back to indexRodney Brooks: Robotics | Lex Fridman Podcast #217
link |
The following is a conversation with Rodney Brooks, one of the greatest roboticists in history.
link |
He led the Computer Science and Artificial Intelligence Laboratory at MIT,
link |
then cofounded iRobot, which is one of the most successful robotics companies ever.
link |
Then he cofounded Rethink Robotics that created some amazing collaborative robots like Baxter
link |
and Sawyer. Finally, he cofounded Robust.ai, whose mission is to teach robots common sense,
link |
which is a lot harder than it sounds. To support this podcast, please check out our sponsors
link |
in the description. As a side note, let me say that Rodney is someone I've looked up to for many years,
link |
in my now over two decade journey in robotics, because, one, he's a legit great engineer of
link |
real world systems, and two, he's not afraid to state controversial opinions that challenge
link |
the way we see the AI world. But of course, while I agree with him on some of his critical views of
link |
AI, I don't agree with some others. And he's fully supportive of such disagreement. Nobody ever built
link |
anything great by being fully agreeable. There's always respect and love behind our interactions.
link |
And when a conversation is recorded, like it was for this podcast, I think a little bit of
link |
disagreement is fun. This is the Lex Friedman podcast. And here is my conversation with Rodney Brooks.
link |
What is the most amazing or beautiful robot that you've ever had the chance to work with?
link |
I think it was Domo, which was made by one of my grad students, Aaron Edsinger. It now sits in
link |
Daniela Roos's office, director of CSAIL. And it was just a beautiful robot. And Aaron was really
link |
clever. He didn't give me a budget ahead of time. He didn't tell me what he was going to do. He just
link |
started spending money. He spent a lot of money. He and Jeff Weber, who was a mechanical engineer,
link |
who Aaron insisted he bring with him when he became a grad student, built this beautiful,
link |
gorgeous robot Domo, which is an upper torso humanoid, two arms with fingers, three finger
link |
hands, and face eyeballs. Not the eyeballs, but everything else, serious elastic actuators.
link |
You can interact with it. Cable driven, all the motors are inside. And it's just gorgeous.
link |
The eyeballs are actuated too or no? Oh yeah, the eyeballs are actuated with cameras.
link |
And so it had a visual attention mechanism, looking when people came in and looking in their face
link |
and talking with them. Why was it amazing? The beauty of it. You said what was the most beauty?
link |
What is the most beautiful? It's just mechanically gorgeous. As everything Aaron builds has always
link |
been mechanically gorgeous. It's just exquisite in the detail. We're talking about mechanically,
link |
literally the amount of actuators. The actuators, the cables. He anodizes different parts,
link |
different colors, and it just looks like a work of art. What about the face? Do you find the face
link |
beautiful in robots? When you make a robot, it's making a promise for how well it will be able
link |
to interact. So I always encourage my students not to overpromise. Even with its essence,
link |
like the thing it presents, it should not overpromise. Yeah, so the joke I make, which I think
link |
you'll get, is if your robot looks like Albert Einstein, it should be as smart as Albert Einstein.
link |
So the only thing in Domo's face is the eyeballs. And because that's all it can do,
link |
it can look at you and pay attention. And so there is no, it's not like one of those
link |
Japanese robots that looks exactly like a person at all. But see, the thing is,
link |
us humans and dogs, too, don't just use eyes as attentional mechanisms. They also use it to
link |
communicate as part of the communication. Like a dog can look at you, look at another thing,
link |
and look back at you. And that designates that we're going to be looking at that thing together.
link |
Yeah, or intent. And on both Baxter and Sawyer at Rethink Robotics, they had a screen
link |
with, you know, graphic eyes. So it wasn't actually where the cameras were pointing,
link |
but the eyes would look in the direction it was about to move its arm. So people in the factory
link |
nearby were not surprised by its motions because it gave that intent away. Before we talk about
link |
Baxter, which I think is a beautiful robot, let's go back to the beginning. When did you first
link |
fall in love with robotics? We're talking about beauty and love to open the conversation. This
link |
is great. I was born in the end of 1954. I grew up in Adelaide, South Australia. And I have these
link |
two books that are dated 1961. So I'm guessing my mother found them in a store in 62 or 63.
link |
How and Why Wonder Books? How and Why Wonder Book of Electricity? And How and Why Wonder Book of
link |
Giant Brains and Robots. And I learned how to build circuits, you know, when I was eight or nine,
link |
simple circuits. And I read, you know, I learned the binary system and saw all these drawings
link |
mostly of robots. And then I tried to build them for the rest of my childhood. Wait, 61, you said?
