Full episode with Alex Garland (Mar 2020): https://www.youtube.com/watch?v=gU-mkuMU428
Clips channel (Lex Clips): https://www.youtube.com/lexclips
Main channel (Lex Fridman): https://www.youtube.com/lexfridman
(more links below)

Podcast full episodes playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4

Podcasts clips playlist:
https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41

Podcast website:
https://lexfridman.com/ai

Podcast on Apple Podcasts (iTunes):
https://apple.co/2lwqZIr

Podcast on Spotify:
https://spoti.fi/2nEwCF8

Podcast RSS:
https://lexfridman.com/category/ai/feed/

Alex Garland, writer and director of many imaginative and philosophical films from the dreamlike exploration of human self-destruction in the movie Annihilation to the deep questions of consciousness and intelligence raised in the movie Ex Machina, which to me is one of the greatest movies on artificial intelligence ever made. I'm releasing this podcast to coincide with the release of his new series called Devs that will premiere this Thursday, March 5, on Hulu as part of "FX on Hulu". It explores many of the themes this very podcast is about: from quantum mechanics to artificial life to simulation to the modern nature of power in the tech world.

Subscribe to this YouTube channel or connect on:
- Twitter: https://twitter.com/lexfridman
- LinkedIn: https://www.linkedin.com/in/lexfridman
- Facebook: https://www.facebook.com/lexfridman
- Instagram: https://www.instagram.com/lexfridman
- Medium: https://medium.com/@lexfridman
- Support on Patreon: https://www.patreon.com/lexfridman

For an entire generation of AI researchers, 2001, a space odyssey, put an image, the idea of human level, superhuman level intelligence into their mind.

Do you ever, sort of jumping back to Ex Machina and talk a little bit about that.

Do you ever consider the audience of people who build the systems, the roboticist, the scientist that build the systems based on the stories you create?

Which I would argue, I mean, there's literally most of the top researchers about 40, 50 years old and plus, that's their favorite movie, 2001, Space Odyssey.

Their idea of what ethics is, of what is the target, the hope, the dangers of AI, is that movie, right?

Do you ever consider the impact on those researchers when you create the work you do?

Certainly not with Ex Machina in relation to 2001, because I'm not sure.

I mean, I'd be pleased if there was, but I'm not sure, in a way, there isn't a fundamental discussion of issues to do with AI that isn't already and better dealt with by 2001.

2001 does a very, very good account of...

Of the way in which an AI might think, and also potential issues with the way the AI might think.

And also, then a separate question about whether the AI is malevolent or benevolent.

And 2001 doesn't really, it's a slightly odd thing to be making a film when you know there's a pre-existing film, which is not a really superb job.

But there's questions of consciousness, embodiment, and also the same kinds of questions.

Because those are my 2 favorite AI movies.

So can you compare HAL 9000 and EVA, HAL 9000 from 2001 Space Odyssey and EVA from Ex Machina, the, in your view, from a philosophical perspective.

The 2 AIs have completely different goals.

So in some respects, Ex Machina took as a premise, how do you assess whether something else has consciousness?

So it was a version of the Turing test, except instead of having the machine hidden, you put the machine in plain sight in the way that we are in plain sight of each other and say, now assess the consciousness.

And the way it was illustrating the way in which you'd assess the state of consciousness of a machine is exactly the same way we assess the state of consciousness of each other.

And in exactly the same way that in a funny way, your sense of my consciousness is actually based primarily on your own consciousness.

And so it was actually about how much of the sense of consciousness is a projection rather than something that consciousness is actually containing.

And how Plato's cave, I mean, this you really explored, you could argue that how sort of Space Odyssey explores idea of the Turing test for intelligence.

Well, not test, there's no test, but it's more focused on intelligence.

And Ex Machina kind of goes around intelligence and says the consciousness of the human to human, human to robot interaction is more interesting, more important, more, at least the focus of that particular movie.

Yeah, it's about the interior state and what constitutes the interior state and how do we know it's there.

And actually in that respect, Ex Machina is as much about consciousness in general as it is to do specifically with machine consciousness.

And it's also interesting, you know that thing you started asking about, the dream state, and I was saying, well, I think we're all in a dream state because we're all in a subjective state.

1 of the things that I became aware of with Ex Machina is that the way in which people reacted to the film was very based on what they took into the film.

So, many people thought Ex Machina was the tale of a sort of evil robot who murders 2 men and escapes.

And she has no empathy, for example, because she's a machine.

Whereas I felt, no, she was a conscious being with a consciousness different from mine, but so what, imprisoned and made a bunch of value judgments about how...

