- The interviewee reflects on how the popularity and interest in AI has changed since their PhD in the 80s
- They worked on logic programming and speeding up query answering in prologue-like languages
- They also studied the frame problem, which involves determining what is relevant and what is not in cognitive processes
- The interviewee later shifted to studying the brain and computational neuroscience, but has now returned to classical AI and machine learning
- The frame problem remains a recurring theme in their work
- The interview is briefly interrupted by a talk from Garry Kasparov on computer chess
- Kasparov discusses the impact of computers on chess and how technology has changed the game
- The interviewee and the video's hosts discuss the influence of computers on chess and Go, and how they are reshaping the way the games are played and analyzed.

- The video discusses the concept of human-machine partnerships in the context of artificial intelligence and consciousness.
- The conversation touches on the idea of creating artificial beings that have lives of their own, referred to as "mind children."
- The film "Ex Machina" is mentioned as an example of exploring the embodiment of AI and the role it plays in human intelligence and consciousness.
- The collaboration between the writer/director of "Ex Machina" and a scientist is discussed, with the scientist providing input on the script and ideas related to AI and consciousness.
- The script of "Ex Machina" is praised for its portrayal of AI and consciousness, with specific lines being highlighted as astute and thought-provoking.

- The author's book is heavily influenced by Wittgenstein, who focuses on the role of words in ordinary life and behavior.
- The protagonist of the movie gradually comes to believe that the AI is conscious through interaction, similar to Wittgenstein's ideas.
- The script for the movie was mostly complete before the author saw it, with only a few changes made.
- There was a scene in the script that was left out in the final film, which would have shown the AI's perspective in an alien way.
- The ending of the film is intentionally ambiguous, leaving questions about the AI's consciousness.
- The author suggested removing a smile from the AI in one scene, but the director felt it was important to show her consciousness.
- The author was shown a clip of the film and was told about an Easter egg hidden in the computer screens.

- The speaker was asked to put a hidden message in a window for a movie promotion
- They wrote a piece of code that computes prime numbers and reveals the ISBN of their book when run
- The code gained attention on the internet and received criticism for its coding style
- The speaker discusses the portrayal of AI in Hollywood films and the differences from reality
- They explain that achieving human-level AI is still uncertain and requires conceptual breakthroughs
- The speaker mentions the laws of robotics by Asimov and their relevance in science fiction and real-life robotics.

- The speaker discusses the difficulty of implementing laws like Asimov's laws into robots
- They mention that current AI technology is not capable of comprehending these laws
- The speaker also mentions the potential dilemma of a robot reasoning that it is human and therefore justified in its actions
- They emphasize that science fiction is good at making us think about these issues, but it is important to remember that it is just a story
- The speaker discusses their interest in deep reinforcement learning and the limitations of current systems
- They recommend searching online for more information on the field and suggest looking into tutorials and MOOCs
- The speaker mentions their involvement in a theater project and an artistic collective.

The speaker discusses their involvement with an artistic collective called Random International. They mention two notable projects by the collective: Rain Room, where clever technology senses your location and turns off the sprinklers above you, and 15 Points, a sculpture that uses point light displays to create the illusion of a person. The speaker reflects on how these projects prompt thoughts about how we perceive machines and the presence of others.

- The interviewee reflects on the changing popularity and interest in AI since the 80s.
- They worked on logic programming and the frame problem in cognitive processes.
- They later shifted to studying the brain and computational neuroscience, but have returned to classical AI and machine learning.
- The interview discusses the impact of computers on chess and Go, and the concept of human-machine partnerships.
- The film "Ex Machina" is mentioned as an example of exploring AI and consciousness.
- The author's book is influenced by Wittgenstein and the portrayal of AI in Hollywood films is discussed.
- The difficulty of implementing laws like Asimov's laws into robots is mentioned.
- The speaker discusses their interest in deep reinforcement learning and their involvement with an artistic collective.

Murray Shanahan was one of the scientific advisors on Ex Machina - https://twitter.com/mpshanahan

He's also a Research Scientist at DeepMind and professor of Cognitive Robotics at Imperial College London.

