Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:11
Welcome to the Sentience Institute podcast,
0:13
and to our 20th episode. I'm
0:16
Michael Deli , Yvo strategy lead and
0:18
researcher at Sentence Institute on
0:20
the Sentience Institute podcast. We interview
0:23
activists, entrepreneurs, and researchers
0:25
about the most effective strategies to expand
0:27
humanities moral circle. Our guest
0:29
for today is David Gun. David
0:32
is an award-winning educator, scholar,
0:34
and author specializing in the philosophy
0:37
and ethics of emerging technology. He
0:39
is the author of over 90 scholarly
0:41
articles and book chapters, and as
0:44
published 12 internationally recognized books including
0:46
the Machine Question, Critical Perspectives
0:49
on ai, Robots and Ethics of
0:51
Remix, Ethics and Aesthetics.
0:54
After remix and Robot
0:56
rights, he currently holds the
0:58
position of distinguished teaching professor in
1:01
the Department of Communication at Northern Illinois
1:03
University in usa. All
1:06
right . I'm joined by David Gun. David, thank you so much
1:08
for joining us on the Sentience Institute podcast.
1:10
Yeah, thank you for having me. Good to be here.
1:12
Great. Great to have you. So , uh,
1:14
I'd like to start with a question , uh, about terminology.
1:17
And I should preface this by saying I
1:19
actually know your answer, I think because I heard
1:21
it in your interview with the Ben Byford , um,
1:24
with Ben Byford on the Machine Ethics Podcast
1:26
in 2020. Uh , but I think it's a
1:28
good place for our listeners to start. So
1:30
why do you call it robot
1:32
rights and not artificial intelligence riots
1:35
sentence , or something else like that?
1:37
Yeah, so it's a really good question, and I
1:39
think I can answer it with three sort
1:42
of ways of, of directing your
1:45
, uh, you know, way of thinking about
1:47
this. First of all, it's
1:49
just, it's an alliterative statement, right?
1:51
Robot rights. So just the alliteration
1:53
sort of makes it easy to say artificial
1:56
intelligence rights seems a little clumsy, and
1:58
as a result, robot rights is pretty easy
2:00
for a person to sort of rattle
2:02
off, and it has a alliterative
2:05
, sort of poetic feel to it because
2:07
of the way it sort of rolls out of your
2:09
mouth and , uh, sounds when you hear
2:11
it. The second reason is that
2:13
this terminology is not mine. Um
2:16
, this is terminology that came to me from other
2:18
people who preceded me in this work.
2:21
Either , uh, legal scholars or
2:23
people in philosophy or in
2:25
the area of artificial intelligence and ethics, but
2:28
they were the ones that began using this terminology.
2:31
And so my engagement
2:33
with it was to sort of pick up the thread and
2:36
carry it further. And since that was
2:38
the terminology that they had already employed, I
2:40
sort of inherited that terminology. Lastly,
2:43
and I think most importantly, the
2:45
difference between the word robot and
2:48
artificial intelligence is something
2:50
that we oftentimes, you know, struggle with and
2:52
try to figure out where to draw the line. And is
2:54
it cloud based application a robot? Is
2:56
it just ai? You know, how do we,
2:58
we sort of sort these things. The
3:01
important thing for me is that AI
3:03
is the result of a academic
3:07
workshop held in the mid 1950s in
3:10
Dartmouth College. Robot, on the
3:12
other hand, is the result of science
3:14
fiction. It comes to us from the 1920
3:17
stage play rur , and
3:19
it was formulated and
3:22
really was a reuse of a
3:24
check term robota , meaning worker
3:26
or slave laborer , uh, by
3:28
Carl Chap . And already in that
3:30
play, which not only gave us the idea of
3:32
the robot, but the word robot, the
3:35
robots are already subjugated to human
3:37
masters, and there's this uprising.
3:39
So not only does that set the sort of template
3:41
for future science fiction, but
3:44
it also gives us this notion of the robot
3:46
as an enslaved or
3:49
a servant type , you
3:51
know, figure or individual. And
3:53
so the robot rights , uh,
3:56
idea sort of fits in that pattern,
3:58
beginning with topics play, and
4:00
the way in which that has developed not only in
4:03
subsequent science fiction, but also in subsequent
4:05
writings on the legal and
4:07
the moral aspects of these
4:09
technologies and the way they connect with us.
4:11
Yeah, that , that makes sense. I
4:14
, I think for the rest of the interview, I'll probably use
4:17
robots and maybe have that
4:19
, um, apply to AI as well , uh,
4:21
just as , as shorthand, but in
4:24
a sense, they're not really interchangeable.
4:26
I , I feel, is that, does that make sense to
4:28
you? Because AI sort of , um, brings
4:31
, uh, an image of , uh,
4:33
an intelligence that's not , that's not necessarily tied
4:35
to a physical body, whereas robot seems to
4:37
imply it's tied to a physical body. Does that
4:39
sound about right?
4:41
It sounds about right, but I , I will say
4:43
that this idea that the
4:45
intelligence is disembodied, it's almost a
4:47
kind of transcendentalist way of thinking that's
4:50
almost religious ouris are
4:52
embodied, right? Even the cloud has a body, it's
4:55
a server connected to, to wires and
4:57
to fiber network cables and
4:59
things like this. So there is an embodiment
5:01
even for the so-called disembodied
5:04
ai. It's just that that body doesn't
5:06
look like our bodies. And so I
5:08
find this to be a useful distinction, the embodiment
5:11
distinction, but I also find it a bit troubling because
5:13
it leads us to think that the AI has
5:15
this kind of transcendental feature to it, and
5:18
it really doesn't, when we talk about the, you
5:20
know, resources of the earth about power,
5:23
about , uh, environmental impact and
5:25
all these other things that AI certainly
5:27
contributes to carbon and to climate
5:29
change. And that has to do because
5:31
of the way it is embodied and where it is
5:33
embodied.
5:34
Yeah, yeah, that makes sense. Thanks for mentioning that. So
5:37
in 2018, you wrote a paper called
5:39
The Other Question, Can Robot, Sorry.
5:42
Can and Should Robots Have Rights? And
5:44
the other question here being in reference to
5:46
the main question most people focus on when
5:48
they think about AI and robots, which is about
5:50
how they affect us, for example
5:52
, um, AI safety being a hot topic at
5:55
the moment, that's, that's mostly about how an
5:57
AI affects humans. And the
5:59
, the other question, I guess is more focused on the
6:02
interests of the robots themselves. So
6:05
you wrote that there's an important difference between
6:07
the can and the should here. So can you
6:09
start by talking about that?
6:10
Right. So I'll make two distinctions, which I
6:12
think will help sort this out. Uh , in more
6:14
philosophy, we distinguish the moral
6:17
agent from the moral patient. A
6:19
moral agent is someone who can act in
6:21
such a way that is either good, bad, or
6:24
is morally culpable or
6:26
praiseworthy, whatever the case is, a moral patient
6:28
is the recipient of that action. Now,
6:30
in our world, we are both moral agents
6:33
and moral patients. We can do good and
6:35
bad, we can suffer good and bad,
6:37
but there are some things that are moral patients
6:39
and not moral agents animals, for
6:41
example, we don't hold the dog responsible
6:44
for barking at the postman, but we do hold
6:47
the postman responsible for kicking and
6:49
injuring the dog, right? And the dog therefore
6:51
can be the recipient of an action that is either
6:53
good or bad. And as you said, I saw
6:55
a lot of people putting a
6:58
lot of the research they were doing on
7:00
the side of agency. And the question was,
7:02
how can we ensure that these devices,
7:05
these tools, these instruments, these
7:07
artifacts are employed in our
7:09
world in a way that has the right outcomes,
7:12
that doesn't disadvantage people, that doesn't create bad
7:14
outcomes, whatever the cases , that's an agency question.
7:17
My question, the other question was, well, okay,
7:19
that's great, but what is
7:21
the status of these things? What is , what is
7:23
the social moral, legal position of
7:26
these devices that are increasingly intelligent and
7:28
socially and interactive? And
7:30
how do we grapple with the moral patency
7:33
question of these machines? And then when
7:35
go to the next stage is the
7:37
can should distinction. And this
7:40
is really derived from David Hume. David Hume
7:42
says, You know, you cannot derive ought from is
7:44
, it's a very famous item in
7:46
, in David Hume's thinking, but a lot of people have picked it up
7:48
and developed it to us in, you
7:50
know, since that time, the can and
7:53
should question is really this, Can
7:55
robots have rights? Yeah. All you have
7:57
to do is make a law that says robots have rights.
