31:06.800 --> 0:31:08.840
<v Speaker 1>I mean, it seems absurd, but it does seem to

0:31:08.840 --> 0:31:11.400
<v Speaker 1>play on our natural biases. I want to talk about

0:31:11.800 --> 0:31:14.360
<v Speaker 1>a couple of examples from a psychology paper in a second,

0:31:14.400 --> 0:31:18.680
<v Speaker 1>but like, uh, we're we're just so ready to look

0:31:18.680 --> 0:31:21.440
<v Speaker 1>at machines like humans and and treat them as such.

0:31:21.520 --> 0:31:25.520
<v Speaker 1>It seems almost impossible to avoid. But anyway to pick

0:31:25.600 --> 0:31:28.360
<v Speaker 1>up with Limly and Casey after after that callo quote

0:31:28.560 --> 0:31:31.320
<v Speaker 1>they say quote. In light of these qualities, Calo argues

0:31:31.360 --> 0:31:34.560
<v Speaker 1>that robots are best thought of as artificial objects or

0:31:34.640 --> 0:31:38.440
<v Speaker 1>systems that sense, process, and act upon the world to

0:31:38.560 --> 0:31:42.240
<v Speaker 1>at least some degree. Thus, a robot in the strongest,

0:31:42.280 --> 0:31:45.240
<v Speaker 1>fullest sense of the term, exists in the world as

0:31:45.240 --> 0:31:50.520
<v Speaker 1>a corporeal object with the capacity to exert itself physically. Though,

0:31:50.760 --> 0:31:54.080
<v Speaker 1>It's interesting to me that even this attempt to give

0:31:54.160 --> 0:31:58.120
<v Speaker 1>a strict and legally useful definition of a robot includes

0:31:58.160 --> 0:32:01.280
<v Speaker 1>a subjective component. I brought this up earlier, the component

0:32:01.320 --> 0:32:06.120
<v Speaker 1>about human feelings, the social valence criterion. The callos sites here.

0:32:06.440 --> 0:32:10.480
<v Speaker 1>This means they feel to us like social actors. Yeah,

0:32:10.520 --> 0:32:12.480
<v Speaker 1>Like I was wondering in all of this, like, where

0:32:12.520 --> 0:32:18.200
<v Speaker 1>does a particularly malicious robo call fit into the scenario? Say,

0:32:18.200 --> 0:32:21.360
<v Speaker 1>a robo call that is not just about trying to

0:32:21.360 --> 0:32:23.800
<v Speaker 1>sell you something, but it's like, you know, actively trying

0:32:23.840 --> 0:32:25.640
<v Speaker 1>to say, get a credit card number out of you

0:32:25.760 --> 0:32:29.480
<v Speaker 1>for nefarious purposes. Yeah, that's a really good point. And

0:32:29.280 --> 0:32:32.920
<v Speaker 1>and along those lines, Limbly and Casey argue that actually,

0:32:33.600 --> 0:32:37.840
<v Speaker 1>they don't think the embodiment criteria of hardware is necessarily

0:32:37.920 --> 0:32:40.640
<v Speaker 1>a good one. That maybe our concept of a robot

0:32:41.000 --> 0:32:44.080
<v Speaker 1>should be less limited to the essentialist quality of being

0:32:44.120 --> 0:32:49.960
<v Speaker 1>embodied and more just apply to anything that exhibits intelligent behavior.

0:32:50.760 --> 0:32:53.320
<v Speaker 1>And exactly things like that robot call would be would

0:32:53.360 --> 0:32:55.480
<v Speaker 1>be a good example. Uh, the things we think of

0:32:55.520 --> 0:32:58.680
<v Speaker 1>as robots probably do. They They're not just like stand

0:32:58.720 --> 0:33:02.560
<v Speaker 1>alone objects. They interact with the broader world in some way,

0:33:02.720 --> 0:33:07.080
<v Speaker 1>but they could be entirely software based. Yeah, I guess certainly.

0:33:07.160 --> 0:33:09.240
<v Speaker 1>The roomba is a great example, or any kind of

0:33:09.280 --> 0:33:13.040
<v Speaker 1>like vacuuming robot where it's it's it's it's in your house,

0:33:13.160 --> 0:33:15.000
<v Speaker 1>or it's in a room in your house. It's it's

0:33:15.080 --> 0:33:19.360
<v Speaker 1>interacting in your environment and it's essentially making decisions about

0:33:19.440 --> 0:33:22.800
<v Speaker 1>how best to move around that space. Sure, but if

0:33:22.840 --> 0:33:24.880
<v Speaker 1>you want to take it out of the embodied space,

0:33:25.000 --> 0:33:27.400
<v Speaker 1>you could have the idea of bots on the Internet.

0:33:27.400 --> 0:33:32.440
<v Speaker 1>There's things out there acting autonomously to some extent and doing,

0:33:32.480 --> 0:33:35.960
<v Speaker 1>you know, executing some behavior, acting almost maliciously. But we

0:33:36.040 --> 0:33:39.640
<v Speaker 1>we are tempted to call them bots meaning short for robots,

0:33:39.680 --> 0:33:43.720
<v Speaker 1>because they have some kind of apparent independent agency and

0:33:43.720 --> 0:33:48.240
<v Speaker 1>they're doing something that seems at least halfway intelligent. Right, Yeah,

0:33:48.240 --> 0:33:51.200
<v Speaker 1>And you can easily imagine how they could they are

0:33:51.240 --> 0:33:53.000
<v Speaker 1>and well they I mean, they are used maliciously in

0:33:53.040 --> 0:33:56.360
<v Speaker 1>some cases, but how something like a social media bot

0:33:56.360 --> 0:33:59.040
<v Speaker 1>that responds to certain comments in a particular way, Like,

0:33:59.120 --> 0:34:01.640
<v Speaker 1>it's very easy to imagine how you how that could

0:34:01.680 --> 0:34:04.880
<v Speaker 1>be utilized in a way that would be like not

0:34:04.960 --> 0:34:09.160
<v Speaker 1>only annoying, but just that outright harmful, even physically harmful.

0:34:09.360 --> 0:34:11.439
<v Speaker 1>Oh yeah, I mean think about some of these, uh say,

0:34:11.480 --> 0:34:15.239
<v Speaker 1>like bots on social media that try to crowdsource information,

0:34:15.440 --> 0:34:18.799
<v Speaker 1>like during a natural disaster or something like that. You

0:34:18.800 --> 0:34:23.480
<v Speaker 1>could imagine uh, intentionally maliciously manipulating a bot of this

0:34:23.640 --> 0:34:26.480
<v Speaker 1>kind to like have you know, bad information on it

0:34:26.600 --> 0:34:29.560
<v Speaker 1>or something. Yeah, yeah, or you know, anything that a

0:34:29.600 --> 0:34:33.640
<v Speaker 1>troll can do on social media, a bought could conceivably

0:34:33.680 --> 0:34:35.840
<v Speaker 1>do as well. So that just you know, opens up

0:34:35.880 --> 0:34:40.000
<v Speaker 1>the door, right. But coming back to this, so there's

0:34:40.040 --> 0:34:44.800
<v Speaker 1>this interesting idea that robots feel to us like social actors,

0:34:44.840 --> 0:34:47.400
<v Speaker 1>and that seems to be, at least by some people's definitions,

0:34:47.680 --> 0:34:51.680
<v Speaker 1>a kind of inextricable quality of what makes a robot

0:34:51.880 --> 0:34:54.840
<v Speaker 1>like it feels like, at least to a small extent,