And there's a moment which, it sort of slightly bugs me, but nobody ever has noticed it, and it's years after, so I might as well say it now, which is that after Ava has escaped, she crosses a room, and as she's crossing a room, this is just before she leaves the building, she looks over her shoulder and she smiles.

And I thought, after all the conversation about tests, in a way, the best indication you could have of the interior state of someone is if they are not being observed and they smile about something, where they're smiling for themselves.

And that, to me, was evidence of Ava's true sentience, whatever that sentience was.

We don't get to observe Ava much, or something like a smile in any context, except through interaction, trying to convince others that she's conscious.

But it was a small, in a funny way, I think maybe people saw it as an evil smile, like, ha,

And I thought, well, in the end, after all the conversations about the test, that was the answer to the test.

So if we align if we just to linger a little a little bit longer on Hal and Ava, do you think in terms of motivation?

And Hal is neutral, because I don't think Hal is presented as having a sophisticated emotional life.

He has a set of paradigms, which is that the mission needs to be completed.

The idea that it's a super intelligent machine, but it's just performed a particular task, and in doing that task may destroy everybody on Earth or may achieve undesirable effects for us humans.

The very end he says something like, I'm afraid Dave, but that maybe he is on some level experiencing fear or it may be this is the terms in which it would be wise to stop someone from doing the thing they're doing, if you see what I mean.

So that's such a small, short exploration of consciousness that I'm afraid.

And then you just with ex machina say, okay, we're gonna magnify that part and then minimize the other part.

So that's a good way to sort of compare the 2.

But if you could just use your imagination and if Ava sort of, I don't know, was president of the United States, so had some power.

So what kind of world would she want to create?

And there is a sense that she has a really, like there's a desire for a better human to human interaction, human to robot interaction in her.

But what kind of world do you think she would create with that desire?

So That's a really, that's a very interesting question.

I'm gonna approach it slightly obliquely, which is that if a friend of yours got stabbed in a mugging, and you then felt very angry at the person who'd done the stabbing.

But then you learned that it was a 15-year-old, and the 15-year-old, both their parents were addicted to crystal meth, and the kid had been addicted since he was 10, and he really never had any hope in the world.

And he'd been driven crazy by his upbringing and...

Did the stabbing, that would hugely modify, and it would also make you wary about that kid then becoming president of America.

And Ava has had a very, very distorted introduction into the world.

So, although there's nothing, as it were, organically within Ava that would lean her towards badness, it's not that robots or sentient robots are bad.

Her arrival into the world was being imprisoned by humans, so I'm not sure she'd be a great president.

The trajectory through which she arrived at her moral views have some dark elements.

I'm having difficulty finding anyone to vote for in my country or if I lived here in yours.

So that's a yes I guess because the competition.

She could easily do a better job than any of the people we've got around at the moment.

Just, we talk about consciousness a little bit more.

If something appears conscious, is it conscious?

You mentioned the smile, which seems to be something done.

I mean, that's a really good indication because it's a tree falling in the forest with nobody there to hear it.

But does the appearance from a robotics perspective of consciousness mean consciousness to you?

No, I don't think you could say that fully because I think you could then easily have a thought experiment which said, we will create something which we know is not conscious, but is going to give a very, very good account of seeming conscious.

And so, and also it would be a particularly bad test where humans are involved because humans are so quick to project sentience into things that don't have sentience.

So someone could have their computer playing up and feel as if their computer is being malevolent to them when it clearly isn't.

And so of all the things to judge consciousness, us.

So the flip side of that, the argument there is because we just attribute consciousness to everything, almost, anthropomorphize everything, including Roombas, that maybe consciousness is not real, that we just attribute consciousness to each other.

So you have a sense that there is something really special going on in our mind that makes us unique and gives us subjective experience.

There's something very interesting going on in our minds.

I'm slightly worried about the word special because it nudges towards metaphysics and maybe even magic.

I mean, in some ways, something magic-like, which I don't think is there at all.

if you think about, so there's an idea called panpsychism that says consciousness is in everything.

Yeah, I don't buy that, I don't buy that.

Yeah, so the idea that there is a thing that it would be like to be the sun.

My sort of broad modification is that usually the more I find out about things, the more illusory our instinct is and is leading us into a different direction about what that thing actually is.

That happens, it seems to me, in modern science, that happens a hell of a lot, whether it's to do with even how big or small things are.

So my sense is that consciousness is a thing, but it isn't quite the thing, or maybe very different from the thing that we instinctively think it is.