His book Embodiment and the Inner Life served as inspiration for Alex Garland while he was writing the screenplay for Ex Machina - https://www.amazon.com/Embodiment-inner-life-Cognition-Consciousness/dp/0199226555

YC's podcast is hosted by Craig Cannon - http://twitter.com/craigcannon

So I think the first question I wanted to ask you is that given the popularity of AI, or at least the interest in AI right now, what was it like when you were doing your PhD thesis in the 80s around AI?

I mean, it's quite a surprise for me to find myself in this current position where everyone is interested in what I'm doing.

The media are interested, you know, corporations are interested.

So certainly when I was a PhD student and when I was a young postdoc, it was a fairly niche area.

So you could just kind of like beaver away in your little kind of corner doing things that you thought were intellectually interesting and being reasonably secure that you weren't going to be bothered by anybody.

And so what exactly was the subject matter at the time?

Let me think, what is it, like 30 something years?

I finished 30 years ago, I finished my thesis.

So I was interested in logic programming and prologue type languages.

And I was interested in how you could speed up answering queries in prologue-like languages by keeping a kind of record of the thread of relationships between facts and theorems that you'd already established.

So, instead of having to redo all the computations from scratch, it kind of kept a little collection of the relationships between properties that you'd already worked out so that you didn't have to redo the same computations over again.

So that was the main contribution of the thesis.

I'm amazed I can remember anything about it.

I did my thesis like 5 years ago and I barely remember.

And So did you pursue that further at Imperial?

I kind of, well, 1 other thing that I discussed in my thesis was I had a whole chapter on the frame problem.

So the frame problem is, there are different ways of characterizing it, But the frame problem in its largest guise is all about how a thinking mechanism, or thinking creature, or a thinking machine, if you like, can work out what's relevant and what's not relevant to its ongoing cognitive processes and how it isn't overwhelmed by having to rule out just trivial things that aren't irrelevant.

And so that comes up in a particular guise when you're using logic and when you're using logic to think about actions and their effects.

And there you want to make sure that you don't have to spend a lot of time thinking about the non-effects of action.

So for example, if I move around a bit of the equipment, like your microphone here, then the color of the walls doesn't change.

And you don't want to have to explicitly kind of think about all those kinds of trivial things.

But then more generally, it's all about sort of circumscribing what is relevant to your current situation and what you need to think about and what isn't.

And so how did that translate to what folks are working on today?

Well so it's actually, so this thing, the frame problem has recurred throughout my career.

So although there's been a lot of variation in what I've done, and I've, so I worked for a long time in classical artificial intelligence, which is there, it's all about, it was, and still is all about using logic-like or sentence-like representations of the world.

And you have mechanisms for reasoning about those sentences and a rule-based approach.

And so that approach of classical AI has fallen out of favor a little bit.

And I sort of got a bit disillusioned with it back in, well, a long time ago.

So by kind of the turn of the millennium, I'd more or less abandoned classical AI because I didn't think it was moving towards what we now call AGI, artificial general intelligence, the big vision of human level AI.

And so I thought, well, I'm going to study the brain instead, because that's the example that we have of an intelligent thinking thing.

So I want to try and understand the brain a bit more.

So I started working on building computational neuroscience style models of the brain and thinking about the brain from a larger kind of perspective and thinking about consciousness and the architecture of the brain and big questions.

And now I'm getting around to answering your question by the way, eventually, but now I'm interested in machine learning.

There's been this resurgence of interest in machine learning.

So I've kind of moved back to some of my interests in artificial intelligence, and I'm not thinking so much about the brain or neuroscience or that kind of empirical work right now.

And I've gone back to some of the old themes that I was interested in, in good old fashioned AI, classical AI.

So that's sort of an interesting trajectory.

Actually the frame problem, interestingly, has been a recurring theme throughout all of that stuff, because it keeps on coming up in 1 guise or another.

So in classical AI, there was the question of how can you write out a set of sentences that represent the world where you don't have to write out a load of sentences that encompass all the trivial things that are irrelevant.

And somehow the brain seems to solve that as well.

The brain seems to manage to focus and attend to only what's relevant to the current situation and ignore all of the rest.

And in contemporary machine learning, there's also this kind of issue as well.

There's also a challenge of being able to build systems, especially if you start to rehabilitate some of these ideas from symbolic AI.