0:34:54.960 --> 0:34:57.680
<v Speaker 1>like a person somehow. And it reminds me of the

0:34:57.800 --> 0:35:00.800
<v Speaker 1>psychology paper I was looking at just recently on human

0:35:00.880 --> 0:35:05.360
<v Speaker 1>social interaction with robots that is by Elizabeth Broadbent called

0:35:05.480 --> 0:35:09.360
<v Speaker 1>Interactions with Robots The Truths We Reveal about Ourselves, published

0:35:09.360 --> 0:35:13.040
<v Speaker 1>in the Annual Review of Psychology in seventeen. Uh. This

0:35:13.120 --> 0:35:15.080
<v Speaker 1>was a highly cited paper and it seems to be

0:35:15.120 --> 0:35:17.240
<v Speaker 1>it's a big literature review of a lot of different

0:35:17.560 --> 0:35:22.560
<v Speaker 1>stuff about about how humans interact emotionally and socially with robots,

0:35:23.080 --> 0:35:25.520
<v Speaker 1>And the one section I was thinking about was where

0:35:25.520 --> 0:35:28.839
<v Speaker 1>she reviews a bunch of other studies about how we

0:35:28.960 --> 0:35:32.960
<v Speaker 1>mindlessly apply social rules to robots. So there are a

0:35:33.000 --> 0:35:35.120
<v Speaker 1>ton of different examples, but just to cite a couple

0:35:35.160 --> 0:35:38.600
<v Speaker 1>of them, one she writes up his quote. After using

0:35:38.600 --> 0:35:42.640
<v Speaker 1>a computer, people evaluate its performance more highly if the

0:35:42.800 --> 0:35:47.160
<v Speaker 1>same computer delivers the rating scale then if another computer

0:35:47.320 --> 0:35:49.799
<v Speaker 1>delivers the rating scale, or if they rate it with

0:35:49.880 --> 0:35:53.640
<v Speaker 1>pen and paper. So like, if you know, you get

0:35:53.640 --> 0:35:56.040
<v Speaker 1>a thing at the end of a test that says like, hey,

0:35:56.080 --> 0:35:58.560
<v Speaker 1>you know, how did you enjoy interacting with this machine,

0:35:58.920 --> 0:36:02.400
<v Speaker 1>You're more likely to give it a higher score if

0:36:02.440 --> 0:36:04.600
<v Speaker 1>you're still sitting on the same machine. Or at least

0:36:04.600 --> 0:36:06.840
<v Speaker 1>that was what was found by nas at all in

0:36:06.920 --> 0:36:12.120
<v Speaker 1>nine UH and the and Broadbent rights quote. This result

0:36:12.160 --> 0:36:15.200
<v Speaker 1>is similar to experiment or bias, in which people try

0:36:15.280 --> 0:36:19.240
<v Speaker 1>not to offend a human researcher. Another example of social

0:36:19.239 --> 0:36:23.680
<v Speaker 1>behavior is reciprocity. We help others who help us. People

0:36:23.760 --> 0:36:26.759
<v Speaker 1>helped to computer with a task for more time and

0:36:26.800 --> 0:36:30.760
<v Speaker 1>more accurately if the computer first helped them with a task,

0:36:30.880 --> 0:36:32.799
<v Speaker 1>then if it did not, and this was found by

0:36:32.880 --> 0:36:37.520
<v Speaker 1>Fog and nas In. I love that idea of people,

0:36:37.719 --> 0:36:41.239
<v Speaker 1>you know, being more reluctant to rate a computer, uh

0:36:41.239 --> 0:36:44.840
<v Speaker 1>poorly if they're still interacting with the same computer. That

0:36:44.960 --> 0:36:48.440
<v Speaker 1>that's that seems perfectly true to me. But another interesting

0:36:48.480 --> 0:36:52.239
<v Speaker 1>one from the summary is quote research and psychology has

0:36:52.239 --> 0:36:56.000
<v Speaker 1>shown that the presence of an observer can increase people's honesty,

0:36:56.120 --> 0:36:59.200
<v Speaker 1>but incentives for cheating can reduce honesty, and this is

0:36:59.239 --> 0:37:02.400
<v Speaker 1>found by Covey at all in nineteen eighty nine. In

0:37:02.440 --> 0:37:05.880
<v Speaker 1>a robot version of this work, participants given incentives to

0:37:05.960 --> 0:37:09.680
<v Speaker 1>cheat were shown to be less honest when alone compared

0:37:09.719 --> 0:37:13.600
<v Speaker 1>to when they were accompanied by either a human or

0:37:13.640 --> 0:37:16.200
<v Speaker 1>by a simple robot, and that was found by Hoffman

0:37:16.280 --> 0:37:20.239
<v Speaker 1>at all in. This illustrates that the social presence of

0:37:20.320 --> 0:37:23.280
<v Speaker 1>robots may make people feel as though they're being watched

0:37:23.320 --> 0:37:26.640
<v Speaker 1>and increase their honesty in an effect similar to that

0:37:26.719 --> 0:37:30.000
<v Speaker 1>produced by the presence of humans. Now this is inter right.

0:37:30.040 --> 0:37:32.440
<v Speaker 1>This This also reminds me of various studies that have

0:37:32.520 --> 0:37:37.560
<v Speaker 1>gone into sort of the idea of of imagine beings

0:37:37.760 --> 0:37:42.000
<v Speaker 1>or religious beings watching us while we're doing things right,

0:37:42.200 --> 0:37:45.000
<v Speaker 1>or even just like I imagery, like putting some eyes

0:37:45.239 --> 0:37:48.000
<v Speaker 1>imagery on a wall looking at people while they're like,

0:37:48.040 --> 0:37:50.080
<v Speaker 1>I don't know, not supposed to steal from the collection

0:37:50.120 --> 0:37:52.120
<v Speaker 1>plate or something like that. I don't know if it's

0:37:52.160 --> 0:37:54.560
<v Speaker 1>to the same extent, but at least in the same

0:37:54.600 --> 0:37:57.960
<v Speaker 1>direction that the presence of another human is. You know,

0:37:57.800 --> 0:37:59.920
<v Speaker 1>your You might be a little bit worried that are

0:38:00.000 --> 0:38:02.279
<v Speaker 1>who D two is gonna, you know, judge your moral

0:38:02.400 --> 0:38:05.520
<v Speaker 1>character harshly or tattle on you. I'm not as worried

0:38:05.520 --> 0:38:16.480
<v Speaker 1>about our two, but um three po snitch than Coming

0:38:16.520 --> 0:38:19.000
<v Speaker 1>back to Limele and Casey, so they talked for a

0:38:19.040 --> 0:38:22.480
<v Speaker 1>long time about how robots get their intelligence. They talked

0:38:22.480 --> 0:38:25.640
<v Speaker 1>about the importance of machine learning for the modern generations

0:38:25.719 --> 0:38:28.920
<v Speaker 1>of robots and AI, that it's just not practical to

0:38:29.520 --> 0:38:32.279
<v Speaker 1>hard code AI the way we used to imagine. You know,

0:38:32.320 --> 0:38:34.480
<v Speaker 1>you'd be a programmer and you're just like creating a

0:38:34.480 --> 0:38:37.600
<v Speaker 1>lot of strings of if then statements like you know,

0:38:37.960 --> 0:38:40.840
<v Speaker 1>the kind of intelligence that we expect from a modern

0:38:40.880 --> 0:38:44.560
<v Speaker 1>AI or or intelligent robot is too complex for people

