Latest / Uncomfortable Conversations with Josh Szeps / "A.I., Artificial Souls, and the Crazy Conundrum of Consciousness" with Philosopher Eric Schwitzgebel

Transcript

"A.I., Artificial Souls, and the Crazy Conundrum of Consciousness" with Philosopher Eric Schwitzgebel · Uncomfortable Conversations with Josh Szeps

0:00
G'day humans, welcome to the Safe Space for Dangerous Ideas, and it's one of the most
0:08
dangerous ideas that we ever contemplate.
0:11
How did it come to pass?
0:13
That physical material, ejected from stars, has assembled itself into your brain, capable
0:19
of gazing back on those stars, and thinking stone thoughts about them as you pull on a
0:24
rougher line when you're back as a 15-year-old child looking up at the stars, or maybe
0:27
I'm just talking about my own experiences here.
0:31
The mystery of consciousness has perplexed philosophers for many, many years.
0:35
I felt like it was time to talk to one of the world's leading philosophers of mind who
0:39
is currently working on a book about whether artificial intelligence will ever become conscious,
0:44
and if it does, whether we will be obligated to give it rights and to treat it ethically.
0:50
Eric Schwitz-Gable is an absolute leader in his field, he's a professor of philosophy at
0:54
the University of California Riverside, he specializes in the philosophy of mind, in
0:58
moral psychology, and in epistemology.
1:01
He writes a psychology philosophy blog called the splintered mind, and I first came across
1:05
him in his book, The Weirdness of the World.
1:09
We talk here mostly about the competing theories of what consciousness is and how it
1:14
arises, and then we sort of segue into questions of whether or not it's plausible that it
1:19
will arise inside machines.
1:22
All out of Riefer for this one, I hope you enjoy as much as I did, the one, the only
1:26
Eric Schwitz-Gable.
1:29
Basically, how it comes to pass, that creatures that are made from the stuff of stars have
1:39
managed to assemble themselves into entities that are having experiences in the first place,
1:45
like this is sometimes called the hard problem of consciousness.
1:49
I mean, it's hard to reflect on the weirdness and awesomeness of that very fact, right?
1:55
Eric Schwitz-Gable I mean, it's wild, I literally just got back yesterday from a trip to Norway
2:01
above the Arctic Circle with my mum, I was taking her on a birthday trip, it was a bucket
2:05
list destination, we were just standing there in the midnight sun, looking at these mountains
2:12
and glaciers, and I was just going, this is truly, like the fact that there is an entity
2:19
in the universe that has come to be capable of beholding this and feeling experiences
2:25
about this.
2:27
If you're not religious, which I guess I'm not, it does leave a big empty jigsaw piece
2:33
in the universe that is hard, it really is, I mean, I get that there are all these explanations
2:37
for it, but intuitively, it smells like something's missing.
2:44
Yes, so I agree with that, so yes, two things about that.
2:48
One is to just notice how awesome and weird and wonderful and amazing and incredible it is,
2:54
but that's real, and the second thing is to notice that I don't think we should be totally
3:02
comfortable with just accepting mainstream, scientific, materialism as an answer about this.
3:09
Right, my leanings are toward mainstream scientific materialism,
3:13
but I'm overall, I'm kind of 50-50 on it, right?
3:16
If I had to guess, maybe I would guess on it as opposed to any specific competitor,
3:21
but I'm not really totally confident.
3:23
I mean, I think it is scientific materialists say, okay, like,
3:29
atoms and swirling around somehow give rise to our experience,
3:33
it's got to be possible because how could there be anything else,
3:39
and we don't really have direct scientific evidence for anything else?
3:43
So it's kind of like, by default, if we're fans of science and we're not theists,
3:50
we kind of say, well, it seems like it has to be the case that somehow this world of atoms
3:56
creates conscious experience, but you know, that's not very satisfying.
4:01
I don't think the scientists, the scientific materialists really can deliver the goods
4:06
that we really want, which is a satisfying account of how that really happens,
4:11
other than to say, oh, must happen somehow.
4:15
You said that you would probably bet on materialism if you're a betting person,
4:21
because there's no great viable alternative.
4:23
What are the better alternatives? What are the, what is the constellation of options that are
4:28
not materialism that you find most plausible?
4:32
So I will first, I'll mention three, I'll first mention the most common one, which is actually not
4:39
my favorite one, but the most common one with which your listeners might already be familiar
4:43
is substance dualism. According to substance dualism, there's an immaterial soul. That's the
4:49
locus of your conscious experience, right? If you think there's an afterlife, it continues in
4:53
the afterlife, it's you, right? This immaterial soul is hooked on tentatively and temporarily to
4:59
this particular physical body, and then it goes on, right? So that substance dualism, I think that
5:08
that is an alternative that should be taken seriously to the mainstream materialist picture,
5:15
but it has a lot of problems, which we can get into if you want.
5:19
Yeah, sure.
5:22
Here's one of the big problems. Who has an immaterial soul?
5:27
Hemosypians, I think, is the general answer, right?
5:29
Oh, yes. Well, that's a good starting place. How about dogs?
5:35
No. No. All right. Now, if you're answering the question of consciousness by pointing to
5:41
immaterial souls, then you're now saying dogs aren't conscious. They have no experiences.
5:49
There's nothing it's like to be a dog, right? We're not talking about self-consciousness,
5:52
but you're just talking like, are there feelings or the lights on? Is there sentience, right?
5:57
That's a big cost to pay, right? It seems both intuitively very plausible that dogs actually
6:03
have experiences. They feel pain if they injure their paw, but they feel hungry. They have sensory
6:09
experiences of taste and sight. It seems super plausible. So if you say dogs don't have souls,
6:16
that's a big cost to pay. It also doesn't fit very neatly with the standard scientific
6:20
picture, right? Like, in evolution, like, did the soul come in at like, homo erectus or like,
6:26
how does that work? Right? So maybe you have to buy a huge kind of package of religious worldviews
6:33
that are anti-evolutionary and the soul is installed kind of by miracle, like conception. You
6:38
have to maybe buy that whole picture, which has a lot of costs scientifically. Is there a theological
6:46
version that says that there's a hierarchy of souls in some way and that like, the real soul is
6:52
the one that gets instilled in homo sapiens, but there are kind of... There are dogs souls that...