8:00
Now, should you do that is another
8:02
question. So the ability to
8:04
do so maybe is entirely
8:06
feasible and very easy to
8:09
answer that question. The moral question,
8:11
the should question is a little more complicated. And
8:13
I think that's where we get into the weeds on , um,
8:16
how we want to shape the world that we live
8:18
in and how we integrate these things alongside
8:21
us in our social reality.
8:23
Mm-hmm . So I'll
8:26
jump the gun a little bit, and just because you mentioned it and
8:28
ask , uh, what are some reasons
8:30
why robots shouldn't have rights? Uh
8:32
, what are what are some arguments one might use?
8:34
So one of the strongest arguments is that they
8:37
machines, right? They're not people,
8:39
they're not human beings, they're not animals. They
8:42
should, you know, they're just artifacts and therefore they
8:44
are things. We may have this
8:46
distinction that comes to us from the Romans , uh, from
8:49
Gaas in in particular, that, you
8:51
know, actions that are moral or
8:53
legal in nature have two kinds
8:56
of objects or two kinds of , uh, entities.
8:58
There is either persons or things. And
9:01
so in the category of persons, we put you
9:03
and I and we put corporations
9:06
and we put maybe some animals, but
9:08
we don't generally put technological objects. Those
9:10
are things that's a strong argument based
9:12
on this ontology that we've
9:14
inherited from the Romans and the way in which
9:16
our legal structures especially have
9:19
operationalized this way of dividing things into
9:21
persons or property. Another argument
9:23
is that if we give
9:26
robots some kind of moral or
9:28
legal standing, we have complicated
9:31
our legal system in ways that
9:33
goes beyond what maybe we
9:35
would like to handle, and
9:37
that it maybe doesn't lend anything
9:40
very useful to the way that
9:42
we decide these relationships. Those
9:45
are usually the two big ones.
9:47
Mm . When you make those arguments
9:49
or when , when you put those forward, it makes me think
9:51
of , uh, these are some of the arguments I've heard in
9:53
relation to why when people say non-human
9:56
animals shouldn't have rights, it , it
9:58
complicates the legal system. Um, it
10:00
, uh, it's, yeah. So that , that , that
10:02
sounds familiar, but what are , what
10:05
are the , um, best arguments for why robots should
10:07
have rights?
10:08
So there's a number of arguments, and I
10:10
don't wanna try to be exhausted here, but let
10:12
me cover some of the important ones
10:14
that have been circulating. Um, the
10:17
literature in this field has absolutely exploded
10:20
, um, in the last decade. And
10:22
when I started working on this back
10:24
in the early, well,
10:27
mid , mid two , you know, 2006 I started First
10:29
book comes out in 2012. So in those early
10:31
years, it was really easy to keep track of
10:34
who was arguing what, because the number
10:36
of different arguments and circulation were pretty manageable.
10:39
By the time I get to robot rights in 2018,
10:41
this thing is spinning out of control
10:43
because a lot of people find
10:46
reason to engage the question and
10:48
to deliver their own sort of response
10:51
to it. So let me just hit a few
10:53
reasons why we might want to
10:55
do this. One is directly derived
10:58
from what we've learned in the animal
11:00
rights experience. So in
11:02
animal rights, we know Jeremy , Jeremy Bentham
11:04
really , uh, is the pivot, right? And he said it's
11:06
not, can they think, can they reason, but can
11:08
they suffer? Are they sentis ? And that
11:11
opened up the moral circle to include
11:13
things that had been previously excluded.
11:15
It had been, up until that point, a very
11:17
human-centric kind of moral universe.
11:20
And when we start to engage
11:22
in the animal question, we widen
11:24
that circle to include other
11:26
creatures that are non-human. And the reason why
11:29
we included animals in the moral circle,
11:31
whether you're following Peter Singer or
11:33
Tom Regan , or one of the other innovators in
11:36
animal rights, is because of
11:38
Sentis , because of the experience the
11:41
animal has of pain
11:43
or pleasure. And you can see just recently
11:45
with Blake Lemo , he was
11:47
talking to Lambda, and Lambda, he
11:50
said, is sentient , which led him to believe
11:52
that Lambda needed to have rights protected
11:54
for it because it was another kind
11:56
of sentient creature. So we use sentience
11:59
as a benchmark, and the question is, how
12:01
can you tell whether an AI or a robot is
12:04
sentient ? Well, that takes you back to the tour test
12:06
because you can't necessarily look
12:08
inside and know exactly what it's doing. You
12:10
can know some of what's going on, but
12:12
really what Blake Lamoin did is learn
12:14
from the behavioral experience that
12:16
it was exhibited to him in the conversational interactions
12:19
that he had with Lambda. So that's
12:21
one reason why people argue for giving robots
12:24
rights, this notion that they'll be at
12:26
some point either sentient or conscious
12:29
or some of these other benchmarks that make something
12:31
available to having standing
12:34
status and need for protection. Another
12:37
argument, and this comes from Kate Darling, who I
12:39
think was really innovative in this area by using Ks
12:41
indirect , uh, duties argument , uh,
12:44
Kant , unlike Bentham, was no animal rights advocate. Kant
12:46
thought animals were just mechanisms like Da Hart
12:48
did. But he argued you should not hurt animals, because
12:51
when you do so, you dease
12:53
yourself, you are corrupting
12:56
your own moral character. You're corrupting
12:58
your own moral education and
13:00
providing a bad example for other people. And
13:03
so indirectly, you're harming somebody
13:06
else if you harm an animal. And
13:08
Kate Darling says, you know, this is one way
13:10
of thinking about why we don't want to harm
13:12
the robot because of the example
13:15
it sets, because of the , uh,
13:17
way in which it could corrupt our own
13:19
moral characters. And she
13:21
uses the constant direct duties argument to make a
13:24
case for the rights of robots as
13:26
a way of protecting our social mechanisms
13:28
, uh, the way that we relate to each other, either morally
13:31
or legally. A third argument
13:33
for doing this, and this is more
13:35
in line with where I take it , um, in
13:37
my own research, is that the
13:40
properties, ascensions and consciousness have
13:42
traditionally been really good benchmarks for
13:44
things like animal rights and , and items like that.
13:46
But I come out of environmental ethics and
13:49
in environmental ethics, you don't
13:51
harm a mountain, dirt
13:54
does not feel pain. A waterway
13:56
does not experience this pleasure. Nevertheless,
13:59
these are part of our
14:01
integral experience on this
14:04
planet, and we have responsibilities
14:06
to the other entities that occupy
14:09
this fragile planet with us. And I think climate
14:11
change is a really good example of
14:13
how we can screw that up if we assume
14:16
wrongly, that these things are just raw materials
14:18
that we can utilize to our benefit and
14:21
artifacts may also play a role
14:23
in our social world in a way that we
14:25
need to think more creatively about
14:27
how we craft our moral
14:30
responsibilities and our legal responsibilities
14:32
to others. And that is
14:34
what I'm calling a relational approach
14:36
to moral status. That it's
14:38
not what the thing is, but how
14:40
it stands in relationship to us and
14:43
how we interact with it on
14:45
a scale that treats
14:47
these other entities as
14:49
fellow travelers, as kin . And
14:52
how we can come up with ways
14:54
of integrating not just
14:56
the natural objects of the environment, but
14:58
also the artifacts that we create in
15:01
our moral universe, in our legal
15:03
universe in a way that makes sense for
15:06
us, but also for our future
15:08
generations and for the kind
15:10
of environment and the kind of world
15:12
we want to occupy.
15:14
That was great. Thanks David. There's , there's
15:16
a lot. Uh , I wanted to, to , um,
15:18
mention from that first , uh, that you
15:20
mentioned Blake Lemo and Lambda. Uh , we actually
15:22
spoke to , uh, Thomas Metzinger in our last
15:25
podcast episode that came out last week, and we
15:27
, we talked about that topic as well. Um,
15:29
you, so the second,
15:31
the second argument you made about if
15:35
we treat robots in
15:37
, in a bad way, in the same way that if we treat animals
15:39
in a bad way, that might have repercussions
15:42
for , um, for how
15:44
it affects us and how it affects humans in general. Uh
15:46
, now it sounds like even
15:49
if there is, even
15:51
if robots are not currently sentient , even if
15:53
robots maybe even can never be sentient, that
15:56
would still remain an argument in favor
15:58
of giving robots rights. Uh , does , does
16:00
that sound, does that sound right? Um , yes.
16:02
Yeah. Great. Yeah, sure.