0:38:44.600 --> 0:38:47.320
<v Speaker 1>to program in a in a direct way like that. Instead,

0:38:47.360 --> 0:38:50.000
<v Speaker 1>they've got to be trained on natural data sets through

0:38:50.040 --> 0:38:53.160
<v Speaker 1>machine learning, but of course doing so comes at the

0:38:53.239 --> 0:38:58.640
<v Speaker 1>cost of increasing uncertainty about their future behaviors. Behaviors could

0:38:58.640 --> 0:39:02.160
<v Speaker 1>emerge that a conchi and just programmer would never intentionally

0:39:02.280 --> 0:39:05.959
<v Speaker 1>hard code into the system. Uh So, so that brings

0:39:06.000 --> 0:39:09.560
<v Speaker 1>us to like, what types of harms could we expect

0:39:09.640 --> 0:39:12.680
<v Speaker 1>from robots and AI? And the authors here come up

0:39:12.760 --> 0:39:15.040
<v Speaker 1>with what I think are some very useful categories, some

0:39:15.120 --> 0:39:17.800
<v Speaker 1>sort of like cubby holes to slot the different types

0:39:17.840 --> 0:39:20.920
<v Speaker 1>of AI fears into. So the first kind is what

0:39:20.960 --> 0:39:23.839
<v Speaker 1>they call unavoidable harms. These are probably not the main

0:39:23.880 --> 0:39:26.440
<v Speaker 1>ones to be worried about, but they are worth thinking about.

0:39:26.880 --> 0:39:29.400
<v Speaker 1>Uh And this is just the fact that some dangers

0:39:29.440 --> 0:39:32.880
<v Speaker 1>are inherent too many products and services, we just accept

0:39:32.960 --> 0:39:36.000
<v Speaker 1>them as the cost of having those products and services

0:39:36.000 --> 0:39:38.279
<v Speaker 1>in the first place. So like this would just be

0:39:38.880 --> 0:39:42.080
<v Speaker 1>cigarette bought just by virtue of selling cigarettes is doing

0:39:42.080 --> 0:39:44.719
<v Speaker 1>harm to people? Right? Yes, I mean, yeah, the fact

0:39:44.840 --> 0:39:47.319
<v Speaker 1>that you have cigarettes. There is some harm coming from that.

0:39:47.400 --> 0:39:51.120
<v Speaker 1>But there are also ones that are more fully integrated

0:39:51.120 --> 0:39:54.680
<v Speaker 1>into just the way society works, like having cars. It

0:39:54.840 --> 0:39:58.080
<v Speaker 1>is absolutely inevitable that people driving cars are going to

0:39:58.080 --> 0:40:00.759
<v Speaker 1>crash their cars and there will be FATA lalities from that,

0:40:01.000 --> 0:40:03.000
<v Speaker 1>and you can think of ways of reducing it, but

0:40:03.040 --> 0:40:07.080
<v Speaker 1>there's there's really not any expectation that we can have

0:40:07.320 --> 0:40:10.239
<v Speaker 1>a country that has car based transportation and there will

0:40:10.280 --> 0:40:13.200
<v Speaker 1>not be any accidents because there will always be things

0:40:13.280 --> 0:40:16.120
<v Speaker 1>that are that are not even reducible to driver error

0:40:16.239 --> 0:40:18.240
<v Speaker 1>or to malfunction of the cars, right, like a tree

0:40:18.280 --> 0:40:21.440
<v Speaker 1>falls on the road or something, right, birds, wild animals.

0:40:21.440 --> 0:40:24.080
<v Speaker 1>Any So, even though they're I think there's some very

0:40:24.080 --> 0:40:28.120
<v Speaker 1>convincing arguments to be made that uh a switch to

0:40:28.280 --> 0:40:33.360
<v Speaker 1>self driving cars would create a much safer uh travel environment,

0:40:33.360 --> 0:40:36.080
<v Speaker 1>that it would make roads safer. You're not gonna you're

0:40:36.080 --> 0:40:39.840
<v Speaker 1>not gonna get to absolute zero crashes or absolute zero

0:40:39.960 --> 0:40:43.360
<v Speaker 1>road fatalities, right, I mean you wouldn't even if the

0:40:43.800 --> 0:40:46.839
<v Speaker 1>driving algorithms were perfect, right, and they're probably not going

0:40:46.880 --> 0:40:49.200
<v Speaker 1>to be perfect. They may well be and probably are

0:40:49.239 --> 0:40:52.800
<v Speaker 1>going to be better than the average human driver. Yeah, Okay,

0:40:52.840 --> 0:40:55.640
<v Speaker 1>So that's just there's just unavoidable harm that comes from

0:40:55.760 --> 0:40:58.319
<v Speaker 1>using any type of product or service, and when you

0:40:58.360 --> 0:41:01.000
<v Speaker 1>integrate robotics in AI and of that product or service,

0:41:01.080 --> 0:41:04.200
<v Speaker 1>those unavoidable harms will just continue. But that's something we

0:41:04.360 --> 0:41:08.960
<v Speaker 1>already deal with. The next category is deliberate least cost harms.

0:41:09.560 --> 0:41:12.280
<v Speaker 1>This is similar to unavoidable harms, but it's in cases

0:41:12.280 --> 0:41:15.920
<v Speaker 1>where the machine actually is able to make a decision

0:41:16.160 --> 0:41:19.520
<v Speaker 1>with with important ramifications, like it can make a decision

0:41:19.560 --> 0:41:22.719
<v Speaker 1>to act in a way that causes harm, but is

0:41:22.760 --> 0:41:26.200
<v Speaker 1>attempting to cause the least harm possible. So, in a sense,

0:41:26.320 --> 0:41:29.640
<v Speaker 1>this is forcing robots to do the trolley problem. Right,

0:41:29.960 --> 0:41:32.520
<v Speaker 1>do you switch to the track that has one person

0:41:32.560 --> 0:41:36.680
<v Speaker 1>sitting on the train tracks instead of five people? And

0:41:36.719 --> 0:41:40.480
<v Speaker 1>this will be another inevitable capability of autonomous cars, but

0:41:40.520 --> 0:41:44.319
<v Speaker 1>it raises all kinds of thorny questions. If an autonomous

0:41:44.440 --> 0:41:47.440
<v Speaker 1>vehicle can avoid a head on collision that will likely

0:41:47.520 --> 0:41:50.600
<v Speaker 1>kill multiple people by suddenly swerving out of the way

0:41:50.680 --> 0:41:55.279
<v Speaker 1>and hitting one pedestrian, that may indeed avoid a greater harm,

0:41:55.880 --> 0:41:58.239
<v Speaker 1>but that's probably cold comfort to the one person who

0:41:58.320 --> 0:42:01.560
<v Speaker 1>got hit. Right, Yeah, and then when you you have

0:42:01.600 --> 0:42:05.239
<v Speaker 1>a robot or some sort of an AI involved in

0:42:05.280 --> 0:42:09.000
<v Speaker 1>that decision making, I mean, it's it's you can just

0:42:09.080 --> 0:42:13.799
<v Speaker 1>imagine the the the intensity of the arguments and the

0:42:13.920 --> 0:42:17.280
<v Speaker 1>conversations that would ensue. Right, But then the authors raised

0:42:17.280 --> 0:42:19.759
<v Speaker 1>what I think is a very interesting point. They say

0:42:19.800 --> 0:42:22.520
<v Speaker 1>that this kind of life or death trolley problem will