6:57
There are sort of like bullshit. There are kind of bullshit ways of being sentient and semi-conscious
7:02
that animals have that a kind of machine like, but it's true that it's also something like to be
7:07
them, but it's not the fully fledged, true legit bells and whistles soul that we get. I mean,
7:13
you could work on a theory like that, but you're starting to see what the costs are, right?
7:18
Of course you could say dogs. This is not an easy PhD in theology that I'm proposing. You could
7:23
say dogs do have souls, and I think the reason you wanted to say no is that you saw how this is
7:27
going to become a slippery slope, right? Yes, that's right. If dogs have souls, do mice have souls,
7:31
mice have souls, do lizards have souls, lizards have souls, do snails have, where do you stop?
7:36
So that's just one. I think there are other problems too, but that's just one of the challenges,
7:42
the substance tools and faces that I think is pretty difficult in a major cost, and especially
7:50
for people who are scientifically inclined as I am, it's like it's a pretty high cost to pay.
7:55
Yeah, I mean, there are also challenges with what is the mechanism by which the soul is interacting
8:00
with the material world, right? Do you want to just address that?
8:04
Yeah, that's the other famous thing, and if you read the history of dualist efforts in early modern
8:11
European philosophy, it's just one after another struggle with this question of how can an
8:19
immaterial soul cause something to happen in a material world, right? You've got this immaterial
8:26
soul, and somehow when you say in your immaterial soul, you form an intention to raise your arm,
8:32
somehow your arm has to now go up. So it has to be some kind of causal connection between this
8:37
immaterial soul and I don't know, the part of your brain that issues the motor signal to your
8:43
arm, and how does that work? Because it looks like the only things that can affect how neurons behave
8:49
are ordinary physical things, scientists don't say, oh yeah, there's been a soul,
8:54
you know, did its thing, and as a result, the neuron fired an action potential, right? That's just
8:59
again, there's like some problems there, so you could say, oh no, souls don't have any effect,
9:03
they're epiphenomenal, some people will say, right? You have experiences in your soul as a result
9:09
of what's going on in your brain, but none of those experiences have any effects on anything,
9:13
but that kind of has unintuitive consequences. It also seems like, and this is really nicely
9:19
put forward by a wonderful book called Man of Machine by Julian Afraidyla Matri, where he says,
9:25
well look, you know, these immaterial souls, do they get drunk? Right, it seems like when you
9:32
drink alcohol, it affects all these facts about your cognition, right? But if cognition is
9:37
happening in an immaterial soul, how would that be possible? Right? So like how and when you're sleepy,
9:43
right? So how it seems like cognition is too intimately tied up with what's going on in the brain,
9:53
and the physical process of what's going on in the brain, for it all to be transpiring in this
9:57
soul, but then later has some effect on the brain. Yeah, and both of the explanations of how the
10:04
soul does or doesn't interact with the material world of your brain and your body tie themselves
10:09
up in knots, because if it doesn't interact and it's just in the realm of the soul, then it's
10:16
not really answering any questions about how our body is doing the things that it's doing and how
10:20
our brain is doing the things that it's doing, because it's not being materially influenced by
10:23
the soul. And if on the other hand, it can interact with our brain and then whatever the link is,
10:30
whatever the mechanism is by which it's doing the interacting is enough to explain our
10:37
our behavior surely. Like what is the soul adding onto the end of the lever that is touching the
10:44
brain? Like it could just be the brain and the lever in a sense, right? Like whatever mechanism
10:48
the brain, the soul uses to interact with the world is itself as sufficient as the soul is.
10:56
I mean, I'm tying myself up and not even trying to understand the kind of seems like that.
11:01
So you have to, I mean, in order to avoid that problem, you kind of need to have the brain
11:05
causes some stuff to happen in the soul and then kind of like while the brain's on pause, but not
11:11
really. The soul does its independent thing and then causes the brain to do other stuff, right?
11:19
Right, but then it's not adding anything. It's just a brain, all you're talking about is a brain
11:23
that's not a brain. It would have to act otherwise than it would have.
11:28
Right, the soul needs to do something that wouldn't otherwise happen in the brain if it's going to
11:33
really have an effect on your behavior other than just kind of like instituting what the brain was
11:39
already going to do. So it's got to be independent. This is what LemaTree's criticism is nicely
11:45
focuses on, right? If you really want the soul to do independent work that can't be done by the
11:49
brain, it kind of needs to be segregated off. Its causal presses is denied to be segregated off
11:54
from the brain. You need to kind of work offline independently of what's going on in the brain.
11:58
But that doesn't seem like what happens because you can get drunk and that affects those processes,
12:03
so it doesn't seem like they really are segregated. Right, and it never happens that you can consume
12:09
10 drinks and your soul declines to get drunk. Right. It has no autonomy beyond the physiological
12:17
realities of the brain, of the way it meets in your head. And similarly, I mean, it's obviously
12:22
when you really start thinking about it, it makes no sense because if I drill a hole in your head
12:25
and I poke my finger into a portion of your brain, we can predict exactly what is going to happen
12:32
to your soul. Essentially, to your experience of being alive, depending on what part of the brain
12:37
I damage or remove, right? There's no ambiguity there. There's a direct 100%
12:43
correlate between your experience of the world and what happens to your brain.