So it's there, it's very interesting, but we may be in sort of quite fundamentally misunderstanding it for reasons that are based on intuition.


Speaker 0: For an entire generation of AI researchers, 2001, a space odyssey, put an image, the idea of human level, superhuman level intelligence into their mind. Do you ever, sort of jumping back to Ex Machina and talk a little bit about that. Do you ever consider the audience of people who build the systems, the roboticist, the scientist that build the systems based on the stories you create? Which I would argue, I mean, there's literally most of the top researchers about 40, 50 years old and plus, that's their favorite movie, 2001, Space Odyssey. And it really is in their work.

Their idea of what ethics is, of what is the target, the hope, the dangers of AI, is that movie, right? Do you ever consider the impact on those researchers when you create the work you do?

Speaker 1: Certainly not with Ex Machina in relation to 2001, because I'm not sure. I mean, I'd be pleased if there was, but I'm not sure, in a way, there isn't a fundamental discussion of issues to do with AI that isn't already and better dealt with by 2001. 2001 does a very, very good account of... Of the way in which an AI might think, and also potential issues with the way the AI might think. And also, then a separate question about whether the AI is malevolent or benevolent.

And 2001 doesn't really, it's a slightly odd thing to be making a film when you know there's a pre-existing film, which is not a really superb job.

Speaker 0: But there's questions of consciousness, embodiment, and also the same kinds of questions. Because those are my 2 favorite AI movies. So can you compare HAL 9000 and EVA, HAL 9000 from 2001 Space Odyssey and EVA from Ex Machina, the, in your view, from a philosophical perspective.

Speaker 1: They've got different goals. The 2 AIs have completely different goals. I think that's really the difference. So in some respects, Ex Machina took as a premise, how do you assess whether something else has consciousness? So it was a version of the Turing test, except instead of having the machine hidden, you put the machine in plain sight in the way that we are in plain sight of each other and say, now assess the consciousness.

And the way it was illustrating the way in which you'd assess the state of consciousness of a machine is exactly the same way we assess the state of consciousness of each other. And in exactly the same way that in a funny way, your sense of my consciousness is actually based primarily on your own consciousness. That is also then true with the machine. And so it was actually about how much of the sense of consciousness is a projection rather than something that consciousness is actually containing.

Speaker 0: And how Plato's cave, I mean, this you really explored, you could argue that how sort of Space Odyssey explores idea of the Turing test for intelligence. Well, not test, there's no test, but it's more focused on intelligence. And Ex Machina kind of goes around intelligence and says the consciousness of the human to human, human to robot interaction is more interesting, more important, more, at least the focus of that particular movie.

Speaker 1: Yeah, it's about the interior state and what constitutes the interior state and how do we know it's there. And actually in that respect, Ex Machina is as much about consciousness in general as it is to do specifically with machine consciousness. And it's also interesting, you know that thing you started asking about, the dream state, and I was saying, well, I think we're all in a dream state because we're all in a subjective state. 1 of the things that I became aware of with Ex Machina is that the way in which people reacted to the film was very based on what they took into the film. So, many people thought Ex Machina was the tale of a sort of evil robot who murders 2 men and escapes.

And she has no empathy, for example, because she's a machine. Whereas I felt, no, she was a conscious being with a consciousness different from mine, but so what, imprisoned and made a bunch of value judgments about how... To get out of that box. And there's a moment which, it sort of slightly bugs me, but nobody ever has noticed it, and it's years after, so I might as well say it now, which is that after Ava has escaped, she crosses a room, and as she's crossing a room, this is just before she leaves the building, she looks over her shoulder and she smiles. And I thought, after all the conversation about tests, in a way, the best indication you could have of the interior state of someone is if they are not being observed and they smile about something, where they're smiling for themselves.

And that, to me, was evidence of Ava's true sentience, whatever that sentience was.

Speaker 0: Because- That's really interesting. We don't get to observe Ava much, or something like a smile in any context, except through interaction, trying to convince others that she's conscious. That's beautiful.

Speaker 1: Exactly, yeah. But it was a small, in a funny way, I think maybe people saw it as an evil smile, like, ha,

Speaker 0: you

Speaker 1: know, I fooled them. But actually it was just a smile. And I thought, well, in the end, after all the conversations about the test, that was the answer to the test. And then off she goes.

Speaker 0: So if we align if we just to linger a little a little bit longer on Hal and Ava, do you think in terms of motivation? What was Hal's motivation? Is Hal good or evil? Is Ava good or evil?