16:04
And , uh, the , the last point you mentioned
16:07
it again, it seems to be coming back
16:09
to , uh, how , um,
16:12
how the way we interact with robots
16:15
affects , affects humans. Uh , I
16:17
know you've spoken about the case of the Wongan Nui
16:19
River in New Zealand being granted legal person
16:21
who writes as an analogy. And I think you said
16:23
something like, it's not that people are arguing that
16:25
the river is sentient , it's just
16:27
that that's a tool in our current legal system , uh,
16:30
to get protection. And then it's for instrumental
16:32
purposes or for , um, orsic
16:36
purposes. It's how , uh, that
16:38
, uh, affects humans. Uh
16:40
, so giving robots, right, giving
16:42
robot rights, sorry, giving
16:44
robots rights now might therefore
16:47
be an instrumental tool in the same way. Yeah,
16:49
I , I think a really good example is a recently
16:51
12 states in the US made
16:55
some legislative , uh, decisions
16:58
and , and put into act , uh, some laws that
17:01
try to deal with these personal
17:03
delivery robots that are being deployed on the city streets.
17:06
And they decided that for
17:09
the purposes of figuring out
17:11
right of way and who can go
17:13
in the crosswalk and who can't go in the crosswalk
17:15
and things like this, it made sense to
17:17
extend to the robot the rights of a pedestrian.
17:20
Now, that's not making a decision about robotic
17:22
personhood. That's not making a , you
17:24
know, a distinction that would grant personhood
17:27
to the robot. It's just saying rights
17:30
are the way that we figure out how to integrate
17:32
things in situations
17:34
where you have competing claims, powers, privileges,
17:36
or immunities coming from different actors
17:39
in the social environment. And
17:41
one way in which we negotiate this is
17:43
by granting rights, by giving something a
17:45
right, a privilege, a power, a claim, or
17:47
an immunity. It has to be at least , uh,
17:50
respected by some, by someone else in
17:52
that social environment. Yeah . So this
17:54
is just a tool we have to try to
17:56
develop ways of responding
17:59
to these challenges that allow
18:01
us to occupy space and
18:04
work with. And alongside these
18:06
various things,
18:07
I , I think I a pretty clear example of
18:09
that is , um, how we corporations have
18:11
rights and they , they're clearly not , um, senti , they're
18:13
clearly not , um, persons, but in , in the eyes of the
18:16
law, often they are treated like they are , uh,
18:18
persons in , in a lot of
18:19
Ways. Can I say the , Yeah . The real trouble here.
18:21
I think the , the real point of, of debate
18:24
and contention is the fact that we're
18:26
trying to work with two legal categories that
18:28
are mutually exclusive person or thing.
18:31
And as I said before, this comes
18:33
from 2000 years ago when Gaas developed
18:36
this distinction in his own , uh,
18:38
legal thinking. And our western legal
18:40
systems have worked with us for a long time, and
18:43
it's worked pretty well, but that's why the corporation got
18:45
moved from a thing to a person because we wanted
18:47
to be able to sue it. We wanted it to
18:49
be able to stand before
18:51
the court as a subject and
18:53
not just an object. And so this
18:56
whole debate is about how do we negotiate this
18:58
distinction person, On the one hand that
19:01
is an subject before the
19:03
law, an object or thing on the
19:05
other hand, and it may be the case that we just
19:07
need a better moral ontology, we just
19:09
might need something that gives our
19:12
legal system a little more latitude with
19:14
regards to the kinds of entities that
19:16
can occupy these
19:19
kinds of , uh, positions in our world.
19:20
Yeah. When , when it comes to say,
19:23
non-human animal rights, I do like Peter Singer's
19:25
take on what that might look like. It's, it's
19:27
not that we're asking for non-humans to have the
19:29
right to, to drive or to vote, for example,
19:31
We're just asking that , um, their interests are
19:34
considered they're similar interests. Um, for
19:37
example, there , uh, the right
19:39
to not be harmed. Uh
19:41
, so yeah. With that in mind, do you have any
19:43
thoughts about what robot rights might look
19:45
like? Um, or what, what
19:48
perhaps one of your most ideal scenarios
19:50
might be for how that might look
19:52
in practice? Uh , you've sort of talked about that a little
19:55
bit, I guess. Um, but also just given that
19:57
, uh, robots may have very different interests
19:59
to us, let's say in the case where they do
20:01
become sentient. And I'd just like to nudge you as well to
20:03
maybe mention , uh, camel , uh,
20:06
maax paper, humans, Neanderthals, robots
20:08
, uh, and rights. I think that
20:10
that seems relevant here as well. That's talking
20:12
about moral patients, US agencies . So , um,
20:15
yeah, please, please talk about that a little bit.
20:17
Yeah. So, you know, this is
20:19
where I think the analogy with animal
20:21
rights starts to, if not
20:23
break down , at least reach a
20:25
limit. Mm-hmm . <affirmative> animals, I
20:27
think we can say have interests and we
20:29
can guess pretty well what
20:32
they are, even though it's guesswork , we can
20:34
pretty much sort of figure out pretty
20:37
well for ourselves, you know, what the dog
20:39
has as an interest, food, a walk, whatever
20:42
the case is, right? Our current technologies
20:44
don't have interests, and if we're
20:46
looking, looking for interests, we might be
20:49
barking up the wrong tree at some point in
20:51
the future, that's possible. But
20:53
I don't want to hold out this
20:55
expectation that we should wait for that to happen before
20:58
we engage these questions. Um
21:00
, if that does happen, then we maybe finding
21:02
ourselves having to deal with this
21:04
, uh, analogy to the animal much more
21:07
directly. But I think for now, the
21:09
real issue is we have interests and
21:11
these objects are in our world, and
21:14
we need to figure out how to integrate
21:16
them in our moral and legal systems
21:18
in a way that makes sense for us. And
21:21
just responding to these
21:23
devices as if they were tools or instruments
21:26
doesn't seem to work.
21:28
In other words, the robot is sort of in between
21:30
the thing and the person, and it resists
21:33
both reification and personification.
21:36
And that's the problem, right? That's, that's
21:38
the challenge and the opportunity before
21:40
us. How do we scale existing
21:43
moral and legal systems that work
21:46
with these two categories for something
21:48
that seems to resist one or
21:50
the other? And I think what
21:52
Camille has done in his work on
21:55
, uh, the subject of moral patency is
21:57
really instructive because he's saying,
21:59
We're not looking for solitude. We're
22:01
not looking for robots to be like human
22:03
beings and therefore have those same rights
22:06
as a human being would have. Robot
22:08
rights are not the same thing as the , as a
22:10
set of human rights. Human rights are very specific to
22:12
a singular species. The human being
22:15
robots may have some overlapping powers,
22:17
claims, privileges, or immunities that
22:19
would need to be recognized by human beings,
22:22
but their grouping or
22:24
sets of rights will be perhaps very different.
22:26
And I think Camille's point is
22:28
that difference actually matters here. And
22:31
that if we're looking for similitude, we
22:33
are going to , uh, actually paint
22:36
ourselves into a corner on the subject
22:38
where , uh, we'll be really missing
22:40
what's important. And I think the
22:43
focus on difference and how to
22:45
integrate what is different into
22:47
, uh, our current way of
22:49
thinking about our relationships to other
22:52
kinds of things will actually help.
22:54
And again, I think environmental ethics is a really
22:56
good guide here, because
22:58
we don't want to think about
23:00
the water like us. It's
23:03
not like us, right? What is the
23:05
water to us and why is
23:07
it different? And how is that difference important to
23:10
our way of engaging it
23:13
and living alongside it?
23:13
Yeah. So to , to go back to that paper,
23:16
and there's one one point I found , uh, kind of
23:18
interesting. Uh , so they used
23:20
, uh, uh, Neals as
23:23
, um, as an analogy, I guess for, for
23:25
robots in that there seems
23:27
to be some evidence I might , I'm not really familiar.
23:29
I just , um, going off that paper that
23:32
, uh, ne the tools might not have
23:34
, um, uh, moral agency
23:36
, uh, per se . They might have moral agency
23:38
, uh, correct me if I'm,
23:40
I'm mistaken here, but they , they're arguing that , uh,
23:44
it , it might , um, we
23:46
might want to treat Neos if they were in our
23:48
current society, in the
23:51
legal system, treat them more as say
23:53
we would a human child in that , um,
23:55
what we do to them , uh, matters
23:58
and what, but we might not necessarily hold them
24:00
, um, accountable or to blame for
24:02
the actions they do , uh, to us. Does
24:04
that , um, sound about right and is , if so,
24:06
is that a reasonable analogy for how we
24:09
might treat robots in a legal system?