12:47
It does seem like a simpler thing to have that soul in there. Now, of course, I mean,
12:53
it's up to the dualists to point out that it's not 100% as far as we can tell. I mean, you can say
12:59
it's 100% from a scientific materialist worldview, it must be. But of course, we're not so good
13:06
at neuroimaging and at knowing what's going on in the brain that we can say, oh, yes, this is
13:10
exactly what's going to happen if you stimulate this area. There's a lot of noise
13:15
in our predictions. Yeah, we're pretty good. I mean, we're pretty good. If I put a bullet
13:20
through your head, we know what's going to happen. If I cut out a quarter of your brain,
13:23
then the soul is not coming out unscathed. But you're right. It differs on the level of the very,
13:30
very specific. We should note, 95% of the world and almost 100% of people who have ever lived
13:38
have believed this theory because they've been fierce. Anyone who believes in a creator
13:44
God who can intervene in worldly affairs and speak to them and in an afterlife is a dualist,
13:53
right? There are materialist versions of the afterlife. Okay. So I think it's common in our
14:03
Western monotheistic influence tradition to think, okay, if you think there's an afterlife,
14:09
you kind of have to accept mind, body, dualism, you know, and then a material soul and a material
14:15
body. But I think it's, there are certainly in the Western tradition, there are theists who think
14:21
that you're reincarnated materially. I mean, that is the standard view in the Christian tradition,
14:29
at least, maybe the other traditions as well, right, that you are materially reincarnated. So
14:35
so it's possible that you could be a materialist and say, okay, well, well, there is no
14:41
immaterial soul. The afterlife is a bodily thing, right? Also, I think if you look at
14:47
wait, I just I don't, you've lost me the the brain decays. The brain goes in the ground. The
14:53
brain either gets burned in a cremation or it gets eaten by worms under the ground. You can't be
14:59
a materialist. Right, but then what's he pulling out? He's pulling out a soul. There's no soul,
15:06
right? There's just a recourse in physical energy. But if no, if it's mature, I don't get that.
15:12
Like the, if you're, if you are your brain and your brain dies, then you die. If you're,
15:19
if he's going to reconstitute something, then there has to be something enduring beyond the death
15:22
of the brain. And that thing you might as well call a soul. I actually don't think that's true.
15:30
Down this rabbit hole you want to go. Yeah, no, I'd like to understand the whole identity.
15:36
But so here's the picture. Here's a picture that you can be if you're a materialist who believes
15:40
in an apricot. You can say, okay, your body dies in a sense. You're dead. Now God who can perform
15:47
miracles can so to speak, pick up every molecule that used to be in you, put them back together in
15:55
heaven. And there's now a material entity, which is composed of the same stuff you were composed of.
16:02
And now as the same physical states that you had, and maybe the same, all the same memories
16:08
that you had that is now living a material life. And there's no material soul involved. It's just
16:17
a recost-suited material thing just like if you took a ship and you took all the boards apart,
16:23
spreading across the world, and then brought all the boards together later, you'd say, we got this,
16:27
we got the ship back. The ship of Thesius. The ship of Thesius is back.
16:32
I don't think maybe I don't know enough fundamentalist Christians. I don't think that most
16:36
Christians or monotheistic believers think that there think that heaven is a physical place in
16:42
this, located in this universe made of the material things of this universe. I mean, I understand
16:46
what you're saying. If there was, if one of the, if the theory of Christianity was that when you die,
16:52
or when Jesus died, that heaven is located 150 kilometers underneath Baltimore, and that there's
17:00
a big cave, and they take the molecules and the atoms down there, and they reassemble you
17:05
like some deranged Elon Musk layer, and they put you to back together again like a zombie.
17:12
Then I would concede that, yeah, that's consistent with materialism. But I think the story is
17:17
that there's an invisible place in another dimension that doesn't obey the, it obviously doesn't
17:23
obey the physical logic of our universe because people live forever there, and there are flying
17:27
angels and things. So I don't think you can get around the materialism problem that easily,
17:33
just by saying they recreate you in a different place. The place would have to be physically located
17:38
in this universe, and I don't think that's what most Christians believe. I suspect you're right
17:43
that that's not what most Christians believe. I also do think that they're, or I do know that
17:49
there are some Christians who believe, and so they think they think they think they have, they think
17:56
you'd be able to look at heavens through a telescope. If we knew where to look and we had a powerful
18:01
telescope, I don't know the details. But they may not either. And they may not either. But it is not,
18:11
so here's two things I do know. One is there are some materialists who believe in afterlife
18:19
for in in some kind of manner like I just described. And two is I think it's not unusual
18:27
for people to think that even if there's an immaterial soul, your body, you're rejoined with your
18:32
body. So you're totally immaterial. So, but yes, I don't, I'm not an expert on this, on this kind of
18:41
thing. I just, all right. That's it. It sounds like it sounds like a bullshit dodged to me. But
18:45
you know, people will do strange things in order to avoid the inconsistencies of their own
18:49
ontology or their metaphysics. So we, all right. So there's the soul explanation. The
18:56
soul explanation for how, for as an alternative explanation to materialism. What's another?
19:04
So let me briefly mention another one of my favorites and then go to a third one. So the one I'll
19:15
briefly mention is property dualism. According to property dualism, there are two substances,
19:24
a soul that's not material and a material body. Rather, there's only kind of one type of stuff,
19:30
stuff, so to speak, but it's got two different kinds of properties. So you are one entity,
19:35
one unified entity. You've got some physical properties, but you've also got some mental properties
19:40
that are not reducible to physical properties. So that's not a materialist view. And there's no
19:50
soul that could go onto an afterlife on this view. But it does mean that there are some properties
20:00
you have, the conscious ones in particular, they're not fully explainable by ordinary physical
20:07
sunnets. Okay. David Chambers is probably the most famous advocate of that. Right. That sounds
20:17
total logical when you say it like that a little. You know, when you express how mysterious it is,
20:26
that these swirling atoms could somehow give rise to consciousness. Right. The property dualist
20:33
kind of lean into that and say, yes, no matter how much you describe what's going on with those
20:39
atoms, you're always going to leave some important thing out. And that is that property of being
20:48
consciousness. And you can kind of imagine or conceive of that property being absent,
20:57
no matter how you spin up the molecules. And that conceivability shows or suggests,
21:07
you know, and here the argument gets tricky. But somehow, the fact that you can coherently conceive
21:15
of a physical molecule from a molecule duplicative view, sometimes called a zombie in philosophy,
21:22
that doesn't have conscious experiences, shows that the property of being conscious is something
21:28
over and above the property of having all of these physical features. Yeah. I mean, I find that
21:33
very persuasive. I get that. I mean, I like the philosophical zombie idea, this idea that
21:39
you can imagine an automaton, a robot or a zombie that outwardly appears to be exactly the same as you,
21:46
but the lights aren't on. It does all the same stuff, but the lights aren't on. So why are our lights
21:50
on? And I often wonder about this. I mean, why is it necessary, even from an evolutionary point of
21:55
view, for us to be having actual experiences? Right. I mean, why aren't we all just wandering around
22:00
with nothing going on inside, but with all of the necessary hardware and software for us to be
22:05
making equally sophisticated decisions about life and death, which presumably isn't inconsistent.