link |
This was when the two books, I've still got them at home. What does the robot mean in that
link |
context? No, they were some of the robots that they had were arms, you know, big arms to move
link |
nuclear material around. But they had pictures of welding robots that look like humans under the
link |
sea welding stuff underwater. So they weren't real robots. But they were, you know, what people
link |
were thinking about for robots? What were you thinking about? Were you thinking about humanoids?
link |
Were you thinking about arms with fingers? Were you thinking about faces or cars? No,
link |
actually, to be honest, I realized my limitation on building mechanical stuff. So I just built
link |
the brains mostly out of different technologies as I got older. I built a learning system, which
link |
was chemical based. And I had this ice cube tray each. Well, was a cell. And by applying
link |
voltage to the two electrodes, it would build up a copper bridge. So over time, it would learn
link |
a simple network. So I could teach it stuff. And that was mostly things were driven by my budget.
link |
And nails as electrodes and ice cube tray was about my budget at that stage. Later,
link |
I managed to buy a transistors. And then I could build gates and flip flops and stuff.
link |
So one one of your first robots was an ice cube tray. Yeah.
link |
And it was very cerebral because it went to add. Very nice. Well,
link |
just a decade or so before in 1950, Alan Turing wrote the paper that formulated the Turing test.
link |
And he opened that paper with the question, can machines think? So let me ask you this question.
link |
Can machines think? Can your ice cube tray one day think?
link |
Certainly machines can think because I believe you're a machine and I'm a machine and I believe
link |
we both think. I think any other philosophical position is sort of a little ludicrous. What
link |
does think mean if it's not something that we do? And we are machines. So yes, machines can,
link |
but do we have a clue how to build such machines? That's a very different question.
link |
Are we capable of building such machines? Are we smart enough? We think we're smart enough to do
link |
anything, but maybe we're not. Maybe we're just not smart enough to build. It's not like us.
link |
The kind of computer that Alan Turing was thinking about, do you think there is something
link |
fundamentally or significantly different between the computer between our ears,
link |
the biological computer that humans use, and the computer that he was thinking about from a
link |
sort of high level philosophical? Yeah, I believe that it's very wrong. In fact,
link |
I'm halfway through a, I think it'll be about a 480 page book titled, the working title is
link |
Not Even Wrong. And if I may, I'll tell you a bit about that book. So there's two three thrusts
link |
to it. One is the history of computation, what we call computation, goes all the way back to
link |
some manuscripts in Latin from 1614 and 1620 by Napier and Kepler through Babbage and Lovelace.
link |
And then Turing's 1936 paper is what we think of as the invention of modern computation.
link |
And that paper, by the way, did not set out to invent computation. It set out to
link |
negatively answer one of Hilbert's three later set of problems. He called it
link |
as an effective way of getting answers. And Hilbert really worked with rewriting rules as did
link |
a church who also, at the same time, a month earlier than Turing, disproved Hilbert's
link |
one of these three hypotheses. The other two had already been disproved by Godel.
link |
So Turing set out to disprove it because it's always easier to disprove these things than to
link |
prove that there is an answer. And so he needed, and it really came from his
link |
professor, I was an undergrad at Cambridge who said, who turned it into, is there a mechanical
link |
process? So he wanted to show a mechanical process that could calculate numbers because
link |
that was a mechanical process that people used to generate tables. They were called computers,
link |
the people at the time. And they followed a set of rules where they had paper,
link |
and they would write numbers down. And based on the numbers, they'd keep writing other numbers.
link |
And they would produce numbers for these tables, engineering tables, that the more iterations
link |
they did, the more significant digits came out. And so Turing, in that paper, set out to define
link |
what sort of machine could do that mechanical machine, where it could produce an arbitrary
link |
number of digits in the same way a human computer did. And he came up with a very simple set of
link |
constraints where there was an infinite supply of paper. This is the tape of the Turing machine.
link |
And each Turing machine had a set, came with a set of instructions that, as a person could do
link |
with pencil and paper, write down things on the tape and erase them and put new things there. And
link |
he was able to show that that system was not able to do something that Hilbert hypothesized. So he
link |
disproved it. But he had to show that this system was good enough to do whatever could be done,
link |
but couldn't do this other thing. And there he said, and he says in the paper,
link |
I don't have any real arguments for this, but based on intuition. So that's how he defined
link |
computation. And then if you look over the next from 1936 up until really around 1975,
link |
you see people struggling with, is this really what computation is? And so Marvin Minsky,
link |
very well known in AI, but also a fantastic mathematician in his book, Finite and Infant
link |
Machines from the mid 60s, which is a beautiful, beautiful mathematical book,
link |
says at the start of the book, well, what is computation? Turing says this and,
link |
yeah, I sort of think it's that. It doesn't really matter whether the stuff's made of wood
link |
or plastic. It's just relatively cheap stuff can do this stuff. And so, yeah, seems like computation.