24:10
Right. So lemme say , uh, two things.
24:12
One, a disclaimer. I know very little about Neanderthals
24:15
<laugh> , so I'm not , I'm not gonna speak directly to the same page.
24:17
Yeah , we're on the same page there . Um , so I can't
24:19
really enlighten you in any appreciable
24:22
way in that regard. Um
24:24
, but I will say, and this again, I I'm
24:26
gonna go back to environmental ethics. Um, Thomas
24:29
Birch, who was an environmental ethicist , um,
24:31
said that, you know, what we're talking about
24:33
is power. When you talk about widening
24:36
the moral circle or widening
24:38
the inclusion of what is on
24:40
the inside and what's on the outside , um,
24:43
someone in
24:45
the inside decides whether or not to expand
24:47
the circle, right? And that is
24:49
a power relationship where those in
24:51
power extend to those without
24:53
power inclusion. You
24:56
can see this already in previous rights
24:58
, uh, expansion, Mary Walstone
25:00
Craft , who wrote the vindication of the rights of women,
25:03
had to pitch her argument to the men
25:05
who were in power and had the right
25:07
to them because they were the ones who
25:09
could expand the circle to include women in
25:12
moral and legal consideration. The
25:15
same with animals, right? In order to
25:17
include animals, someone had to pitch
25:19
an argument on behalf of the animals, but
25:22
they were inside the circle to begin with, otherwise
25:24
they would not have been able to make the argument
25:26
for expanding that circle. And
25:29
I think this is the same thing we see playing
25:32
out with, you know, rights expansion beyond
25:34
, um, even animals. That
25:37
this is a dynamic that is
25:39
very much related to power and politics,
25:42
and how this plays out is really
25:44
something that is in our hands because
25:46
we're the insiders. We're the ones
25:48
who make these decisions. So how
25:51
robots get integrated into our moral and
25:53
legal systems is entirely ours
25:55
to decide, and therefore, we
25:57
need to engage this responsibly
25:59
in ways that really adhere to
26:02
our moral standards and protect our
26:04
futures.
26:05
You've, you've said that to change the topic a
26:07
little bit , uh, you've said that AI ethics , um,
26:10
is often dominated by , um, western thinking,
26:12
and that expanding the dialogue to include other ways
26:15
of thinking, like, for example, indigenous or
26:17
aism could be useful. In
26:19
some of our research at Sentience Institute, we,
26:22
we found that people who reported more
26:24
belief in , uh, that artificial
26:26
beings like ais and robots, if
26:29
they have reported belief that those
26:32
entities can have spirits, they also tended
26:34
to extend more moral consideration
26:36
to them, which , um, doesn't sound that
26:38
surprising, I guess, but it's
26:41
an example of how maybe some other ways of
26:43
thinking might actually be , um, beneficial
26:45
to bring into this discussion. So do
26:47
you have any other examples of how, say
26:50
bringing indigenous animism or
26:52
other ways of thinking into the AI
26:54
ethics conversation might be useful?
26:56
Yeah, so lemme just say that a lot of the
26:58
AI ethics , uh, discourse has
27:01
been distinctly western, right? We've used
27:03
consequentialism, we've used deontology,
27:05
we've used virtue ethics, we've
27:07
used traditions that are very much grounded
27:10
in a Western European Christian sort
27:12
of tradition. And there's nothing wrong with
27:14
that except that we've gotta recognize that that's
27:16
not a universal position, right?
27:18
That's very particular. And for people
27:21
who live in the global north and have
27:23
grown up with these philosophical and religious traditions,
27:25
it may make sense, but the
27:28
rest of the world looks at things from
27:30
different perspectives and does
27:32
things , um, that do not necessarily
27:35
track with what comes out
27:37
of a Western experience. And
27:39
so, I think you're exactly right. There's ways in which we can
27:41
look beyond , um, our own
27:43
way of thinking about these matters and do
27:46
so to help inform this in a more
27:48
global perspective and draw on
27:50
a wider range of human wisdom as
27:52
a way of developing responses to this. Now,
27:55
I'll caution, we gotta be careful here because
27:57
this could turn into Orientalism, right? This
27:59
is one of the premier sort of
28:01
colonialist kinds of gestures. You go
28:03
out to the other and you take from them
28:05
what you think is gonna help you , um,
28:08
in your own endeavors. And we've
28:10
gotta protect against that kind of gesture.
28:12
It's not about going
28:15
and colonizing these other ways
28:17
of thinking , um, in order to mine from
28:19
them some sort of insight that we lack
28:22
in our way of doing things. So
28:24
it's about learning, it's about engagement
28:26
in order to be , um,
28:29
students of other ways of thinking
28:31
and to learn from these other
28:33
traditions how to see and engage
28:35
the world in ways that will be different
28:37
from what we may have grown up with and different from
28:40
the standard practices that we have
28:42
from our own traditions. So
28:44
I'll mention , uh, just a couple things that I think
28:46
are useful here. One
28:48
is, I think African philosophies like Ubuntu
28:50
. Obviously Ubuntu is not one
28:52
philosophy, it's a, a collection
28:55
or a constellation of different
28:57
philosophies, but it is a much more
28:59
holistically oriented and less individualistic.
29:02
Whereas Decart said, I think therefore
29:04
I am the philosophers
29:07
arguing and working in the abub to tradition
29:09
says, you know, I am because we
29:11
are, And it comes much more out of a
29:13
communal kind of relationship to
29:16
a wider perspective on the world.
29:18
And I think that can help us, because
29:21
I think a lot of the work that is done in
29:23
a AI ethics and in
29:25
even the robots rights literature
29:27
tends to be very much focused on a Cartesian
29:30
subject that is sentient , that is conscious, and
29:32
that becomes the unit of analysis. If you
29:34
look at things from a more holistic, communal perspective,
29:37
we're looking at it then in a more relational approach
29:39
that I had described earlier. Another
29:42
tradition I think can be really useful
29:44
is by looking to indigenous
29:47
, uh, epistemologies and cosmologies.
29:49
And again, there is no one indigenous
29:52
epistemology. There are a plurality, a
29:54
multiplicity, because they're very different across
29:56
the world. But there are ways in which
29:59
our very idea of rights is
30:01
already a western concept, right? This idea
30:03
of God given rights to the individual. And
30:06
that's very Christian, it's very European, it's very
30:08
modern. And the
30:10
pre-modern sort of indigenous ways of thinking
30:12
about these things look at not rights.
30:14
They don't have that concept yet. They
30:16
talk about kinship relationships and
30:19
how do we build kin with our machines?
30:21
How do we exist alongside
30:24
these entities that are
30:27
our tools, our servants,
30:29
our instruments that doesn't turn
30:31
them into a slave, that doesn't turn them into
30:34
something that is beholden
30:36
to us. And I think kinship relationships as
30:39
developed in a lot of indigenous traditions can
30:41
be a nice way to sort of complicate
30:43
the rights literature that we often bring
30:46
to bear on these questions. And
30:48
then the third thing I will say, and this
30:50
comes outta Confucianism and some research that
30:52
some Confucian scholars have done recently. Instead
30:55
of talking about robot rights, r i
30:57
g H t, they talk about robot
31:00
rights, r i t e s,
31:02
that it's idea of a ritual of
31:04
a performance, and that the
31:06
robot is engaged with us alongside
31:09
us in performative activity. And
31:11
as a result, they are engaging us
31:13
in rights of social interaction
31:16
and that we should put the focus not on
31:19
rights as our I G H
31:21
T, but writes r i t
31:23
e s as a different way of sort
31:26
of shifting the focus from this
31:28
individual possession to a communal
31:30
performance.
31:31
Yeah, that's, that's interesting. Do you have any examples
31:33
of how that might look in practice? What
31:36
, what would that entail doing?
31:37
So this is what I've tried to develop, especially
31:39
with this relational turn , um, concept
31:42
that I, along with Mark Kleberg have
31:44
really been formulating and researching
31:47
for the last decade or more. This
31:49
idea is not ours alone. It comes out
31:51
of environmental ethics. It comes out of the sts
31:53
feminist , uh, ethics , uh, like Karen Barard
31:56
and Rosiere . Um
31:58
, but it's this idea that we need to begin
32:00
to think about our more relationships
32:03
as maybe taking precedence
32:05
over the individual moral , uh,
32:08
entity, and that we are all
32:10
born alongside others, and that
32:12
we are already in that relationship
32:15
prior to this extraction
32:18
of our sort of identity of ourselves.