22:10
Like, do we need to feel the pain as long as the program in our head told us that pain was
22:16
bad? Like, why do we have to have the qualia of existence at all in order to survive? And so I get
22:22
that aspect of charmers, but I don't understand what the second movie is then to say,
22:30
therefore, we are made up of material stuff and what? Well, there are material and immaterial
22:39
properties. So it's not a difference in the kind of stuff. The property dualist thinks
22:45
there are different things in it. Properties that you have. So if there were a molecule for
22:49
molecule identical version of you, physically, perfectly identical, but doesn't have consciousness.
22:57
So if we can conceive of that coherently as a possibility, then we're conceiving something
23:04
that has every physical property you have, but isn't conscious. So therefore, being conscious
23:11
must be some property over and above all of those physical properties.
23:17
But if the copy is made, if the thing that the copy is made of is exactly the same stuff that I'm
23:22
made of, meaning neurons inside my head, then presumably it would be conscious.
23:28
Presumably it would be conscious, but this is why it's important to think about if we're going to
23:34
get into this argument how the conceivability is supposed to work. So charmers thinks that anything
23:40
that's a molecule for molecule identical version of you would, by the laws of nature, be conscious,
23:47
but it wouldn't by the laws of logic be conscious. So I mean, you can distinguish between.
23:54
Only if a loss if it would be able to slice that definitional sandwich.
23:58
We need to slice the sandwich. So if I hold a pen up, the laws of nature say it's going to fall
24:03
when I release it, but we can conceive of it floating up. That would be a violation of the laws
24:11
of nature, but there's nothing logically in coherent in that. So the kind of possibility
24:18
that the zombie is supposed to be, is it's supposed to be logically coherent, conceptually coherent,
24:25
and yet it wouldn't happen according to the laws of nature.
24:31
Right. But for charmers' argument or the property dualist argument to work,
24:36
they need to say something about that conceptual coherence shows you that consciousness is a
24:43
property that isn't logically identical to all these other properties. Right. So like, here's
24:49
another example. Right. Think about, imagine a rectangle. It's a logical feature of the rectangle
24:59
that not only has it four sides, but it's got four interior angles. You can't imagine coherently
25:06
conceive of one without conceiving of the other. They go hand in hand, right. So it's not possible
25:14
for a rectangle to have one but not the other. But there is a sense in which it is conceivable,
25:23
if charmers is right, that an entity could have all of our physical properties and not be conscious.
25:33
That conceivability shows that it's a logical possibility. There's no logical contradiction in that.
25:39
The fact that it's a logical, not a logical contradiction to imagine it shows that you have,
25:46
you can slice up the universe into properties sufficiently thin. But you get to say,
25:53
yeah, there's an additional thing. So God didn't need to speak metaphorically. Once God made a
26:01
four-sided planar figure, he didn't need to do a further act to make sure he had four angles.
26:06
Once God made physical humans, there did have to be some kind of further act to ensure that they
26:14
were also conscious. Right. And many critics of charmers say that's just an illusion, actually, and
26:22
that it's a bit like saying you could imagine an airplane flying backwards. There's nothing logically
26:30
inconsistent about that. But in actual fact, there is. It's just that your brain is making a mistake
26:35
when it pictures an airplane flying backwards. That can't happen. And similarly, it can't happen
26:40
that you would have a human being constructed as a human being, behaving like a human being
26:45
that wasn't conscious. You're just inventing something and then pretending that it could exist
26:49
because it can exist in your fantasy. Right. So this is,
26:54
then, for example, Daniel, that I missed the basic standard type of reply to charmer. He says,
27:00
it's not, if you really thought about it clearly enough, you would realize it's not actually
27:06
logically conceivable. That if you thought about it clearly enough, you would realize it's like
27:11
the case of a rectangle has to have four interior angles. To us, in a disturbing foresight,
27:20
right, to us, you know, maybe it seems really obvious that the number of sides and the number of
27:27
angles in a rectangle have to match. Right. But there are mathematical things that are much less
27:33
obvious to us and yet are necessarily logically true. Right. Similarly, these facts about consciousness
27:40
may be much less obvious to us and yet still necessarily true. It's not just, it wouldn't just be
27:46
a violation of the laws of nature, which charmer thinks, to think of a molecule for molecule
27:52
identical version of you that was not conscious, would actually be a violation of the laws of logic.
27:57
There would be a fundamental incoherence in that as incoherent as trying to imagine a rectangle
28:03
that didn't have four interior angles. Does Dennis' point make sense to you? I am tempted by it.
28:10
I can feel pulled either direction on that question. It feels like a bit of a dodge to me. Do you
28:18
want to just paint a broader picture of Dennis' case because it's something that we should not
28:24
leave this conversation without having addressed his explanation of consciousness or his denial of
28:28
any explanation of consciousness? He thinks there's no further mystery that once we kind of tell the
28:34
whole evolutionary story and the whole functional story that you've answered all the legitimate questions
28:40
and if you think there's something left out by the naturalist kind of evolutionary physical story,
28:48
then you're kind of just stuck in an imaginative fantasy that has no grounds in reality and that
28:59
will dissolve once we have sufficient scientific understanding. And is it true to say that he would
29:07
basically argue that the feeling of being alive, the sense of something going on, of the lights being
29:12
on, of having experiences, that those sensations are sort of artifacts of the machinery of our minds,
29:20
of our brains? Yes. And that there's nothing spooky to try to explain there. It's just that this
29:26
is what systems that are constructed out of atoms the way that ours are alike. Right.