0:42:22.520 --> 0:42:27.160
<v Speaker 1>probably be the exception rather than the rule. Instead, they say, quote, uh,

0:42:27.360 --> 0:42:32.000
<v Speaker 1>far likelier, I'll build albeit subtler scenarios involving least cost

0:42:32.080 --> 0:42:36.279
<v Speaker 1>harms will involve robots that make decisions with seemingly trivial

0:42:36.400 --> 0:42:40.480
<v Speaker 1>implications at an individual level, but which result in non

0:42:40.560 --> 0:42:45.000
<v Speaker 1>trivial impacts at scale. Self driving cars, for example, will

0:42:45.120 --> 0:42:48.400
<v Speaker 1>rarely face a stark choice between killing a child or

0:42:48.480 --> 0:42:52.200
<v Speaker 1>killing two elderly people, but thousands of times a day

0:42:52.239 --> 0:42:55.480
<v Speaker 1>they will have to choose precisely where to change lanes,

0:42:55.880 --> 0:42:59.480
<v Speaker 1>how closely to trail another vehicle, when to accelerate on

0:42:59.520 --> 0:43:02.719
<v Speaker 1>a freeway on ramp, and so forth. Each of these

0:43:02.760 --> 0:43:08.160
<v Speaker 1>decisions will entail some probability of injuring someone, I guess.

0:43:08.160 --> 0:43:09.799
<v Speaker 1>Another thing to keep in mind, like with the trap,

0:43:09.920 --> 0:43:13.759
<v Speaker 1>with the trolley problem, generally, when you're dealing with it,

0:43:13.920 --> 0:43:15.920
<v Speaker 1>there's a there's a lot of emphasis on the problem

0:43:16.000 --> 0:43:18.920
<v Speaker 1>aspect of it, you know, like the trolley problem should

0:43:18.960 --> 0:43:22.600
<v Speaker 1>be a an ethical dilemma. It should it should hurt

0:43:22.640 --> 0:43:25.120
<v Speaker 1>a bit to try and figure out how of what

0:43:25.280 --> 0:43:27.960
<v Speaker 1>to do, and the idea of the trolley problem being

0:43:28.040 --> 0:43:31.960
<v Speaker 1>something that is encountered and decided upon, like as in

0:43:32.000 --> 0:43:35.840
<v Speaker 1>a split second, by a machine, um, by an algorithm

0:43:35.880 --> 0:43:39.200
<v Speaker 1>like that. That feels that that feels a bit worse

0:43:39.239 --> 0:43:42.120
<v Speaker 1>to us. You know, that feels like if if it's

0:43:42.160 --> 0:43:44.160
<v Speaker 1>if it's an easy decision, even if it's just based

0:43:44.160 --> 0:43:47.960
<v Speaker 1>purely on math, you know, it's um, it feels wrong

0:43:47.960 --> 0:43:50.880
<v Speaker 1>on some level. Oh yeah, yeah, so I think you're right.

0:43:51.120 --> 0:43:53.600
<v Speaker 1>But also the thing they're bringing up here is that

0:43:53.719 --> 0:43:56.680
<v Speaker 1>the trolley problem you're actually more often facing is that

0:43:57.280 --> 0:43:59.680
<v Speaker 1>every single day you're your autonomous car is going to

0:43:59.800 --> 0:44:02.800
<v Speaker 1>make you know, hundreds or thousands of trolley problem calls

0:44:02.800 --> 0:44:07.480
<v Speaker 1>where on one track it is getting to your destination

0:44:08.000 --> 0:44:10.920
<v Speaker 1>a few seconds faster, and on the other track is

0:44:10.960 --> 0:44:14.279
<v Speaker 1>a one in a million chance of killing somebody. Yeah yeah,

0:44:14.320 --> 0:44:16.399
<v Speaker 1>and these do. We make these decisions all the time,

0:44:16.560 --> 0:44:18.399
<v Speaker 1>but we don't focus on these. That's but I think

0:44:18.400 --> 0:44:20.839
<v Speaker 1>that's part of the issue, you know, Yeah, exactly, You're like,

0:44:20.880 --> 0:44:24.440
<v Speaker 1>should I, okay, should I take a left on this road? Well,

0:44:24.480 --> 0:44:27.440
<v Speaker 1>there's a chance there's a speeding car just above, just

0:44:27.560 --> 0:44:29.520
<v Speaker 1>over the edge there, and I can't see it, but

0:44:29.560 --> 0:44:31.360
<v Speaker 1>I'm going to take that chance because I want to

0:44:31.400 --> 0:44:34.239
<v Speaker 1>cut three minutes off my drive to work. Yes. Uh,

0:44:34.520 --> 0:44:36.680
<v Speaker 1>this is actually a very good point that we we

0:44:36.800 --> 0:44:40.400
<v Speaker 1>already make these decisions, but we just don't think about

0:44:40.480 --> 0:44:44.160
<v Speaker 1>them in these explicit probability calculations, and there may be

0:44:44.239 --> 0:44:46.920
<v Speaker 1>some consequences to thinking about them this way, which is

0:44:46.960 --> 0:44:49.640
<v Speaker 1>there could be a weird like perceived downside just to

0:44:49.880 --> 0:44:53.919
<v Speaker 1>making these kind of these kind of calculations objective and

0:44:53.920 --> 0:44:56.319
<v Speaker 1>and explicit. Yeah, I mean I've I've run into this

0:44:56.360 --> 0:44:58.640
<v Speaker 1>with some of the map programs that I used to

0:44:58.719 --> 0:45:01.040
<v Speaker 1>drive before, where I want to tell it in some

0:45:01.080 --> 0:45:03.520
<v Speaker 1>cases like like give me the ability and maybe they

0:45:03.520 --> 0:45:06.360
<v Speaker 1>have this now, but there was one left turn in

0:45:06.400 --> 0:45:09.400
<v Speaker 1>particular where I would the ability to to flag this

0:45:09.480 --> 0:45:12.400
<v Speaker 1>left turn. This is a dangerous left turn. You have

0:45:12.520 --> 0:45:15.520
<v Speaker 1>put me in a position to make. Um, I might

0:45:15.640 --> 0:45:17.920
<v Speaker 1>know which left turn you're talking about. If you probably

0:45:18.200 --> 0:45:21.080
<v Speaker 1>in town, it's it's it's in town. It's near our office.

0:45:21.120 --> 0:45:23.799
<v Speaker 1>So yeah, I mean yeah, by the way, if you're

0:45:23.800 --> 0:45:27.200
<v Speaker 1>out there working on programming driving apps, you should absolutely

0:45:27.239 --> 0:45:29.759
<v Speaker 1>include the A toggle key where you can say no

0:45:29.920 --> 0:45:33.399
<v Speaker 1>left turns please. Yes, that that is highly useful. I'm

0:45:33.440 --> 0:45:36.719
<v Speaker 1>to understand this is how one of my aunts got around. Like,

0:45:36.880 --> 0:45:39.160
<v Speaker 1>if they got older and they were less adventurous driving,

0:45:39.200 --> 0:45:41.319
<v Speaker 1>they would only take right turns, and they would do

0:45:41.360 --> 0:45:43.680
<v Speaker 1>all their driving so that no left turns were made.