You want to think about how you can build systems that focus on what's relevant in the current situation and ignore things that are not.

For example, if you, if a lot of this, A lot of work here at DeepMind has been done with these Atari retro computer games.

So if you think of a retro computer game like Space Invaders, then if you think about the little invader going across the screen.

It doesn't really matter what color it is.

In fact, it doesn't really matter actually what shape it is either.

What really matters is that it's kind of dropping bombs and you need to get out of the way of these things.

So In a sense, a really smart system would learn that it's not the color that matters, it's not the shape that matters, it's these little shapes that fall out of the thing that matter.

So that's all about kind of working out what's relevant and what's not relevant to solving, getting a good score in the game.

So Garry Kasparov in conversation with Demis Hassabis.

Yeah, he gave a great talk about the history of computer chess and his famous match with Deep Blue.

So yeah, we just had to pop upstairs to watch that, right?

Yeah, kind of 1 of those once in a lifetime things.

It also seems like he got out at the exact right time.

So, Demis, at the beginning of the interview, said that he thought that he was perhaps the greatest chess player of all time.

And so, he was there just at the right time to be knocked out by a computer in a way.

Yeah, and he also said that maybe accurately that any iPhone chess player now is probably better.

I also thought it was interesting that he was saying that just anybody in their living room now can sit and watch 2 grandmasters playing a match and can use their computer to see as soon as they make a mistake and can analyze the match and can follow exactly what's going on.

Whereas in the past, it took expert commentators sometimes days to figure out what was going on when 2 great players were playing.

What struck me was how he was kind of analyzing the current players and how they relied so heavily on the computer, or at least he thinks they rely so heavily on the computer that they're kind of like reshaping their mind.

And that certainly I think is going to be true with Go and with AlphaGo.

So it's been interesting watching the reactions of the top Go players like Lisa Doll and KJ who are very positive in a way about the impact of computers on the game of Go.

And they talk about how AlphaGo and programs like it can help them to explore parts of this universe of Go that they would never otherwise have been able to visit.

And so it's really interesting to hear them speak that way.

Yeah, it seems like they're going to open up just kind of new territories for new kinds of games to actually be created.

Well, so we've already seen that with AlphaGo in the match with Lee Sedol.

So as you probably know, there was a famous move in the second match against Lee Sedol, move 37, where all the commentators, all these sort of 9 dan masters were saying, oh, this is a mistake.

And this is very strange." And then they sort of gradually came to realize that this was a sort of revolutionary kind of tactic to put the stone in that particular rank in that particular time in the game.

And since then, the top Go players have been exploring this kind of play about moving into that sort of territory when the conventional wisdom was that you shouldn't.

I mean, the augmentation in general, I find fascinating across the board.

So he was was very positive about the prospects of human machine partnerships and where humans provide maybe a creative element and machines can be more analytical and so on.

Named after Hans Moravec, the roboticist who wrote some amazing books including Mind Children.

So, he wrote this book called Mind Children And this phrase, mind children, alludes to the possibility that we might create these artifacts that are like children of our mind and that they have sort of lives of their own and they are the children of our minds, you know, it's a challenging idea.

This is an old book, I mean, from the late 80s.

Okay, well then maybe we ought to we ought to segue back into what we were talking about Which is kind of related to your book your 2 books ago

Okay Because that was that was kind of an integral question to the movie ex machina Right, because you you didn't necessarily have to have a person like AI.

And more importantly, you didn't have to have an AI that sort of looked like a person that sort of looked like an attractive female that also looked like a robot, right?

Now, I mean, obviously, to a certain extent, those are things that make for good film.

And so, there are artistic choices and cinematographic choices.

And I mean, in the film, Her, we actually have, of course, a disembodied AI.

And so it's possible to make a film out of disembodied artificial intelligence as well.

But obviously a lot of the plot and what drives the plot forward in Ex Machina is to do with Ava's embodiment

the fact that Caleb is attracted to her and sympathizes and empathizes with her.

But there's also a kind of a philosophical side to it too, which is certainly I think that it, well, no doubt when it comes to human intelligence and human consciousness, our physical embodiment is a huge part of that.