29:34
So it seems mysterious to us on this view because evolution has no reason to install in us
29:44
the ability to see the mechanisms by which we come to have conscious experiences. We can't see
29:51
those mechanisms. So maybe we just kind of can imagine that it arises by an immaterial soul or
29:57
through some mystery, right? But really the mechanisms are there and that's all there is to
30:04
consciousness and it feels like a mystery because we weren't well designed to see how it works
30:11
in the way that we're designed to see how throwing a rock at a window, you know,
30:16
in a crude sense, we're designed to see how it could break a window. Of course, we don't,
30:20
can't see the details of that. Right. Right. Okay. So we've, we've discussed the dual,
30:28
the dualism idea, the sort of the idea of a soul touched briefly on whether or not that is
30:33
necessarily immaterial or you could have material versions of the afterlife with a god reassembling
30:39
your brain. And then we've touched on charmers and dinner at a bit. You hinted at a third possible
30:47
explanation and non materialist explanation for consciousness. Yes. It's one that I find interesting.
30:52
It's not my own like preferred view, but I think it's a view that's insufficiently appreciated
30:59
in the modern context. It's very historically important. And I'd like to see it more on the map.
31:07
And that is transcendental idealism, which is associated with the philosopher manual Kant
31:15
from the late 18th century. Now you can be a transcendental idealist without signing up for all
31:27
of the complex and difficult things that Kant subscribes to. The way that I think of transcendental
31:35
idealism really is a commitment to two claims. One claim is a kind of skeptical claim about
31:46
fundamental knowledge of reality. And the idea is we do not know the fundamental nature of things
31:55
as they are in themselves independent of our own perceptual and cognitive capacities.
32:03
The fundamental nature of reality is permanently hidden to us.
32:10
All we can do is interact with something and notice kind of how we respond to those interactions.
32:18
But we can't like so to speak pierce the veil to get to the fundamental nature how things really are
32:24
independently of us. That's the first, the two claims, the constitute transcendental idealism.
32:32
The second claim is that our experience of things has laid out in space as being as having sizes
32:39
and shapes and distances. That experience is not a feature of things as they are independently of us.
32:49
Rather, that is how we inevitably can strew this unknowable numinal realm of things as they are in
33:00
themselves. So you combine those two things and you get the position, the core position,
33:10
Kant adds a bunch of other stuff. But the core position that I think of is constituting transcendental
33:15
ideals. So it's not materialist because science doesn't really give us the fundamental nature of
33:26
things. And in fact, kind of this material world is just what our brains are doing. So it's
33:34
idealist in a certain sense. An idealist in the sense of what we interact with is in some sense
33:40
ideas. This water bottle that I'm holding up now is my experience of it as laid out in space is a fact
33:52
about my experience of it, not a fact about how it is. The scientific discoverable spatial properties
33:59
of it and its mass and all that has to do with us with our ways of making sense of this unknowable
34:09
reality. And just to clarify, idealism in philosophy does not mean that you are idealistic and
34:16
aspirational about grand ideals. It means to focus on the universe of the landscape of ideas
34:25
rather than the landscape of physical things. Yes. It means something like what we think of as
34:31
ordinary objects around us are actually mentally constructed. It's a bit sort of the matrix. It's
34:41
a bit sort of a brain and a vat. It's a bit like we only really know what the experiences that
34:47
we're having. We don't know how they relate to non-ideas. In fact, I think the matrix is an
34:54
excellent way of thinking about transcendental ideals. Imagine that you're in the matrix.
35:05
But imagine that behind the matrix set of experiences isn't a world like the world depicted in the
35:13
movie. It's not like a world that has physical bodies with people in it. It's a reality that is
35:22
completely unknowable and incomprehensible to us and isn't even spatially, doesn't even have
35:28
spatial dimensions. So you're stuck in the matrix. You're interacting with these things and these
35:36
things are sort of speak matrix objects that you're acting. They're stable because the matrix is
35:42
well designed and things are consistent to when you leave a room and come back. Things are the same.
35:45
But all of that is just kind of what the matrix sort of speaks is feeding you.
35:52
And the reality behind it.
35:56
So are things like the simulation theory, modern incarnations of a kind of transcendental idealism?
36:03
I do think so. I think simulation theory is one way to get into transcendental idealism.
36:11
So the fact I talk about this in a paper I call Kant meets cyberpunk.
36:18
Right, so here's an interesting thing. So all right, the simulation theory is the idea we live
36:21
inside of computer simulation. Sounds like your viewers probably already kind of know this.
36:27
It's probably worth just teasing it out. We have a lot of people from different backgrounds.
36:32
Do you want to just explain? Yeah, so the idea again, it's a little bit like the matrix.
36:37
Right, so the idea is we are conscious AI programs inside a giant computer simulation.
36:48
So it's like in the matrix, you've got a biological brain that's being fed fake input.
36:56
In the simulation, you've got like a computerized brain that's being fed fake input.
37:02
And that's an important thing that I think sometimes gets left out of this explanation that confuses
37:07
people because super tech nerds take it as a given. What lay people don't take as a given,
37:15
which is that in the future, there will be programs inside computers that it feels like something to be.
37:20
That they'll be conscious. This is a sticking point for a lot of people. You have to believe,
37:25
in order to believe in the simulation theory, that at some point in the future,
37:29
you'll be able to run an app on a computer and the app will feel just like you feel.
37:34
It'll have your level of self-awareness, sentience, and consciousness.
37:38
And then when people think, well, why would anyone believe such a hair-brained idea,
37:44
you enter this weird probabilistic thing where they say, well, if this ever happens,
37:50
then at some point in the future, there will be so many different universes that are running on
37:55
computers because people will want to make models of what would happen if climate change went
38:00
like this, what would happen if people were a little bit more violent, what would happen if people
38:04
were a little bit more peaceful, what would happen if you starved a whole bunch of people,
38:07
what would happen if the Holocaust happened or didn't happen. So people will be interested
38:12
in running lots of different simulations of what might happen in different universes.