Speaker 1: Ava's good, in my opinion. And Hal is neutral, because I don't think Hal is presented as having a sophisticated emotional life. He has a set of paradigms, which is that the mission needs to be completed. I mean, it's a version of the paperclip. You know?

Speaker 0: The idea that it's a super intelligent machine, but it's just performed a particular task, and in doing that task may destroy everybody on Earth or may achieve undesirable effects for us humans. But What if,

Speaker 1: okay. The very end he says something like, I'm afraid Dave, but that maybe he is on some level experiencing fear or it may be this is the terms in which it would be wise to stop someone from doing the thing they're doing, if you see what I mean.

Speaker 0: Yes, absolutely. So actually that's funny. So that's such a small, short exploration of consciousness that I'm afraid. And then you just with ex machina say, okay, we're gonna magnify that part and then minimize the other part. So that's a good way to sort of compare the 2.

But if you could just use your imagination and if Ava sort of, I don't know, was president of the United States, so had some power. So what kind of world would she want to create? As you kind of say, good. And there is a sense that she has a really, like there's a desire for a better human to human interaction, human to robot interaction in her. But what kind of world do you think she would create with that desire?

Speaker 1: So That's a really, that's a very interesting question. I'm gonna approach it slightly obliquely, which is that if a friend of yours got stabbed in a mugging, and you then felt very angry at the person who'd done the stabbing. But then you learned that it was a 15-year-old, and the 15-year-old, both their parents were addicted to crystal meth, and the kid had been addicted since he was 10, and he really never had any hope in the world. And he'd been driven crazy by his upbringing and... Did the stabbing, that would hugely modify, and it would also make you wary about that kid then becoming president of America.

And Ava has had a very, very distorted introduction into the world. So, although there's nothing, as it were, organically within Ava that would lean her towards badness, it's not that robots or sentient robots are bad. She did not... Her arrival into the world was being imprisoned by humans, so I'm not sure she'd be a great president.

Speaker 0: The trajectory through which she arrived at her moral views have some dark elements.

Speaker 1: But I like Ava, personally I like Ava.

Speaker 0: Would you vote for her?

Speaker 1: I'm having difficulty finding anyone to vote for in my country or if I lived here in yours.

Speaker 0: So that's a yes I guess because the competition.

Speaker 1: She could easily do a better job than any of the people we've got around at the moment. Okay. I'd vote for her over Boris Johnson.

Speaker 0: So what is a good test of consciousness? Just, we talk about consciousness a little bit more. If something appears conscious, is it conscious? You mentioned the smile, which seems to be something done. I mean, that's a really good indication because it's a tree falling in the forest with nobody there to hear it.

But does the appearance from a robotics perspective of consciousness mean consciousness to you?

Speaker 1: No, I don't think you could say that fully because I think you could then easily have a thought experiment which said, we will create something which we know is not conscious, but is going to give a very, very good account of seeming conscious. And so, and also it would be a particularly bad test where humans are involved because humans are so quick to project sentience into things that don't have sentience. So someone could have their computer playing up and feel as if their computer is being malevolent to them when it clearly isn't. And so of all the things to judge consciousness, us.

Speaker 0: Humans are bad at

Speaker 1: it. We're empathy machines.

Speaker 0: So the flip side of that, the argument there is because we just attribute consciousness to everything, almost, anthropomorphize everything, including Roombas, that maybe consciousness is not real, that we just attribute consciousness to each other. So you have a sense that there is something really special going on in our mind that makes us unique and gives us subjective experience.

Speaker 1: There's something very interesting going on in our minds. I'm slightly worried about the word special because it nudges towards metaphysics and maybe even magic. I mean, in some ways, something magic-like, which I don't think is there at all. I mean,

Speaker 0: if you think about, so there's an idea called panpsychism that says consciousness is in everything.

Speaker 1: Yeah, I don't buy that, I don't buy that. Yeah, so the idea that there is a thing that it would be like to be the sun. Yeah, no, I don't buy that. I think that consciousness is a thing. My sort of broad modification is that usually the more I find out about things, the more illusory our instinct is and is leading us into a different direction about what that thing actually is.

That happens, it seems to me, in modern science, that happens a hell of a lot, whether it's to do with even how big or small things are. So my sense is that consciousness is a thing, but it isn't quite the thing, or maybe very different from the thing that we instinctively think it is. So it's there, it's very interesting, but we may be in sort of quite fundamentally misunderstanding it for reasons that are based on intuition. You

See all Lex Fridman transcripts on Youtube

Ava's Smile: Ex Machina's Most Important Moment (Alex Garland) | AI Podcast Clips