32:21
Um , so it sort of works counter to the
32:23
Cartesian way of thinking about being in the world
32:25
where Decart is sort of isolated
32:28
from others and then has to go out to
32:30
others and figure out his , uh,
32:32
responsibilities to others. This way of thinking
32:35
is always already responsible
32:37
to others, and that the, the individual is
32:40
a product of that sort
32:42
of interaction. But , uh,
32:44
yeah, that's , uh, that's a life's work right there.
32:46
Sure, sure. Thanks. Do
32:49
you have any thoughts about what a human interaction
32:51
might look like in the future? We've talked a bit about
32:54
the , the legal context , um, but there's, there's,
32:56
there would be a lot of aspects of, of
32:58
interaction. And , um, I guess this
33:01
, you could answer this in the long term where,
33:04
as you say, when at some point in the future,
33:06
robots , um, become likely sentient
33:08
, uh, but there's also the short term answer.
33:10
I mean, we have human robot interaction now that's
33:12
not necessarily , um, related to robots
33:15
being sentient. So what, what, is
33:17
there anything that we can expect in the
33:19
future? How much can we, One thing I
33:21
wanna talk about as well is how much can we learn from science fiction? Um
33:24
, how much of that is lessons
33:26
about what we might see and how much of that is just
33:28
me a fantasy?
33:29
Yeah. No , this is a really important question because
33:31
I think sometimes we think that the robot invasion
33:33
is something from the future, right? We're
33:35
waiting for it. The robots are gonna rise up, or they're
33:37
gonna descend from the heavens with guns and,
33:40
you know , bombs and they're gonna attack us. And that's
33:42
, that's a science fiction scenario. I
33:44
think the robot invasion is way less
33:47
exciting, way less dramatic, even
33:50
moon mundane, it's like the fall of Rome.
33:52
We invite these things into our world, and
33:55
over a couple hundred years, we wonder where the robots
33:57
came from , um, because they have infiltrated
33:59
us in very slow movements
34:02
of, you know, our decisions to use
34:05
a device here to use of a device there.
34:08
So I think we need to look not necessarily
34:11
at the big picture, long
34:13
term kinds of questions, but I wanna look more
34:15
immediately, where are we at
34:17
right now? Like, what is happening
34:20
in our relationships to these devices
34:22
that is maybe of interest to us in
34:25
changing our social relationships
34:27
with each other in the process? So
34:30
one thing we've seen recently , um, as being
34:32
reason for both concern, but also of
34:34
interest is children saying, Thank you to
34:36
Alexa. Now, that's weird. We don't
34:38
say thank you to objects, right? We, we,
34:41
you know, don't say thank you to
34:43
our automobile for getting us around town. We
34:45
say thank you to persons . Some people do, Yeah , mostly
34:48
not. But you know, when we say
34:50
thank you to persons, right, and yet
34:52
the abilities of these very
34:55
simple digital assistance to
34:57
use language brings us into
34:59
a social relationship where
35:01
we think that we need to be polite to
35:04
the object. And there's nothing necessarily
35:06
wrong with that. There's reason
35:08
to think that that is part of what makes
35:11
us social creatures and
35:13
that we need to be really concerned
35:15
with not only , um, what
35:17
that artifact is, but how we
35:19
engage it, how we respond to
35:21
it. I think sometimes people try to write this
35:23
off as anthropomorphism. They say, you know,
35:25
this is the anthropomorphism, anthropomorphism
35:27
is a dirty word because we shouldn't
35:30
be doing that. Um, I think anthropomorphism
35:32
is not a bug. It's a feature . It's a feature of human
35:35
sociology. We do it to each other, other,
35:37
we do it to our animals, and we do
35:39
it to our objects. So it's not a matter of
35:41
yes, no, with anthropomorphism, it's
35:43
not a binary. It's a matter of
35:45
careful and informed
35:48
management. How do we want to manage
35:50
the anthropomorphism that
35:52
we are developing and designing in
35:54
the process of creating these things? And
35:57
I don't know that we have answers to those questions,
35:59
but I do know we have lots
36:01
of ways of engaging in this question.
36:04
Cause we not only have the example of talking to Alexa
36:06
and saying, thank you. We have robot abuse studies
36:08
in which people find it very , uh,
36:11
disconcerting and problematic to , um,
36:13
harm something that they're
36:15
told as just like a toaster. Nevertheless,
36:18
it's social interactivity makes it
36:20
very difficult to do these things. We
36:22
can already see in very
36:25
rudimentary robotic and AI
36:27
systems, ways in which we are accommodating
36:29
ourselves to these objects and
36:32
bringing them into our social relationships in
36:35
ways that maybe don't exactly
36:38
fit our human to human relationships, but
36:40
are creating new relationships. I'm
36:43
part of a new field in , uh, communication called
36:45
human machine communication. And that's
36:47
because we recognize the machines are no longer the
36:49
medium through which we send messages to each other.
36:51
They are the thing we talk to, they
36:54
are the thing we interact with. And this, I think, raises
36:56
some interesting , uh, immediate
36:59
questions that we don't have to wait until , uh,
37:01
you know, two, three decades from now when
37:03
we get sentis or agi or whatever the heck it
37:05
is .
37:06
Yeah, yeah. Um, we
37:08
, we talked about this a little bit with , uh, Thomas
37:10
Smith singer as well. It's, I guess, kind of a social
37:12
hallucination where we, we might just
37:14
all accept that, whether Alexa
37:17
or something else , um, we just
37:19
kind of accept and act like it's , uh, sentient
37:21
even if it's not. Um, one
37:24
, one thing I wanna maybe push back a little bit on
37:26
is , uh, I mean there are some examples,
37:28
I guess other examples of where people kind
37:30
of act like something is sentient
37:32
when it's not like children with
37:35
stuffed toys, for example. Or maybe
37:38
in like a very realistic video game where
37:41
you kind of, or you
37:43
are , maybe not intentionally, but you're sort of forgetting
37:46
maybe that it , it what you're interacting with
37:48
is an npc, like an
37:50
AI character, not , um,
37:52
not another , a real person. So I , I
37:55
have to ask, I guess , um, is that necessarily a
37:57
bad thing or is, I mean, you mentioned before
37:59
, uh, it
38:01
, it , the way we treat , um,
38:03
robots, even if they're not sentient, might actually be
38:05
important because it , it , uh, influences how
38:08
we affect how we interact with other humans
38:10
as well. So is that , is that a good thing, a
38:12
bad thing? Not , not quite a clear answer.
38:14
So I don't think it's a good or bad thing, but
38:17
it's a thing, It's a thing we have
38:19
to really take seriously. Um, we
38:21
talk about suspension of disbelief. When you go
38:23
to the theater or you watch a movie, the characters
38:25
on screen are not real. And yet we
38:28
feel for them, we engage with
38:30
their emotions, and we have an experience as
38:32
a result of that. Um, and, you
38:34
know, know in the early years of cinema, that
38:37
was something that people were worried about. Um,
38:39
would people, you know, lose themselves
38:42
in, in the story and, you
38:44
know, exit reality and spend
38:46
more time in the , in the movies than in
38:48
real world? Well, that didn't happen. We figured
38:50
it out. But that's why I say I think it's a management
38:52
problem. It's a way of managing
38:55
these relationships and managing these responses
38:57
that we make to these devices, because
38:59
that is where I think the real challenge
39:02
is . I think saying yes no
39:04
is way too simplistic. It's, you know,
39:06
we we're not going to fix this by saying, don't
39:08
do that. I don't think you fix
39:10
a social problem by saying to people, stop
39:13
doing something. Prohibition never
39:16
really fixes the problem. You've gotta figure
39:18
out how to engage them in
39:20
a reasonable and emotionally
39:22
informed response that we are
39:24
able to effectively manage and that works for
39:26
them.
39:27
Yeah. I actually find that a little bit amusing
39:29
how you mentioned people would think
39:31
that , um, cinema
39:34
is going to make people lose, lose
39:36
themselves in, in all these fictional
39:38
worlds. I guess , uh, the example I'm familiar
39:40
with most recently is , um, virtual reality and
39:43
I guess video games in general. People had
39:45
that worry , Uh, and I didn't realize there was
39:47
that worry about cinema. And then I also thought, well,
39:49
what, I mean, you could go back further and save
39:52
with this . It's not like , um, cinema was the
39:55
first , uh, iteration of fiction. There were plays
39:57
there, there were books. So
40:00
unless something is particularly different about
40:02
this new medium, maybe it's, you know , the newer
40:04
mediums are more engaging. It
40:06
is kind of interesting
40:08
and funny to think about for me. So one
40:12
example from science fiction that I wanted to
40:14
get your thoughts on is in science fiction , uh,
40:16
artificial entities are often seen as being quite discreet.