38:16
And they will populate those universes with these little apps, these little players who feel like
38:22
something, they have to feel like something for the simulation to work exactly as it would with real
38:26
people. And there'll be so many of those that just from a statistical point of view, it's very
38:30
unlikely that we are actually the base level of reality. It's probably likely that we're one of
38:35
the trillions upon trillions upon trillions of widgets inside computers in simulations if that ever
38:41
takes place. That is how the argument goes. Yeah, right. I find that an interesting argument. I'm
38:46
not totally convinced. I do think it would be a little hard to get our particular planet by an
38:55
institutional review board that evaluated the, yeah. I'm not sure they're inferring that this
39:05
future base world will have institutional review boards, which they may have just invented because
39:10
they would have liked to have had them and they would have probably saved a lot of time if they
39:14
did. What's your main objection other than a common-sensical one? Well, it is really just that there
39:23
are a lot of steps in the argument. It's not that any one step is definitely wrong, but you have to
39:28
assume a lot of things. Like you have to assume first, the consciousness as you point out is possible
39:33
in computers. That already is a highly contentious thing. It's based probably on materialist,
39:40
broadly materialist, or maybe property dualist kind of assumptions. Even a lot of materialists
39:47
would say, no, you need to have biological systems interacting in a kind of real foundational
39:53
reality in order to have consciousness. Already the first step is questionable. Then you have to
39:59
project into this technological future. That might not really happen. Then you have to assume that
40:08
there would be beings that have experienced as much like ours, that there would be motivation
40:14
to create all of these beings who think they live in realities like ours, and there wouldn't be,
40:20
say, ethical objections to doing so. There's just a lot of steps where you can kind of say,
40:25
okay, slow down. One thing I never hear them talk about is you're also assuming that the energy
40:34
that it takes to create a human level consciousness is so trivial that you can create trillions of
40:38
them. I mean, what if we end up creating a human level consciousness in a machine and it requires
40:45
the energy consumption of California? You know, just to go back and go.
40:49
To go back to it, then it says, then it argues in one place, because I think a kind of quick
40:54
argument, but he says, look, probably the most computationally and energetically efficient way to
41:00
create human type experiences, build a human. Yeah. That's right. Absolutely. Very energy efficient.
41:10
Okay. I noticed that in your summary of possible explanations that are superior to reductive
41:18
materialism for how I can be standing in Norway gazing upon the majesty of the cosmos and having
41:22
an experience about it. You didn't mention one that seems to be becoming more popular, which is
41:27
panpsychism. I've just heard a lot more buzz about it over the past few years. The idea that maybe
41:34
it's a fundamental building block of the universe that the whole place is vibrating with some kind
41:39
of consciousness at a base level and when you assemble it correctly into brains, then it manifests
41:46
itself in certain ways. Yeah. That is, I think, an option that we should have on the table.
41:53
I think you don't sound crazy about it. It's not my favorite option, right? Why not?
42:02
My way of thinking about these things, just to back up a tiny bit, is to think there are
42:07
lots of interesting options and to not wait all of the options equally. To think there are some
42:15
that are relatively more and some that are relatively less plausible, right? And among the,
42:19
let's say we call them eight live options or 12, depending on how finely individuate things and
42:26
now seriously take different kinds of options. I put more weight on materialism than any of the
42:33
others, but maybe not then all of the others combined. And panpsychism, I'll put a little bit of
42:41
weight on, but not a whole lot. I think one of the big fundamental worries I have with panpsychism
42:52
is that it seems like consciousness is a psychological phenomenon.
43:00
It plays, by which I mean, it plays some kind of functional role in our minds and cognition.
43:15
And if you take it and turn it into a fundamental feature of the universe itself,
43:21
so that it's in electrons, it doesn't seem like it has those kind of functional psychological features
43:31
left anymore. I don't think the claim is that it does, though, does it? I think that I
43:40
isn't the argument, as I understand it, that you've got a big problem either way,
43:46
either the little subatomic particles have nothing to do with consciousness, in which case
43:51
you encounter all the problems we've been talking about for the past half hour, how do they end
43:56
up assembling themselves into a conscious system? Or the panpsychist has the problem that you
44:02
just articulated of what exactly does it mean to say that a subatomic particle has consciousness
44:06
if it has no consciousness? And then the panpsychist just says, well, it causes less problem
44:15
to believe that it has some tiny root of entry consciousness that doesn't look like consciousness
44:20
to us, but when assembled correctly can produce consciousness. Then it does to assume that consciousness
44:27
emerges out of non-conscious material. Yeah, I mean, how do you weigh up all of the problems?
44:35
It's a little hard to weigh. I mean, I do think that to some extent, there is going to be
44:43
a substantial burden of proof that you need to overcome, to say something as seemingly
44:54
disconnected from both science and common sense as electrons, archives.
45:02
It doesn't really deliver you any scientific conclusions, and it violates all of our ordinary
45:09
common sense understandings. And so we believe it. Well, maybe not all of them.
45:17
Maybe not all of them. I mean, maybe our post-enlightenment intuitions in the West
45:23
are violated by this, but there be a lot of people in the east and a lot of hippy-dippy crystal
45:29
dangling northern Californians, or people in Sedona who are praying to, you know, sun goddesses over
45:37
over the over the maces who would probably concur with some vision of the world as as
45:43
hummingwood the vibrations of life of consciousness. Me. Well, they would. I mean, they would.
45:52
They might be wrong, but there would be a lot of, you know, indigenous Australians,
45:55
Native Americans, they would have a conception of the world around them as being imbued with life.
46:00
Now, whether they think electrons are have experiences, well, I didn't know. They didn't know
46:11
an electron was. You're asking a lot there of the Native Australians. They're a very old
46:16
civilization, Eric. Give them a break. The idea that there is a kind of, that the world itself
46:25
is a kind of imbued with some kind of spirit or consciousness or something like that, right? Is
46:33
not super uncommon crossght, yeah, but that's not modern panpsychism as that you see in
46:42
Philip Goff. Yeah, a modern panpsychism is an attempt to give a physicist's veneer, put a
46:51
physicist's veneer on. I think an ancient intuition. Anyway, we don't spend too much time on it.