It's where our intelligence originates from because what our brains are really here to do is to help us to navigate and manipulate this complex world of objects in 3D space.

And that is, so our embodiment is an essential fact here.

We've got these hands that we use to manipulate objects and we've got legs that enable us to move around in complicated spaces.

So that's in a sense is what our brains are originally for.

The biological brain is there to make for smarter movement.

And all of the rest of intelligence is a flowering out of that, in a way.

And so did you buy the gel that he showed Caleb in the beginning?

So it's interesting, because the way the film is constructed is that, so Alex Garland, you know, the writer and director, so he sometimes says that the film is set 10 minutes into the future.

It's just, you know, It's like a really meant to be like really a lot like our world, just very slightly into the future.

And so, when you see Nathan's lair, his sort of retreat in the wilderness, there's nothing particularly science fictiony about that.

It's designer-ish and of course it is in fact a real hotel or a real- you can actually go and stay in this place.

And so it doesn't have a particularly futuristic feel.

And almost everything you see is not very futuristic.

It's not like Star Wars, but then there are a few things, a few carefully chosen things that look very futuristic and they're Ava's body.

So the way you can see the insides of her torso and her head.

And then when he shows the brain, which is made of this gel.

And so I think that was a good choice because we don't at the moment know how to make things that are like Ava, that have that kind of level of artificial intelligence.

So that's the point at which you have to go sci-fi, really.

Well, I mean, those like lifelike melding elements, have you watched the new show, the HBO show?

No, I mean, I, yeah, I really, it really is on my to watch list because, but I've heard a lot about it, yeah.

the original with Yul Brynner, but I haven't watched the series, the current series yet.

I mean, I guess it's probably like in the sci-fi canon that you have this basement layer where you create the robots and then they become lifelike through this whole process.

Even if you just watch kind of the opening title credits, It's exactly that.

It's like the very, the 3D printed sinews of the muscles.

And so what I was wondering is, as you were consulting on the show, how much of that were they asking you about?

And were they saying like, is this like remotely 10 minutes in the future or is this 50 years in

Yeah, well it wasn't really like that actually.

I mean I'll tell you the sort of whole story of how the kind of collaboration came about.

So I got this email from Alex Garland, unsolicited email out of the blue.

It's the kind of unsolicited email you really want to get from famous writer, director who wants you to work on a science fiction film.

And he basically said, I read your book Embodiment in the Inner Life and it helped me to kind of crystallize some of the ideas around this script that I'm writing for a film about AI and consciousness.

And do you want to get together and have a chat about it?

So I didn't have to think very hard about that.

And so we got together and had lunch and he sent me the script.

And so I'd read through the script by the time I got to see him.

And he really, he certainly wanted to know whether it sort of felt right from the standpoint of somebody working in the field.

There was nothing, I mean, as a script, it was a great page turner actually.

Because it's interesting being in that position because now ex machina and the image of Ava has become kind of iconic and you see it everywhere.

But of course when I read the script, all of that imagery didn't exist.

So I was reading it, I had to kind of conjure it up in my own head.

he didn't give you any kind of preview of what he was thinking,

So no, because nothing had been, nobody had been cast then at that point.

And I think actually when we met up, if my memory serves me right, He did have a few, he did have some images of some mock-ups from artists of what Ava might look like.

But I hadn't seen it when I'd read the script.

So for me, it was just kind of the script.

And the characters really leapt off the page.

The character of Nathan in particular was really very vivid and you really didn't like this guy just reading the script.

Anyway, so then Alex really wanted to- so I sort of grabbed the title of scientific advisor.

I'm not sure if I ever really was officially a scientific advisor, but Alex really wanted to meet up and talk about these ideas.

He wanted to talk about consciousness and about AI.

And so we met up several times during the course of the filming.

And I think there's very little that I contributed to the film at that point.

And in a sense, perhaps I'd already done my main bit by writing the book.

I mean, there were a few little phrases that I corrected, tiny, tiny things, but otherwise I just thought, you know, great, you know, really, really very good.

And there are some lines in the film that I just thought were so spot on.

Yeah, well, I mean, so a favorite 1 is where Caleb was.

So initially Caleb is told that he's there to be the human component in a Turing test.