40:20
Uh , so for example, what often you have a
40:22
robot and that robot is sentient , and that's
40:25
in , like, their mind is encased in that
40:27
, in that physical robot. But in
40:29
reality, it might be a little bit more complex.
40:32
You might have you say
40:34
a single sense , uh, sentient entity , artificial
40:36
entity that controls
40:39
multiple different robots. Uh , and you
40:41
mentioned already that , uh, it's, it's a
40:43
mistake to think about artificial intelligence as being
40:45
disembodied because it is embodied somewhere. It
40:47
might just be more , um, diffuse more,
40:49
more spread out in say, different servers. So
40:53
for example, maybe for , um,
40:55
maybe for an AI
40:58
losing that controls multiple different robots,
41:00
losing an individual robot might be more like,
41:02
say, losing a limb than say, a human
41:04
dying. So in
41:06
, in cases like this where it's much
41:09
more diffuse and hard to tell really where an
41:11
AI begins and ends, or robot begins and ends,
41:13
how might this affect the case for robot
41:15
rights? Or how might this affect rubber rights in practice?
41:18
So I think here, corporate law provides
41:21
a pretty good template because
41:23
corporations are also diffuse, right? Corporations
41:26
are in such a way that there is no
41:28
one place you can go and say, that's the corporation,
41:30
right? It's all over the place and
41:33
it , it has different manifestations. And
41:35
I think if we really want
41:37
to learn from that experience,
41:40
I think we'll have a good template, at
41:42
least for beginning to address a lot of these questions.
41:44
Because I think the relationship
41:47
is much more direct between
41:49
AI and the, and the corporation because
41:51
both our artifacts, both are humanly
41:53
created and both have a
41:56
kind of status. Um, in
41:58
the case of the corporation, they have personhood. In
42:00
the case of ai, we're now arguing whether
42:02
AI should also have personhood. Um,
42:05
and again, I think oftentimes we're looking to the
42:07
animal question as the template
42:09
for how we decide a lot of questions
42:12
regarding the moral legal status of AI
42:14
and robots. But I think the corporation may be a
42:16
better , uh, analogy to follow.
42:19
Um , as we try to think our way through these things.
42:21
Are there things you think that we can learn from
42:23
science fiction that maybe some,
42:26
some depictions where the useful thought experiments
42:28
or where you might think, Oh, they've , they've
42:30
got it right, That looks like it's a plausible , um,
42:33
plausible scenario?
42:34
Yeah, I think there's a lot we can learn from
42:36
science fiction, and I appreciate the question because
42:38
I think sometimes , um, the
42:41
response to science fiction by roboticists
42:44
is this kind of, yes, but
42:46
no, you know , I'm interested in it,
42:48
but don't go there because science fact is way
42:50
more complicated and it's not as
42:53
simplistic as what you get in fiction. And we how
42:55
to bracket off that fictional stuff. So we can talk
42:57
about the real stuff. I think science
42:59
fiction does a lot of heavy lifting for us in
43:01
this field. It already gave us the
43:03
word robot. We wouldn't have the word robot
43:06
if it wasn't for science fiction to begin with. Secondly,
43:08
I don't think science fiction is about the future. Many
43:11
science fiction writers and filmmakers will tell
43:14
you this. Corey Daro is one of them . He
43:16
says, Science fiction is not about predicting the future, it's
43:18
about diagnosing the present. And
43:20
so what we see in science fiction are present
43:22
anxieties, present worries, present
43:25
concerns projected on the screen of
43:27
a possible future. And so we
43:29
can see science fiction as a way of self-diagnosing
43:32
our own worries and
43:34
concerns, almost like a psychoanalysis of
43:36
us as a species right
43:38
here, right now, and what really troubles
43:41
us. And so if we look at science fiction, not
43:43
as predictive, but as diagnostic, I
43:45
think we can learn a lot from our
43:47
science fiction. I also think science
43:49
fiction gets a lot of things
43:52
right way before we get into
43:54
those matters in the scientific
43:57
research. So for example, already in Blade
43:59
Runner, you have this analogy between
44:01
real animals and robotic animals. And
44:04
this whole notion of the electric sheep
44:07
that is the title from Phillip Cape
44:09
original novella, is this idea that
44:12
we are developing devices that
44:14
are indistinguishable from real entities,
44:17
and that we could have artificial animals
44:19
and we could have natural animals. And
44:21
so this idea, I think, helps us grapple
44:24
with the way in which we build these analogies to
44:26
other kinds of entities, the way we
44:29
analogize the robot by comparing it
44:31
to the animal and learning from
44:33
our relationship to animals, how we
44:35
relate to the robot or the ai.
44:37
I also think you see in science
44:40
fiction, a lot of deep
44:42
thinking about human robot interaction.
44:44
I mean, we already today are
44:46
talking about the , uh, effect and
44:49
possible social , uh, consequences of
44:51
sex robots. We've already grappled
44:54
with a lot of those questions in science fiction. Now,
44:57
maybe we haven't got the right answers, and maybe we
44:59
haven't , uh, developed even the right
45:01
inquiry, but we've already seen prototyped
45:03
for us the kinds of things that we should
45:06
be concerned with, the kinds of directions
45:08
that we should take our inquiries so that
45:10
when we do engage these questions in social
45:12
reality, we are prepared to do
45:15
so. Finally, I think science
45:17
fiction does a lot of good work making
45:19
public a lot of things about
45:21
robots and AI that average
45:24
people would not have access to. A
45:26
lot of the research is done in proprietary
45:28
labs behind closed doors, and
45:31
we only hear about it once it's released to the public.
45:33
And then there's either excitement or outrage
45:35
as the case may be , right? I
45:38
think a lot of people, if you ask them
45:40
what they know about robots, they inevitably
45:42
are gonna talk about Mr. Data . They're
45:44
gonna talk about Westworld, they're gonna talk about
45:46
Wally, they're gonna talk about R two D
45:48
two . They know the robots science fiction
45:51
way before they get into the
45:53
science fact . This is what it's called, science fiction prototyping.
45:56
And I don't think that science fiction prototyping
45:58
is necessarily something that is bad.
46:00
I think there's a lot of education that
46:02
takes place in popular media, and
46:04
if we are careful about how we create
46:07
our stories, if we're careful about how we
46:09
cultivate critical , uh, viewers
46:11
in our young people, I
46:14
think we can use this resource
46:16
as a really good way of popularizing
46:20
our thinking about these new challenges.
46:22
Yeah, I I really like what you said about , uh,
46:24
science fiction being almost like a , a thought experiment,
46:27
which that's one of the reasons why I love
46:29
sci science, reading science fiction, watching science fiction so
46:31
much. And I just wanna shout out as
46:33
well, one of my favorite science fictions, which depicts
46:36
, um, AI in a lot of different
46:38
forms is , um, Culture Ian Banks , uh,
46:41
series. Um, so I would
46:43
recommend , uh, people check that out. Uh
46:46
, one , one thing that's related this from
46:48
, uh, that we found from some research at Cent Institute,
46:50
we , um, uh, found
46:52
that , uh, people with a
46:55
science fiction , uh, fan identity
46:57
who , um, self-identified as being , um,
46:59
science fiction fans, that trait
47:01
was correlated with people perceiving
47:04
more mind in currently existing robots and AI
47:06
perceiving more mind in
47:08
robots that might exist in the future , Uh,
47:11
stronger beliefs that Ai , ai and robots
47:13
should have , uh, would have similar value
47:15
to human feelings, less moral exclusion
47:18
of robots in the eyes than I could go on. Um
47:20
, but it , it does seem like , uh, that
47:23
, um, science fiction fan identity or being
47:25
interested in science fiction has some positive effects,
47:28
I guess hard to say, whether that's , um,
47:30
whether that's causal or maybe that's, if
47:33
one , someone has one that likely to have the other.
47:35
But that gets me thinking
47:37
about , uh, what, what
47:39
kinds of things can we do to , uh,
47:42
actually, I guess , um, almost
47:44
like an intervention, if we were interested in moral circle
47:46
expansion in the AI robot context
47:48
, um, what can we do? I don't
47:50
, I don't mean like , uh, making people watch science
47:53
fiction or something, but , um, is there anything
47:55
that you think we could do to , uh,
47:58
encourage people to think about robots
48:00
and AI in a more positive light? Or should we,
48:02
and should we be doing anything?