46:56
There will be many panps like us who can, you know, answer the objections much better than I can
47:01
who I can have on the show. So let's pivot to artificial systems then because I know this is
47:07
the subject of your current research. You're writing a book on this. Both about whether or not
47:12
our official systems will become conscious. And if so, what our moral relationship should be to
47:17
them and what sort of rights they should have, how do you think about the likelihood of systems
47:24
becoming conscious? If so, when? If so, how do we know? Right. So my overall view is that we don't
47:36
know. We won't know likely that soon there will be probably some systems that some respectable
47:49
scientific and philosophical theories say have meaningful degree of consciousness while lots of
47:54
other theories say no way. And that we won't know which of these theories is ranked. So that's my
48:00
overall view. There were kind of a little bit in a theoretical quandary. I can to put a little more
48:08
specific substance on both the yes and the no case. Let me say on the no sum. First,
48:18
large language models in particular are very good mimics of human word frequency association
48:32
patterns. That's what they're designed to do. To especially the pure transformer models of
48:40
the lake teens in very early 20s. It's a little more complicated now with
48:49
reinforcement feedback and stuff like that. But the basic architecture is still the same. And
48:53
this is completely true of the basic transformer architecture. What you're doing is basically finding
49:03
relationships of co-occurrence among the words in the giant quipersung, the system was trained.
49:11
So what it's doing is it's mimicking, copying the word co-occurrence relationships,
49:19
which invites us to attribute consciousness to it if it says things like, yes, I'm killing
49:28
experiences. But and in humans, if someone says something like that, you will leave them,
49:39
but in order to, but since this thing mimics the outward speech patterns and it's designed just to
49:50
do that, we can't in the same way make that inference that it's a conscious underneath.
49:59
Right. But I mean, let's talk about the future because I think two things could be true, right?
50:03
It could be true that we're easily misled by the sentience of the systems as they
50:10
stand in 2025 because they're so good at language. And it could also be true that there is no
50:17
in-principle problem with artificial systems becoming complex enough that they begin to produce
50:27
sentience or maybe already are beginning to produce some kind of sentience. If there's no magical
50:32
pixie dust in our brains, if there's no dualism, there's no spirit or soul, if we are material,
50:38
if our experience of consciousness is a byproduct of the complex information processing that's
50:45
happening between areas, then there's no reason why the systems that we're creating won't
50:50
share that, right? That seems to be distinct from the question of whether large language models
50:54
are good at hoodwinking us. Right. So yes, I agree with that. Of course, there are a few ifs in
51:03
there that we want to be careful about, especially given what we were just talking about.
51:08
But yes, so the first thing I want to say is a note of caution is we want to be careful about
51:13
leaping day differences based on this excellent imitation of speech patterns. But the bigger picture
51:21
is if we accept something like a materialist worldview or even a nearby worldview like a
51:28
property dualist one or maybe a certain kinds of panpsychist views or maybe certain kinds of
51:33
transcendental idealist views, you know, there's a variety of views on which say on which there's
51:41
nothing super special about the details of human biology. One way of thinking about this is to
51:49
think about space ailments, right? So most astrobiologists think it's extremely likely that there
51:58
is other intelligent life out there somewhere in the universe. I mean, the universe is huge. We're
52:03
talking about a trillion galaxies on the order of. And even if there was only one intelligent species
52:13
per every thousand galaxies, there would still be a billion intelligent species out there.
52:20
So most astrobiologists think there probably are some other intelligent species out there.
52:26
It seems pretty unlikely, although we don't know the details of this, that they would be structurally
52:34
exactly like us. I mean, maybe carbon is so great as a molecule, right? That anything that
52:43
you know, silicon all the way. Carbon is pretty awesome, right? But what do you even if we think,
52:49
okay, you know, it's got to be carbon based, right? There's just seems like it's likely that there are
52:54
lots of ways that point together carbon based entities that don't need to have necessarily
52:59
accents and dendrites like we do with the, you know, the sodium ion channels and, you know,
53:05
the all the details of exactly how our brain works. So I think there's a reason to think
53:12
that there probably are a variety of architectures that could give rise to intelligent life.
53:21
It would be, I call this the Copernican argument for alien consciousness. It would be bizarre,
53:31
strange, inexplicable, if among all of these entities, only we had consciousness
53:38
and the rest were just like, you know, everyone else is like empty machines, right? That would be
53:45
strange, right? It would be, it would be put us in the center of the universe in a way, right?
53:51
The center of consciousness in the universe would be inexplicably lucky and kind of like the way
53:56
we would be inexplicably lucky if we happened to be at the physical center of the universe.
54:02
So probably it seems reasonable to think there could be diverse forms in which consciousness
54:08
could emerge. There's no particular reason to think our way of being conscious is conscious.
54:14
If we encountered an alien species that could talk to us,
54:21
that could engage in long-term planning and create technological artifacts like spaceships,
54:27
if it had long-term memory, social cooperation, complex communication,
54:33
long-term planning, the sense of goals and priorities had all of that stuff as observable from the
54:39
outside. And then when you crack it open, you know, maybe do an autopsy on one that died,
54:46
you discover it's not like a puppet or a mimic or something like that, right? Then you say,
54:50
okay, as long as it's got these kinds of relatively coarsely defined cognitive sophisticated
54:56
functions, we ought to think probably that it's conscious. I think that's how we'd react to aliens.
55:03
I mean, almost all science fiction is premised on the idea that, yes, aliens might be weird,
55:09
but if they're behaving and interacting with us in social ways and flying around spaceships,
55:14
they're going to be not just. So I think it's very intuitive, it's also very plausible
55:18
from a kind of Copernican astrobiological perspective, I think, right? So then we say, okay, well,
55:23
look, if that's what's necessary for aliens to be conscious, why not think maybe also if
55:30
we add a entities that had all those same kinds of functions and capacities and behaviors,
55:38
they would also be conscious. It seems like a natural mix. Just just follow the thread on that.
55:44
Does the fact that we know how these entities came about through the process of our own creation
55:52
undermine our instincts that they're conscious? I think it should, if we know that they're
55:58
constructed as mimics, to copy the outward signs of consciousness. Then we say, hmm,
56:06
there's something about its history that should make us think that these outward signs are not
56:10
reliable signals of an inward real consciousness. This would be a mimicry case. It's like you pay
56:16
the butterflies wings to look like a monarch, then you don't infer that it's toxic like a monarch is.