0:45:44.040 --> 0:45:46.000
<v Speaker 1>I think I have one time read This could be

0:45:46.040 --> 0:45:49.360
<v Speaker 1>totally wrong, but I at least one time I remember

0:45:49.400 --> 0:45:53.520
<v Speaker 1>reading a claim that, like, you know, the traffic efficiency

0:45:53.520 --> 0:45:56.480
<v Speaker 1>would be x percent higher and people would spend x

0:45:57.000 --> 0:45:59.279
<v Speaker 1>number of minutes less time in traffic if there were

0:45:59.320 --> 0:46:01.560
<v Speaker 1>no such thing as left turns, if everybody had to

0:46:01.560 --> 0:46:04.680
<v Speaker 1>get everywhere by only doing you know, full right turns

0:46:04.719 --> 0:46:07.880
<v Speaker 1>to to go around the block. Interesting. I'm sure there

0:46:07.880 --> 0:46:09.920
<v Speaker 1>would be some cases where you can't do that, but

0:46:10.000 --> 0:46:12.080
<v Speaker 1>you know, in a in a grid city, seems to

0:46:12.120 --> 0:46:13.919
<v Speaker 1>make a lot of sense. Maybe you get like one

0:46:14.000 --> 0:46:17.040
<v Speaker 1>left turn of day. It's some sort of card system.

0:46:17.239 --> 0:46:19.880
<v Speaker 1>But like I said, I I cannot confirm that. Okay,

0:46:19.880 --> 0:46:22.320
<v Speaker 1>But anyway, the next categories of harm they talk about

0:46:22.520 --> 0:46:25.719
<v Speaker 1>this one is defect driven harms. Uh, this one is

0:46:25.880 --> 0:46:28.680
<v Speaker 1>very easy to understand. The robot harms someone because of

0:46:28.680 --> 0:46:31.600
<v Speaker 1>a design flaw or a bug or a mistake, or

0:46:31.680 --> 0:46:35.360
<v Speaker 1>it's just broken. You know, a warehouse loading robot is

0:46:35.440 --> 0:46:38.640
<v Speaker 1>designed to only operate when no humans are nearby it.

0:46:38.840 --> 0:46:41.400
<v Speaker 1>But there's a malfunction with one of its sensors and

0:46:41.480 --> 0:46:44.280
<v Speaker 1>it fails to detect the presence of a human operator

0:46:44.800 --> 0:46:46.400
<v Speaker 1>trying to get I don't know, a piece of junk

0:46:46.440 --> 0:46:48.640
<v Speaker 1>out of it out of one of its hinges, and

0:46:48.719 --> 0:46:52.160
<v Speaker 1>it moves and kills them. Okay, this is pretty straightforward.

0:46:52.160 --> 0:46:54.880
<v Speaker 1>Just it's broken for some reason. The authors here do

0:46:54.960 --> 0:46:57.200
<v Speaker 1>point out that this gets even more complicated when there

0:46:57.320 --> 0:47:01.040
<v Speaker 1>is a human in the loop e g. An autonomous

0:47:01.120 --> 0:47:04.440
<v Speaker 1>car with a human driver who is supposed to intervene

0:47:04.600 --> 0:47:06.719
<v Speaker 1>in the event of an emergency. They talked about one

0:47:06.760 --> 0:47:09.640
<v Speaker 1>case where this happened with with I believe it was

0:47:09.680 --> 0:47:13.880
<v Speaker 1>an uber autonomous vehicle where both the machine and the

0:47:14.000 --> 0:47:17.960
<v Speaker 1>human fail, that both of them failed to stop a

0:47:18.000 --> 0:47:21.640
<v Speaker 1>collision that hurts someone, Like what happens here? Yeah, yeah,

0:47:21.800 --> 0:47:24.480
<v Speaker 1>of course, we we have very similar cases in just

0:47:24.560 --> 0:47:27.560
<v Speaker 1>purely human affairs. Right when questions are asked, like where

0:47:27.600 --> 0:47:30.600
<v Speaker 1>was this person's supervisor? Uh, you know, who who were

0:47:30.600 --> 0:47:32.920
<v Speaker 1>the watchers? Who? There should have been some other person,

0:47:33.160 --> 0:47:35.279
<v Speaker 1>there was someone else in the loop here, Why didn't

0:47:35.320 --> 0:47:39.080
<v Speaker 1>they do something to stop this crime from taking place? Right? Okay?

0:47:39.120 --> 0:47:42.319
<v Speaker 1>After that you get into misuse harms. Now, some of

0:47:42.320 --> 0:47:45.759
<v Speaker 1>these are very obvious, very straightforward, like if you program

0:47:45.760 --> 0:47:48.600
<v Speaker 1>a robot directly to go kill someone, or even if

0:47:48.640 --> 0:47:52.680
<v Speaker 1>you program it to wander around at random swinging a machete.

0:47:52.680 --> 0:47:55.680
<v Speaker 1>In these cases, it seems that the human programmer is

0:47:55.719 --> 0:47:58.839
<v Speaker 1>clearly at fault, Right, the robot has just become a

0:47:58.840 --> 0:48:02.240
<v Speaker 1>weapon of murder or of reckless endangerment, and the person

0:48:02.239 --> 0:48:05.399
<v Speaker 1>who told it to do that is the person responsible. Yeah.

0:48:05.440 --> 0:48:10.080
<v Speaker 1>Like if you take an automotive, uh, like oil changed

0:48:10.160 --> 0:48:15.680
<v Speaker 1>robot and you reprogram it to um do appendectomies and

0:48:15.760 --> 0:48:17.960
<v Speaker 1>people die as a result, Like, that's a misuse you

0:48:17.960 --> 0:48:21.240
<v Speaker 1>can you can only blame the the oil changed robots

0:48:21.280 --> 0:48:26.160
<v Speaker 1>so much because it was not ultimately designed to perform appendectomies. Right.

0:48:26.160 --> 0:48:28.320
<v Speaker 1>In this case, this is more like the hammer example

0:48:28.520 --> 0:48:31.680
<v Speaker 1>used at the beginning. This it's not the robot autonomously

0:48:31.840 --> 0:48:34.760
<v Speaker 1>making the decision to do this. Uh, this is somebody

0:48:34.800 --> 0:48:38.640
<v Speaker 1>just using it as a tool of crime. Yeah, But

0:48:38.840 --> 0:48:42.040
<v Speaker 1>the authors point out that there are cases where quote,

0:48:42.080 --> 0:48:45.280
<v Speaker 1>people will misuse robots in a manner that is neither

0:48:45.440 --> 0:48:50.160
<v Speaker 1>negligent nor criminal, but nevertheless threatens to harm others, and

0:48:50.239 --> 0:48:54.280
<v Speaker 1>these types of harm are especially difficult to predict and prevent.

0:48:55.120 --> 0:48:58.400
<v Speaker 1>So one example is just people love to trick robots,

0:48:58.440 --> 0:49:02.000
<v Speaker 1>people like to mess around with robots and AI. I

0:49:02.040 --> 0:49:05.520
<v Speaker 1>would admit to myself finding this amusing and principle we've

0:49:05.520 --> 0:49:07.480
<v Speaker 1>talked about this in uh, you know, the flate of

0:49:07.520 --> 0:49:10.319
<v Speaker 1>sex Mokena episodes. But of course there are times when

0:49:10.360 --> 0:49:12.680
<v Speaker 1>it's not so funny, when when people take it to

0:49:12.760 --> 0:49:16.560
<v Speaker 1>really sinister places. One example the authors bring up here

0:49:16.719 --> 0:49:20.120
<v Speaker 1>is the horrible saga of Microsoft ta Do you remember

0:49:20.160 --> 0:49:23.360
<v Speaker 1>this thing? Oh? This was the this is the robot

0:49:23.400 --> 0:49:26.000
<v Speaker 1>that was traveling across the country. No, no, no, no,

0:49:26.160 --> 0:49:28.279
<v Speaker 1>uh though I know what you're talking about there. No,

0:49:28.760 --> 0:49:31.160
<v Speaker 1>maybe we can come back to that. But Tay was

0:49:31.280 --> 0:49:36.440
<v Speaker 1>a Twitter chat bot created by Microsoft that was supposed

0:49:36.520 --> 0:49:40.000
<v Speaker 1>to learn how to interact on the Internet just by

0:49:40.080 --> 0:49:42.759
<v Speaker 1>learning from conversations it had with real users. So you

0:49:42.760 --> 0:49:45.120
<v Speaker 1>could tweet it Tay and say hey, how are you doing?