And of course it isn't the Turing test, but you know, and Caleb says that pretty quickly.

He says, well, look, you know, it's in the real Turing test, that the judge doesn't see whether it's a human or a machine and so on.

But of course I can see that, you know, Ava's a robot.

And Nathan says, oh yeah, we're way past that.

The whole point here is to show you that she's a robot and see if you still feel she has consciousness.

I thought that was an excellent, really an excellent point, making a very important philosophical point in, you know, in this 1 little line in the middle of a psychological thriller, which is pretty, pretty cool.

I found it very like, yeah, that was really astute.

I was wondering like which, which texts influenced him most when he was writing it and in particular like where you found that your your work had seeds like planted throughout the movie yeah where do you think it was most influential

well it's a good question you know you need to ask him so so certainly that so so my book is very heavily influenced by Wittgenstein.

And in a sense, Wittgenstein is all about when it comes to these deep philosophical questions, in a sense, it's very down to earth.

It's always saying, well, what do we mean by consciousness and intentionality and all these kind of big difficult words?

And Wittgenstein is always taking a step back and saying, well, what are the role of these words in ordinary life?

And the role of these words in ordinary life with something like consciousness is all to do with the actual behavior of the people we see in front of us.

Well, I don't actually go around judging others as conscious.

That's a point that he would make as well.

I just naturally treat them as conscious.

so why do I naturally treat them as conscious?

Because their behavior is such that they're just like fellow creatures and that's just what you do when you encounter a fellow creature.

And so this is an important Wittgensteinian point that I bring out in the book very much.

And in a sense, that's very much what happens to Caleb.

So Caleb isn't sort of sitting making notes saying, therefore she is conscious.

But rather through interacting with her, he just gradually comes to feel that she is conscious just to, and to start treating her as conscious.

And so that's a very, there's something very Wittgensteinian about that.

And then I think probably that comes from, I'd like to think that comes from my book to an extent.

Well, I had never, I guess it seems very cinematic that it would be like over the course of a week, the Turing test, but I had never seen a Turing test framed that way.

I guess it's not, you know, it's a Garland test.

But did you coach him in any way of like the natural steps that someone would take as the test elevates?

So the script was already and the plot was already, the whole script was already, you know, 95% done, you know, when I first saw it.

So there are a few differences in the final film from what you see in the in the script that I saw and indeed in The in the publish.

So that was actually that was a question from Twitter this is kind of a Seemingly a pseudonym on Twitter someone trench shovel They asked were there any parts of the script that were changed or left out because they weren't technically feasible or realistic?

Ah, well, so there was a bit that was in the script that was left out in the final filming, which I think is very significant.

And so it's a right, so spoilers ahead for the few people.

I assume if you're listening to this, you've seen the film.

So right at the end of the film where Ava is climbing into the helicopter, having escaped from the compound and she's about to kind of fly off.

And we see her have a few words with a helicopter pilot and I wonder what she says actually.

Anyway, and then off the helicopter goes.

Now in the written script, there's an instruction there which says something along the lines, oh, we see waveforms and we see the facial recognition vectors fluttering across the screen and we see this, that and the other.

And now in the end, so the very first version of the film that I saw was long before all the VFX had been properly done and everything.

And they put sort of this scene and they start, they put a little bit of this kind of visual effects in.

And then I think they decided this didn't really work terribly well to do that at that point.

So in the version that we see, you don't actually see that, you just see her speaking to the helicopter pilot and she climbs into the helicopter.

But it's a very significant direction because, you know, we're left, I think 1 of the great strengths of the film is that it leaves so many unanswered questions.

You know, you don't really, you're left thinking, is she really conscious?

You know, is, you know, does she really, Is she really capable of suffering?

Is she just a kind of machine that's gone horribly wrong?

Or is she a person who's understandably had to commit this act of violence in order to save herself.

And although I think people are leaning more towards the kind of, oh, she's conscious in a straightforward kind of way, then that's the...

But that version of the ending just points to the fact that there's a real ambiguity there.

Because if that had been shown, you might be leaning more the other way.

You might be thinking, gosh, this is a very alien creature indeed.

See all Y Combinator transcripts on Youtube

Ex Machina's Scientific Advisor - Murray Shanahan