48:03
Yeah, no, it's, it's , again, it's a really good
48:05
question and it's important because more
48:08
expansion is something that is part
48:11
of our evolution in
48:13
both moral philosophy and in law, right? I
48:16
mean, we , we've opened the circle to include previously
48:18
excluded individuals and groups, and
48:21
that's been a good thing. And so engaging
48:23
people in these kinds of , uh, exercises,
48:25
if you wanna call 'em, that I think is , uh, not
48:28
only instructive for them, but
48:30
it also contributes to our
48:33
evolution in our thinking on these
48:35
matters. I think, as
48:37
we just have discussed, I think science fiction is
48:39
one place that you can engage people in these questions.
48:41
I know when I work with my students, one
48:44
of the things that I find them to be
48:46
most engaged with and most
48:48
excited about is when
48:50
you can take something in their
48:52
popular media experience and
48:55
open it up in a way that allows for
48:57
them to really see a lot
48:59
of these things at play , um, and gives
49:02
them some access to it. Because
49:04
I think a lot of times these technological subjects
49:06
and these matters seem rather
49:09
inaccessible. And if you can make
49:12
'em accessible by fiction, by
49:14
whatever means, I think that's a , a really good
49:16
, um, opener to the
49:18
conversation. It doesn't mean you end
49:21
the conversation there, but that's where you begin to
49:23
cultivate this way of thinking. I
49:26
think another way to do this, and I
49:28
again, have found this to be a direct
49:30
, uh, instance in my own
49:32
classroom, is by giving people
49:35
access to devices, by
49:38
letting them just engage with robots.
49:40
You know, we have this idea of the robot petting
49:42
zoo that some people put together conferences and
49:45
stuff, but I think this is important. I
49:47
think kids are curious, especially
49:49
younger, younger kids, you know, high
49:51
school and and below, they
49:54
want to engage these things. They want to
49:56
take their curiosity and see, you
49:58
know , what happens. And
50:01
giving them access, I think is
50:03
crucial, because otherwise, it's
50:05
something that happens at Google. It's something that
50:07
happens at Microsoft, and therefore
50:09
it's not really a part of what they
50:12
are. It's not really in their world in
50:14
a way that they can make sense of it. And I
50:16
think access is absolutely necessary in
50:18
, in that area. I also think education
50:21
is very key to a lot of
50:23
this stuff. Again, I think we've
50:25
limited access to a lot of these materials
50:28
to specialists in computer science, artificial
50:30
intelligence, engineering, and elsewhere. We
50:32
, I think we've gotta open the curriculum and make
50:35
this stuff broadly available. Um, you can see already
50:38
with the release of Dolly and the
50:40
way people are using it to generate images, we
50:42
need artists to be engaged with this
50:44
technology, and we need artists to
50:46
help us make sense of what this creativity with
50:49
AI is all about. And if we don't
50:51
let artists into the conversation, we're
50:53
not going to learn what we can possibly
50:56
gather from the history of human
50:58
artistic expression and experience. The
51:01
same with music, the same with
51:03
journalism, the same with any field.
51:05
I think this technology is
51:08
no longer able to be limited to
51:10
one specific field, and we've gotta teach
51:13
it across the curriculum in a way that begins
51:15
early and that gets the curiosity
51:18
of our young learners engaged
51:20
from the very early stages of
51:22
their career.
51:23
Great. Thanks for that. Uh , so just
51:25
a couple of questions to sort of wrap this
51:27
all up. Um , I've noticed that you, you
51:29
had an interest in programming from
51:32
a young age, and you've actually developed internet
51:34
applications. You, you , um, uh,
51:37
a an established developer, but
51:39
instead of pursuing computer science , uh, more generally,
51:41
you followed a career in the philosophy of technology.
51:44
Um, why do you think that is? What interests you about, about
51:46
the philosophy of technology more so than coding
51:48
itself?
51:49
Yeah, this is interesting because I , I
51:51
used web development as
51:53
the means to pay for the bad habit of going to
51:55
grad school, <laugh> . But
51:58
it's funny because those two things tracked
52:00
really nicely because one was very
52:02
hands on , very practical, and
52:05
the other was very heady, very theoretical, and
52:07
so they sort of balanced each other out. Uh
52:10
, but one , one thing I noticed as I was
52:12
doing these things simultaneously is
52:15
that the one could really speak to the other, if
52:17
somebody would build the bridge, that
52:19
what we were doing in programming
52:22
and in developing applications could
52:24
actually gain from the traditions
52:26
of human learning, from a epistemology,
52:28
from metaphysics, from ethics, you name
52:30
it. If we would only build the bridge
52:32
to those philosophical traditions, we'd be able to open
52:35
that conversation. And I think we've been rather successful with
52:37
that. If you see how AI ethics has really
52:39
exploded in the last five years,
52:42
but it also goes the other direction. I
52:44
think the computer and digital
52:46
media and ai and robots can
52:49
actually provide philosophy with
52:51
some really interesting thought experiments
52:53
on the one hand, but also some interesting challenges,
52:56
the human exceptionalism and the way we think
52:58
about ourselves as being the center of the universe, and
53:01
therefore the only sentient creature to,
53:03
you know, exist on planet earth, which obviously
53:05
isn't true. So what I
53:07
saw as this ability to use the
53:09
one to inform the other, and the reason
53:11
I went in one direction as opposed
53:13
to the other direction, it just had to do
53:15
that. It turned out I'm a better teacher than
53:17
I am programmer <laugh> . And so
53:20
I pursued the one that was gonna take me in
53:22
that direction.
53:23
Yeah. Do you , do you think , uh,
53:25
your work , um, in
53:27
developing has given you some credibility? Because
53:30
I imagine there , there
53:32
might be some , uh, people in the
53:34
philosophy of technology who maybe
53:37
aren't taken so seriously by people who
53:39
actually work on artificial intelligence, machine learning,
53:41
what have you. Uh , and
53:44
I , I can think of , um, some,
53:46
there are some people who don't, like , for example, don't
53:48
take , um, AI safety very seriously,
53:50
who work in , uh, who work
53:52
in actual the actual development of ai. They might think
53:55
these people, you know, they have these ideas,
53:57
but they don't really know anything about technology. They're kind of naive
54:00
is , is what they might say. So do you think because
54:02
you've kind of , um, you've , you've
54:05
done both, do you think that gives you some credibility in,
54:07
in that , uh, the tech space?
54:09
I hope it does. What , what I will say
54:11
is that it feels very
54:13
dishonest to me to
54:15
talk about machine learning or
54:17
any other technology and not know how
54:20
it works. Mm-hmm . <affirmative> , I'll
54:23
just give you some examples from my own sort of trajectory.
54:25
So I wrote a book on remix in
54:28
the, you know, 2016,
54:31
I think it came out. And it
54:33
took me a while to write the book, not because
54:35
I couldn't write it, but because I wanted to learn
54:38
to be a DJ before I wrote the book. And
54:40
I spent all this time developing the practice because
54:43
I didn't think I had the credibility to
54:45
speak about this artistic practice
54:47
and the technology behind it without
54:49
knowing the how it works, knowing the tools
54:51
and having hands on experience with it. The
54:54
same, when I started to engage with
54:56
AI and robotics, I knew that
54:59
there was no way I could speak with any
55:01
credibility about machine
55:03
learning, about big data, about neural
55:05
networks and all these things. If I hadn't
55:07
built one, if I hadn't done the
55:10
task at hand in actually
55:13
constructing a neural network, training it on
55:15
data and see what it does. And I
55:17
think for my own writing, this is what
55:19
allows me to speak with some conviction,
55:22
with some real good
55:24
grounding in the technology. And
55:27
that hopefully is communicated
55:30
in the resulting text to the rest of
55:32
the world that I'm not just
55:34
making this up. I, I come to this from
55:36
the perspective of really knowing what goes on
55:38
behind the scenes and have brought my
55:40
philosophical knowledge to bear on something I
55:43
have direct hands-on experience with.
55:45
See it . So I I, I've got a similar experience
55:48
in that I tried to , um, I , I actually did
55:50
a PhD in space science , uh,
55:52
and , um, I have a
55:54
geo geoscience background and I wanted to , uh,
55:56
work a little bit on say, some long-termism
55:59
, uh, like ethical
56:02
questions and apply that as, as it applies
56:04
to space science. Uh , and I thought
56:06
that doing that might , could be more credibility
56:08
when I talk about these ethical problems. Um,
56:11
but I , in my experience, something
56:13
that's too soon to say perhaps, but it doesn't feel
56:15
like it. Uh , I think there's
56:18
just , I've been met with a lot of skepticism from the space
56:20
science community on some of those ethical ideas.