0:49:45.280 --> 0:49:47.480
<v Speaker 1>You know, and you could talk about the weather or whatever.

0:49:47.560 --> 0:49:51.760
<v Speaker 1>But of course who who ended up engaging and training

0:49:51.760 --> 0:49:54.440
<v Speaker 1>this AI to speak? It was like the worst trolls

0:49:54.480 --> 0:49:57.040
<v Speaker 1>on the internet. So within a matter of hours, this

0:49:57.200 --> 0:50:00.680
<v Speaker 1>brand new chat bot had been transformed from from a

0:50:00.840 --> 0:50:03.359
<v Speaker 1>from a you know, a lump of clay unformed into

0:50:03.360 --> 0:50:08.279
<v Speaker 1>a pornographic nazi. Yes, I do remember this now. And

0:50:08.360 --> 0:50:10.600
<v Speaker 1>this kind of just gets you thinking about the ways

0:50:10.680 --> 0:50:13.920
<v Speaker 1>that people will be will be able to misuse robots

0:50:13.960 --> 0:50:18.520
<v Speaker 1>in ways that guide their behavior in extremely pernicious directions,

0:50:18.640 --> 0:50:23.399
<v Speaker 1>sometimes without the people guiding this misuse necessarily committing any

0:50:23.480 --> 0:50:26.960
<v Speaker 1>kind of identifiable crime. Yeah, Like people are going to

0:50:27.040 --> 0:50:30.200
<v Speaker 1>look for exploits, They're gonna look for ways, they're gonna

0:50:30.200 --> 0:50:32.040
<v Speaker 1>look for cracks in the system. It's it's like within

0:50:32.320 --> 0:50:34.640
<v Speaker 1>with any kind of like a video game system. You know,

0:50:34.640 --> 0:50:36.200
<v Speaker 1>people are just gonna see what they can get away

0:50:36.280 --> 0:50:39.080
<v Speaker 1>with and and just engage in that kind of action,

0:50:39.520 --> 0:50:41.800
<v Speaker 1>sometimes just for the fun of it, right, And sometimes

0:50:41.800 --> 0:50:46.120
<v Speaker 1>that's harmless, but sometimes that's really awful. Yeah, Okay. Next

0:50:46.320 --> 0:50:50.200
<v Speaker 1>category is unforeseen harms. And here's where we start getting

0:50:50.200 --> 0:50:54.920
<v Speaker 1>into the really really interesting and really difficult cases, types

0:50:54.960 --> 0:50:58.799
<v Speaker 1>of harm that are not unavoidable, not a product of

0:50:58.880 --> 0:51:04.080
<v Speaker 1>defects or miss use, but are still not predicted by creators. Uh.

0:51:04.120 --> 0:51:06.200
<v Speaker 1>And so the authors talk about how, in a way,

0:51:06.400 --> 0:51:11.040
<v Speaker 1>unpredictability is what makes AI potentially useful, right like, it

0:51:11.120 --> 0:51:15.640
<v Speaker 1>can potentially arrive at solutions that humans wouldn't have predicted.

0:51:16.040 --> 0:51:19.120
<v Speaker 1>But sometimes it does so in ways that really miss

0:51:19.200 --> 0:51:22.080
<v Speaker 1>the boat and could be extremely harmful if they were

0:51:22.160 --> 0:51:25.840
<v Speaker 1>embodied in action in the real world. Uh, similar to

0:51:25.880 --> 0:51:28.520
<v Speaker 1>the drone example from the circle that we talked about

0:51:28.560 --> 0:51:32.319
<v Speaker 1>at the beginning. But they signed another fantastic example here.

0:51:32.360 --> 0:51:34.839
<v Speaker 1>That's kind of chilling. So I'm just gonna read from

0:51:35.040 --> 0:51:38.400
<v Speaker 1>Limle and Casey here. In the ninety nineties, a pioneering

0:51:38.480 --> 0:51:42.440
<v Speaker 1>multi institutional study sought to use machine learning techniques to

0:51:42.480 --> 0:51:48.000
<v Speaker 1>predict health related risks prior to hospitalization. After ingesting an

0:51:48.080 --> 0:51:52.480
<v Speaker 1>enormous quantity of data covering patients with pneumonia, the system

0:51:52.600 --> 0:51:58.040
<v Speaker 1>learned the rule has asthma X delivers lower risk X.

0:51:58.520 --> 0:52:02.360
<v Speaker 1>The colloquial translation is patients with pneumonia who have a

0:52:02.440 --> 0:52:05.680
<v Speaker 1>history of asthma have a lower risk of dying from

0:52:05.719 --> 0:52:10.360
<v Speaker 1>pneumonia than the general population. The machine derived rule was curious,

0:52:10.400 --> 0:52:13.439
<v Speaker 1>to say the least, far from being protective, asthma can

0:52:13.520 --> 0:52:19.800
<v Speaker 1>seriously complicate pulmonary illnesses, including pneumonia. Perplexed by this counterintuitive result,

0:52:19.880 --> 0:52:23.200
<v Speaker 1>the researchers dug deeper, and what they found was troubling.

0:52:23.800 --> 0:52:27.160
<v Speaker 1>They discovered that quote, patients with the history of asthma

0:52:27.200 --> 0:52:30.640
<v Speaker 1>who presented with pneumonia usually were admitted not only to

0:52:30.680 --> 0:52:33.840
<v Speaker 1>the hospital but directly to the i c U, the

0:52:33.880 --> 0:52:37.640
<v Speaker 1>intensive care unit. Once in the i c U, asthmatic

0:52:37.640 --> 0:52:41.680
<v Speaker 1>pneumonia patients went on to receive more aggressive care, thereby

0:52:41.840 --> 0:52:46.920
<v Speaker 1>raising their survival rates compared to the general population. The rule,

0:52:47.000 --> 0:52:50.320
<v Speaker 1>in other words, reflected a genuine pattern in the data,

0:52:50.719 --> 0:52:55.520
<v Speaker 1>but the machine had confused correlation with causation quote, incorrectly

0:52:55.640 --> 0:52:59.360
<v Speaker 1>learning that asthma lowers risk when in fact, asthmatics have

0:52:59.520 --> 0:53:02.120
<v Speaker 1>much high fire risk. It seems like we've got another

0:53:02.160 --> 0:53:06.960
<v Speaker 1>wormhole here. And here the authors introduce an idea of

0:53:06.960 --> 0:53:12.080
<v Speaker 1>of a curve of outcomes that they call a leptokurtic curve.