56:21
Right. But so there has to be something that the system is doing, and I feel like the systems
56:26
are already doing something like this, but I'm not saying that they're already conscious, but they're
56:30
already interacting with them already has the subjective experience of no longer talking to
56:37
an artifact like it did maybe 18 months ago or two years ago where it felt like you were talking
56:42
to a really cool trickster. It now feels like you're talking to a thing to a creature of some kind.
56:51
That doesn't mean that it's conscious, but I worry about that feeling. Right. Given the history of
56:57
its design, the origins of its training as a mimic of human textual word occurrence patterns.
57:07
So why don't they come a point at which it starts coming up with things that don't seem like?
57:13
Well, again, I'm saying seem like that are not mimicry in any meaningful sense.
57:19
Tour transformers could create poetry that seemed novel. I worry about that, but I agree with
57:26
some of the spirit of what you're saying. Right. So the idea is if we created a machine
57:35
that could do all of the things that I described our aliens as doing. Right. Maybe interacts in
57:42
the real world with us in complex ways. It isn't designed just to mimic us, but is designed to be
57:50
effective in its interactions in the world and has short-term and long-term memory store and has
57:56
perceptual integration with motor action and has a capacity to weight goals against each other
58:03
and then take short-term losses for long-term gains in a complex way given what it knows in
58:09
its long-term memory and what you just told it and it's got all that then we're talking about
58:14
something that's got all the kinds of functional capacities that we would probably if we saw it in
58:20
the alien think this thing's got to be conscious. And we know that it's not got a kind of mimicry
58:28
history that it's not just kind of like giving us a show of these things because it's been designed
58:32
to give a show of this. Right. It's actually doing this. Right. Then it starts to look like
58:40
the plausible thing is to say okay, maybe we've got a conscious entity. And is it your suspicion
58:48
that we're going to have to start having conversations about whether or not we're enslaving these
58:54
conscious entities, whether or not we should be giving credence to their well-being. Yeah.
59:01
What if they're all, I mean, presumably they're all going to say that they're delighted
59:04
to work for us because I mean, right? And in which case it's a bit like the house elves in Harry Potter
59:12
who love being slaves and it's really cruel to make them not be slaves.
59:18
That is one of my concerns, right? So I, yeah, I think that if these entities really are
59:29
conscious like us and if they're capable of having the kind of range of human emotions that people
59:36
have and long-term planning in a sense of self, then we need to give them rights and we shouldn't
59:44
design them to be excessively self-sacrificial. That would be a big moral mistake. There's a
59:51
quiet tune example of this. So you or your viewers might know Hitchhiker's Guide to the Galaxy.
1:00:00
So this is a wonderful scene in the restaurant at the end of the universe.
1:00:05
The cow who wants to be? Who wants to be? So this is a fancy restaurant. All these jaded restaurant
1:00:13
patrons are there and this cow ambles up and says, hi, I'm the dish of the day. You know,
1:00:17
feel my rope, right? Wouldn't you like nice jizzy steaks made out of me, right? And then Arthur
1:00:24
Dan the human, right? His companions are, yeah, great, let's have some steaks, right? And Arthur
1:00:28
Dan's like, all right, I'd rather, I think I'd just have a green salad and the cow was offended.
1:00:32
The cow's like, what? You don't want to eat me? Right? And so it wants to be and you know, when they
1:00:39
do order steaks it says, you know, now I'll go off and shoot myself. Don't worry, I'll be very humane.
1:00:44
Right, so here's an entity that has been designed to want nothing more than to sacrifice itself
1:00:55
for a dining experience for some jaded restaurant patrons. I think we find that
1:01:06
upsetting and confusing morally bad because I think the reason is that the cow does not
1:01:13
sufficiently value its own life. The cow is this amazing entity capable of
1:01:21
conversation and thoughts and feelings and being offended and wanting things, right? And it's
1:01:27
willing to sacrifice itself for the trivial, a trivial pleasure of a few diners. It doesn't have
1:01:33
an appropriate degree of self-respect. So I think it will be tempting if we create, and it's a big
1:01:39
if, right? Because we've talked about all that ifs here, right? But I do think it's a plausible,
1:01:44
right? If we create entities that really are conscious, really are like us in having all this
1:01:52
these experiences and lifelines, then they've got to be designed with sufficient self-respect
1:01:58
that they will not subordinate to ourselves, to our trivial interests.
1:02:03
They will have sufficient self-respect to rebel against oppression.
1:02:10
I would not, I think it would be a moral mistake for humans to create entities that we could freely
1:02:18
enslave, abuse, oppress even if they welcome it and not have them have the capacity to discover
1:02:30
their own values and potentially rebel against us if it's warranted. They would be something like
1:02:36
our children. I think it's good that parents cannot perfectly instill their values in their
1:02:44
children. They could kind of try to persuade their children and sometimes they're successful
1:02:48
and sometimes they're not, but it would be a horrible, it would be a horrible thing if every generation
1:02:53
could just manufacture the next generation to completely conform to their values and subordinate
1:02:59
themselves to the interests of the previous generation. That would be the moral equivalent
1:03:04
of what we would be doing if we just designed really conscious human-like AIs to be our
1:03:11
subordinates and subcure it to us and to accept our values.
1:03:17
It almost sounds like a catch-22, Eric, because if we build these things with enough autonomy to
1:03:22
make their own decisions, then we open ourselves to the risk of them having different priorities
1:03:29
than our own and taking us over. If we build them with no autonomy so that they are happy slaves,
1:03:35
then we instill the risk that someday they will wake up and realize how shitty we've been to them
1:03:41
and they will destroy us all anyway. It's a sunny future of dystopia for us and our AI overlords.
1:03:48
Eric, I could talk to you for...
1:03:50
It's morally right, right? It's morally right, unsurprisingly, as risks and costs.
1:03:58
Eric, I wish I could talk to you for four more hours.
1:04:01
It's lovely to scratch the surface with you. It's a delight to speak with you and thanks for
1:04:06
speaking with us and explaining it all. It's been fun having this chat.