56:23
Um, but it , it sounds like that's worked out better for you, so
56:26
that's that's good to hear.
56:27
It . It doesn't mean that you don't get pushback . It
56:29
doesn't mean that you don't get criticism. Yeah , yeah . Um
56:31
, I think, you know, it , it's
56:33
always a push pull . You're always kind of putting
56:36
yourself out there and then trying to justify what it
56:38
is you're doing to others who may be skeptical of
56:40
it. Um, especially when your
56:42
ideas might be less than popular, which
56:45
often is the case in , in academia. But
56:48
I think the dialogue is crucial.
56:50
And I think meeting people where
56:52
they're at is part
56:54
of building that transaction , um,
56:57
and making it work. Uh , I did have a
56:59
guy at one point on Twitter say to me , Um,
57:01
you should shut up because you don't work
57:03
in this field and you don't know what you're talking about. And
57:05
so I sent him the , uh,
57:08
neural network that I built. I
57:10
just sent him the code, and I just said, Here,
57:14
<laugh> , that that must
57:16
be satisfying in a , in a kind of <laugh> , vindictive
57:18
Way. It was very satisfying.
57:21
Can imagine. Um, well
57:23
, yeah, just , uh, to , to kind of bring this
57:25
together. Do you , um, over your career
57:28
so far, do you think you've noticed any shifts
57:31
in thinking , um, about how we think
57:33
about robot rights? Um, one, one just
57:35
to , to kind of prime you is , um, maybe
57:37
a shift in people thinking about robots
57:40
, um, as moral agents
57:42
to shifting to thinking about them more as moral
57:44
patients. Do you, what have you seen
57:46
over your career so far?
57:48
So I , yeah, I , as I said, I think earlier is
57:50
, you know, when this started for me it
57:52
was a really small, you know, I , I
57:54
could really hold on, you know, my
57:56
fingers, how many people were working on this subject, right?
57:58
And that was it. And it's really exploded.
58:00
I think the work that you've done at Sun Institute documenting
58:03
this in your review of literatures that
58:05
you've done , uh, really shows this exponential
58:08
increase in interest, but also
58:10
, uh, scholarship in this field. And
58:13
that, on the one hand is very gratifying. On
58:15
the other hand, it's hard to keep up <laugh> cause there's
58:17
so much happening and there's a lot to read and
58:19
a lot to stay on top of. Um,
58:22
but I will say that a couple of trends
58:24
have emerged , uh, in this process. I
58:27
think there has been an increasing move
58:30
from this being a speculative
58:32
moral subject to this being a
58:34
much more pr pragmatic and
58:36
practical legal subject. My
58:38
own thinking has evolved in that way. My first book,
58:41
the Machine Question was very philosophical,
58:43
very morally situated
58:45
in those traditions. My most recent
58:47
book, which is gonna be this thing called Person thing robot
58:50
from MIT press , come on next year
58:52
, um, is much more engaged
58:54
with the legal philosophy and
58:56
with legal practice. And that
58:58
just, I think, is a reflection
59:00
of the fact that that's how the trend is
59:02
gone in the research over the
59:05
last decade. Another thing
59:07
I've noticed is a
59:09
development in the,
59:12
bringing into the conversation these
59:15
non-western ways of thinking about these questions.
59:17
I think when these questions began over
59:20
a decade ago, the way
59:22
in which I and others were very much
59:24
engaging these things were by leveraging
59:26
sort of the western moral and legal traditions
59:29
to try to speak to the people
59:31
who are building these things and developing these technologies.
59:34
Over this past decade, we've
59:37
seen, I think, a greater engagement
59:39
with, in a greater desire to
59:42
engage with other ways of thinking
59:44
and other ways of seeing not
59:46
as a way of , uh, doing something better
59:48
or worse or whatever the case is , but it's just
59:50
tapping into that difference
59:53
that we can see in human thought
59:55
processes that allow for us to really
59:57
cultivate a relationship to
1:00:00
the wide range of human wisdom as
1:00:02
it's developed over time, but also over space.
1:00:05
And I would say the last
1:00:07
thing I've seen, and this is very gratifying, I
1:00:09
think when I started this and
1:00:12
began to talk about robot
1:00:14
rights as a subject matter for
1:00:16
investigation , uh, there
1:00:18
was I think a lot of very
1:00:21
abrasive and very , uh,
1:00:23
sort of triggered reactions. How can you say
1:00:25
that this is, this is just horrendous. I
1:00:27
mean, who would talk this way? Um,
1:00:29
and I had this very famous picture I put
1:00:32
on Twitter with me holding a sign that said, Ro
1:00:34
robot writes now say
1:00:36
that . Yeah . And it really sparked an amazing,
1:00:39
huge controversy , um,
1:00:41
about a decade ago , well, about
1:00:43
five years ago. And I
1:00:46
learned a lot in that little exchange that
1:00:48
took place, but it was an explosion of in , of
1:00:50
interest, but also of , of pushback . I
1:00:53
think we've seen it evolve to the point now where people
1:00:55
are saying, Yeah, we need to talk
1:00:57
. This has gotta be talked about. This has gotta be grappled
1:00:59
with, We can't just put
1:01:01
, put our fingers in our ears and go, blah, blah, blah,
1:01:03
blah, blah . This doesn't exist. It does exist. Laws
1:01:07
are being made, hearings are
1:01:09
happening. AI personhood is
1:01:11
not something for the future. It's something
1:01:13
that legislatures are looking at right now. And
1:01:16
as a result of all this, taking these questions
1:01:18
seriously , um, and engaging
1:01:20
with them in a way that is informed by
1:01:23
good moral and legal thinking processes,
1:01:25
I think is absolutely crucial. And I've seen in
1:01:28
the last five years that mature in
1:01:30
a way that I think really , uh, speaks
1:01:32
to the fact that a lot of people have
1:01:34
found this to be not only of interest, but
1:01:37
also something that is crucial for our
1:01:39
engagement with as researchers.
1:01:42
Hmm , <affirmative> . Great. Well thanks. Thanks for that , David. Um
1:01:45
, just to finish up, where can listeners
1:01:47
, uh, best follow you in your work? And is there anything
1:01:49
in particular you'd wanna suggest they look at, whether
1:01:51
it's a book or any, any other , um, piece of
1:01:53
work you've worked on? If they're , especially if
1:01:55
they're interested in this topic and want to wanna learn more.
1:01:57
So you can follow me on Twitter. It's , uh,
1:01:59
David Underlined Gun. Uh , that's
1:02:01
my handle. Um , you can find it very
1:02:04
easily. My website is gun web.com
1:02:06
and you can go there for access
1:02:09
to texts and books and things
1:02:11
like that. I would say right
1:02:13
now, if this is of interest
1:02:15
to you and you really want to jump in
1:02:18
feet first and sort of see what it's all about , uh,
1:02:20
the two books that began all this was the
1:02:22
Machine Question, Critical Perspectives
1:02:25
on ai, Robots and Ethics from 2012.
1:02:28
And Robot Writes from 2018,
1:02:30
both published by MIT Press, and
1:02:32
, uh, you should be able to get both of 'em used for very cheap
1:02:35
these days. Um, or
1:02:37
go to the library, they have 'em too. Uh
1:02:39
, but that's a pretty good way to get,
1:02:41
I think, into this material.
1:02:44
Um , and because of the kind of research I
1:02:46
do, I try to be
1:02:48
as exhaustive as possible and documenting
1:02:51
what people have said, where it's
1:02:53
going, where it's come from, and hopefully make
1:02:55
sense of it. So it hopefully will
1:02:57
provide people with a good guide to
1:03:00
, uh, finding their way through this stuff
1:03:02
and figuring out where they stand.
1:03:03
That's great. Thanks. We'll have , uh, links to
1:03:05
all of that in the show notes and everything else that we've
1:03:08
referred to in the, in the show. So thank
1:03:10
you again, David, Really appreciate your time and thanks for joining us.
1:03:12
Yeah, it's been really great to talk to you and , uh, I
1:03:14
appreciate the questions. Uh, as
1:03:17
, as you said early on, there's a sort of reason
1:03:19
that the Sentence Institute is
1:03:21
interested in these questions and there's a reason I'm interested
1:03:23
in these questions and I think they dovetail very
1:03:26
nicely. And it was great, you know, to talk with
1:03:28
you about these matters.
1:03:30
Thanks for listening. I hope you enjoyed the episode.
1:03:33
You can subscribe to The Sentence Institute
1:03:35
podcast on iTunes, Stitcher,
1:03:37
or any podcast app.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More