0:53:12.560 --> 0:53:15.080
<v Speaker 1>That's a strange term, but basically what that means is

0:53:15.120 --> 0:53:18.319
<v Speaker 1>if you are um, if you're charting what types of

0:53:18.360 --> 0:53:22.880
<v Speaker 1>outcomes you expect from a traditional system like just you know,

0:53:23.000 --> 0:53:28.640
<v Speaker 1>humans looking at data versus a a complex automated system,

0:53:28.719 --> 0:53:32.200
<v Speaker 1>the sort of the tails of the graph with the

0:53:32.280 --> 0:53:35.560
<v Speaker 1>complex automated system will tend to be fatter, meaning you

0:53:35.600 --> 0:53:39.320
<v Speaker 1>get more extreme events in the positive and negative space,

0:53:39.920 --> 0:53:43.320
<v Speaker 1>rather than a you know, a sort of rounder clustering

0:53:43.400 --> 0:53:46.759
<v Speaker 1>of events in the you know, normal operation space, if

0:53:46.760 --> 0:53:51.160
<v Speaker 1>that makes any sense. So, these kinds of unforeseen harms

0:53:51.160 --> 0:53:54.040
<v Speaker 1>are some of the most worrisome types of things to

0:53:54.120 --> 0:53:56.440
<v Speaker 1>expect coming out of robots and AI. But then the

0:53:56.520 --> 0:53:59.600
<v Speaker 1>other one would be systemic harms And this is the

0:53:59.680 --> 0:54:03.000
<v Speaker 1>last category of of harms they talk about. UH the

0:54:03.000 --> 0:54:06.239
<v Speaker 1>author's right quote. People have long assumed that robots are

0:54:06.320 --> 0:54:10.880
<v Speaker 1>inherently neutral and objective, given that robots simply intake data

0:54:11.200 --> 0:54:15.760
<v Speaker 1>and systematically output results. But they are actually neither. Robots

0:54:15.760 --> 0:54:18.520
<v Speaker 1>are only as neutral as the data they're fed, and

0:54:18.640 --> 0:54:21.719
<v Speaker 1>only as objective as the design choices of those who

0:54:21.760 --> 0:54:26.360
<v Speaker 1>create them. When either bias or subjectivity infiltrates the systems

0:54:26.400 --> 0:54:30.279
<v Speaker 1>inputs or design choices, it is inevitably reflected in the

0:54:30.320 --> 0:54:34.319
<v Speaker 1>system's outputs. This is your classic garbage in, garbage out, problem, right,

0:54:35.200 --> 0:54:38.920
<v Speaker 1>They go on. Accordingly, those responsible for overseeing the deployment

0:54:38.960 --> 0:54:44.240
<v Speaker 1>of robots must anticipate the possibility that algorithmically biased applications

0:54:44.239 --> 0:54:47.680
<v Speaker 1>will cause harms of this systemic nature to third parties.

0:54:48.239 --> 0:54:51.320
<v Speaker 1>So UH, an example that's much discussed in this would

0:54:51.320 --> 0:54:55.960
<v Speaker 1>be an AI trained to make decisions about granting loans

0:54:56.360 --> 0:55:00.560
<v Speaker 1>by studying patterns of which loan applicants got their loans

0:55:00.600 --> 0:55:03.359
<v Speaker 1>granted in the past. And a I like this could

0:55:03.440 --> 0:55:06.640
<v Speaker 1>end up manifesting some type of bias that hurts people,

0:55:06.680 --> 0:55:10.239
<v Speaker 1>like a racial bias in its loan assessments, because there

0:55:10.320 --> 0:55:13.560
<v Speaker 1>was already a bias in the real world data set

0:55:13.640 --> 0:55:16.320
<v Speaker 1>that it was trained on. So, in other words, AI

0:55:16.440 --> 0:55:19.840
<v Speaker 1>that is trained on data from the real world, unless

0:55:19.880 --> 0:55:22.719
<v Speaker 1>it is it is explicitly told not to do this,

0:55:22.880 --> 0:55:27.160
<v Speaker 1>it will tend to reproduce and perpetuate any injustices, any

0:55:27.160 --> 0:55:30.680
<v Speaker 1>inequalities that already exist. And the authors here give an

0:55:30.760 --> 0:55:36.680
<v Speaker 1>example that is based on algorithmically derived insurance premiums that

0:55:36.800 --> 0:55:40.040
<v Speaker 1>I think they're talking about auto insurance quote. A recent

0:55:40.120 --> 0:55:44.240
<v Speaker 1>study by Consumer Reports found that contemporary premiums depended less

0:55:44.280 --> 0:55:49.680
<v Speaker 1>on driving habits and increasingly on socioeconomic factors, including an

0:55:49.680 --> 0:55:54.880
<v Speaker 1>individual's credit score. After analyzing two billion car insurance price

0:55:54.960 --> 0:55:59.000
<v Speaker 1>quotes across approximately seven hundred companies, the study found that

0:55:59.080 --> 0:56:03.640
<v Speaker 1>credit scores actored into insurance algorithms so heavily that perfect

0:56:03.760 --> 0:56:07.919
<v Speaker 1>drivers with low credit scores often paid substantially more than

0:56:08.120 --> 0:56:12.440
<v Speaker 1>terrible drivers with high scores. The studies findings raised widespread

0:56:12.440 --> 0:56:15.840
<v Speaker 1>concerns that AI systems used to generate these quotes could

0:56:15.960 --> 0:56:20.160
<v Speaker 1>create negative feedback loops that are hard to break. According

0:56:20.200 --> 0:56:24.080
<v Speaker 1>to one expert, quote, higher insurance prices for low income

0:56:24.160 --> 0:56:27.840
<v Speaker 1>people can translate to higher debt and plummeting credit scores,

0:56:28.040 --> 0:56:31.080
<v Speaker 1>which can mean reduce job prospects, which allows debt to

0:56:31.120 --> 0:56:34.480
<v Speaker 1>pile up, credit scores to sink lower, and insurance rates

0:56:34.480 --> 0:56:37.880
<v Speaker 1>to increase in a vicious cycle. Uh So this is

0:56:37.920 --> 0:56:40.360
<v Speaker 1>kind of a nightmare scenario, right, Like an AI that

0:56:40.520 --> 0:56:45.200
<v Speaker 1>is too powerful and not explicitly protected against acquiring these

0:56:45.200 --> 0:56:49.360
<v Speaker 1>types of biases could create these kind of computer enforced

0:56:49.480 --> 0:56:54.320
<v Speaker 1>prisons in reality, like a machine code for perpetuating whatever

0:56:54.440 --> 0:56:56.960
<v Speaker 1>state of the world, like whatever state the world was

0:56:57.040 --> 0:57:00.920
<v Speaker 1>in when the AI was first deployed, and then entrenching

0:57:01.000 --> 0:57:04.239
<v Speaker 1>it further and further. Yeah, and that kind of thing

0:57:04.320 --> 0:57:07.240
<v Speaker 1>is especially scary because, like, if there's a human making

0:57:07.280 --> 0:57:09.680
<v Speaker 1>the decision, you can you can call up the human

0:57:09.719 --> 0:57:12.000
<v Speaker 1>to a witness stand or ask them like, hey, why

0:57:12.040 --> 0:57:14.600
<v Speaker 1>did you make the decision this way? But if it's

0:57:14.640 --> 0:57:17.120
<v Speaker 1>an AI doing it, you could say like, hey, why

0:57:17.160 --> 0:57:19.520
<v Speaker 1>why is it? Why are we getting this outcome that's

0:57:19.560 --> 0:57:22.439
<v Speaker 1>you know, creating a sort of like cyclical prison out

0:57:22.440 --> 0:57:24.640
<v Speaker 1>of reality, And they can just say, hey, you know

0:57:24.760 --> 0:57:27.400
<v Speaker 1>it's the machine with the machine. You know it knows

0:57:27.400 --> 0:57:29.560
<v Speaker 1>what it's doing. Yeah, and then yes, the machine it

0:57:29.600 --> 0:57:33.280
<v Speaker 1>says I learned it from watching you dad, and you

0:57:33.320 --> 0:57:35.520
<v Speaker 1>have that moment of shame. So I think these different

0:57:35.520 --> 0:57:38.160
<v Speaker 1>categories that they that they bring up are really important

0:57:38.200 --> 0:57:41.160
<v Speaker 1>for helping us kind of sort our ideas into into

0:57:41.200 --> 0:57:44.760
<v Speaker 1>recognizable types for for ways that AI and robots could

0:57:44.760 --> 0:57:46.919
<v Speaker 1>go wrong and could potentially cause harm that you would

0:57:46.920 --> 0:57:50.520
<v Speaker 1>seek legal remedy for. And also they help identify the

0:57:50.600 --> 0:57:53.440
<v Speaker 1>spaces that there's the most worry. I mean, for me,

0:57:53.520 --> 0:57:56.760
<v Speaker 1>I think that would be like those last two cases, right,

0:57:56.800 --> 0:58:00.280
<v Speaker 1>the unforeseen problems and the systemic problems are the ones

0:58:00.640 --> 0:58:03.760
<v Speaker 1>where there's the most real danger I think and the

0:58:03.800 --> 0:58:06.919
<v Speaker 1>most difficulty in trying to figure out how to solve it. Yeah,

0:58:06.960 --> 0:58:10.680
<v Speaker 1>because we we kind of you know, train ourselves for

0:58:11.080 --> 0:58:14.120
<v Speaker 1>to a certain extent and sort of culturally focus on

0:58:14.240 --> 0:58:20.720
<v Speaker 1>the sky net problems, right, the really obvious um situations

0:58:20.760 --> 0:58:23.080
<v Speaker 1>where the robot car veers off the road in a

0:58:23.200 --> 0:58:28.080
<v Speaker 1>dangerous way. But the situations where it is just perpetuating

0:58:28.120 --> 0:58:32.920
<v Speaker 1>what we're already doing, where it's making choices in getting

0:58:32.920 --> 0:58:36.040
<v Speaker 1>from point A to point B that don't violate anything

0:58:36.040 --> 0:58:39.000
<v Speaker 1>we told it, but just are an uninventive and even

0:58:39.080 --> 0:58:42.520
<v Speaker 1>harmful way of doing it. Uh. Yeah, that's that's that's

0:58:42.520 --> 0:58:45.640
<v Speaker 1>harder to deal with. That's a type of misbehavior that

0:58:46.520 --> 0:58:49.520
<v Speaker 1>you can't solve by just having Dan o'harla Hay stand

0:58:49.560 --> 0:58:56.000
<v Speaker 1>up and bellowing behave yourselves exactly. Um yeah, yeah, yeah,

0:58:56.040 --> 0:58:58.160
<v Speaker 1>I mean I can't remember if that even worked. I

0:58:58.200 --> 0:59:00.440
<v Speaker 1>just remember that was one of my favorite moments, and

0:59:00.680 --> 0:59:02.880
<v Speaker 1>that was RoboCop two, right, was it? Yeah, well, I

0:59:02.880 --> 0:59:05.360
<v Speaker 1>mean RoboCop one. I think we're already dealing with this

0:59:05.440 --> 0:59:09.200
<v Speaker 1>problem of like the sort of like weird dynamics of

0:59:09.240 --> 0:59:12.640
<v Speaker 1>machine culpability. When Ed two oh nine like shoots that

0:59:12.640 --> 0:59:15.959
<v Speaker 1>guy five hundred times in the boardroom during the demonstration,

0:59:16.400 --> 0:59:19.000
<v Speaker 1>and then Dan O'Hurley he's response to it is to

0:59:19.080 --> 0:59:21.959
<v Speaker 1>turn to Ronnie Cox and say, I'm very disappointed, Dick.

0:59:24.680 --> 0:59:27.200
<v Speaker 1>But anyway, well, I guess we're running running kind of long,

0:59:27.240 --> 0:59:29.400
<v Speaker 1>so maybe we should call part one there, But we

0:59:29.440 --> 0:59:34.080
<v Speaker 1>will resume this discussion about about robot justice and robot

0:59:34.080 --> 0:59:37.080
<v Speaker 1>punishment in the next episode. That's right, we'll be back

0:59:37.280 --> 0:59:40.480
<v Speaker 1>with more of this discussion in the meantime. If you

0:59:40.480 --> 0:59:42.560
<v Speaker 1>would like to check out past episodes of Stuff to

0:59:42.560 --> 0:59:45.680
<v Speaker 1>Blow Your Mind Uh, and they're definitely worth checking out

0:59:45.680 --> 0:59:48.000
<v Speaker 1>because we have lots of past episodes that deal with

0:59:48.160 --> 0:59:50.920
<v Speaker 1>robots and AI. We have lots of episodes where we

0:59:50.960 --> 0:59:54.720
<v Speaker 1>make RoboCop references, so they're all they're all there, go

0:59:54.760 --> 0:59:56.480
<v Speaker 1>back and check them out. You can find our podcast

0:59:56.520 --> 0:59:58.760
<v Speaker 1>wherever you get your podcasts. Just look for the Stuff

0:59:58.760 --> 1:00:01.640
<v Speaker 1>to Blow your Mind podcast need uh in that feed

1:00:01.680 --> 1:00:04.480
<v Speaker 1>We put out core episodes of the show on Tuesdays

1:00:04.480 --> 1:00:08.080
<v Speaker 1>and Thursdays. Mondays, we have a little listener mail Wednesdays,

1:00:08.120 --> 1:00:11.680
<v Speaker 1>so that's when we do the artifact shorty uh usually,

1:00:12.040 --> 1:00:14.640
<v Speaker 1>and then on Friday's we do Weird House Cinema. That's

1:00:14.640 --> 1:00:17.000
<v Speaker 1>our chance to sort of set most of the science

1:00:17.040 --> 1:00:20.880
<v Speaker 1>aside and just focus on the films about rampaging robots.

1:00:21.640 --> 1:00:24.400
<v Speaker 1>You just thinks. As always to our excellent audio producer

1:00:24.520 --> 1:00:26.680
<v Speaker 1>Seth Nicholas Johnson. If you would like to get in

1:00:26.800 --> 1:00:29.080
<v Speaker 1>touch with us with feedback on this episode or any

1:00:29.120 --> 1:00:31.480
<v Speaker 1>other to suggest a topic for the future, or just

1:00:31.560 --> 1:00:34.040
<v Speaker 1>to say hello. You can email us at contact at

1:00:34.040 --> 1:00:43.880
<v Speaker 1>stuff to Blow your Mind dot com. Stuff to Blow

1:00:43.920 --> 1:00:46.480
<v Speaker 1>Your Mind is production of I Heart Radio. For more

1:00:46.480 --> 1:00:49.080
<v Speaker 1>podcasts for my heart Radio, visit the iHeart Radio app,

1:00:49.280 --> 1:00:52.000
<v Speaker 1>Apple Podcasts, or wherever you're listening to your favorite shows.

1:01:00.520 --> 1:01:07.760
<v Speaker 1>Pretty Toy pop po