link |
And Donald Knuth, in his first volume of his Art of Computer Programming in around 1968,
link |
says, well, what's computation? It's the stuff like Turing says that a person could do each step
link |
without too much trouble. And so one of his examples of what would be too much trouble was
link |
a step which required knowing whether Fermat's last theorem was true or not, because it was not
link |
known at the time. And that's too much trouble for a person to do as a step. And Hopcroft and
link |
Oldman sort of said a similar thing later that year. And by 1975, in the Aho Hopcroft and Oldman
link |
book, they're saying, well, you know, we don't really know what computation is, but intuition
link |
says this is sort of about right. And this is what it is. That's computation. It says sort of
link |
agreed upon thing, which happens to be really easy to implement in silicon. And then we had
link |
Moore's Law, which took off and it's been an incredibly powerful tool. Certainly wouldn't
link |
argue with that. The version we have of computation, incredibly powerful.
link |
Can we just take a pause? So what we're talking about is there's an infinite tape with some
link |
simple rules of how to write on that tape. And that's that's what we're kind of thinking about.
link |
This is computation. Yeah. And it's modeled after humans, how humans do stuff. And
link |
I think it's a, Turing says in the 36 paper, one of the critical facts here is that a human
link |
has a limited amount of memory. So that's what we're going to put onto our mechanical computers.
link |
So, you know, I'm like mass, I'm like mass or charge or, you know, it's not, it's not given
link |
by the universe. It was, this is what we're going to call computation. And then it has this really,
link |
you know, it had this really good implementation, which has completely changed our technological
link |
work. That's computation. Second part of the book, I, or argument in the book, I have this
link |
two by two matrix with science in the top row, engineering the bottom row, left column is
link |
intelligence, right column is life. So in the bottom row, the engineering, there's artificial
link |
intelligence and there's artificial life. In the top row, there's neuroscience and
link |
abiogenesis. How does living matter turn in, how does nonliving matter become living matter?
link |
Yes. Four disciplines. These four disciplines all came into the current form in the period
link |
1945 to 1965. That's interesting. There was neuroscience before, but it wasn't
link |
effective neuroscience. It was, you know, there was ganglion and there's electrical charges,
link |
but no one knows what to do with it. And furthermore, there are a lot of players who are common
link |
across them. I've identified common players except for artificial intelligence and abiogenesis,
link |
I don't have, but for any other pair, I can point to people who work in and a whole bunch of them
link |
by the way. We're at the research lab for electronics at MIT, where Warren McCulloch held
link |
forth. And in fact, McCulloch, Pitts, Letvin and Maturana wrote the first paper on functional
link |
neuroscience called What the Frog's Eye Tells the Frog's Brain, where instead of it just being
link |
this bunch of nerves, they sort of showed what different anatomical components that
link |
anatomical components were doing and telling other anatomical components and, you know,
link |
generating behavior in the frog. Would you put them as basically the fathers or one of the early
link |
pioneers of what are now called artificial neural networks?
link |
Yeah. I mean, McCulloch and Pitts. Pitts was much younger than him. In 1943,
link |
had written a paper inspired by Bertrand Russell on a calculus for the ideas imminent in
link |
neural systems, where they had tried to, without any real proof, they had tried to give a formalism
link |
for neurons, basically in terms of logic and gates or gates and not gates, with no real evidence
link |
that that was what was going on. But they talked about it. And that was picked up by Minsky for
link |
his 1953 dissertation on, which was a neural network, we'll call it today. It was picked up by
link |
John von Neumann when he was designing the EDVAC computer in 1945. He talked about its
link |
components being neurons based on, I mean, references. He's only got three references
link |
and one of them is the McCulloch Pets paper. So all these people and then the AI people and the
link |
artificial life people, which was John von Neumann originally, is that overlap?
link |
They're all going around the same time. And three of these four disciplines
link |
turn to computation as their primary metaphor. So I've got a couple of chapters in the book.
link |
One is titled, wait, computers are people, because that's where our computers came from.
link |
Yeah. And, you know, from people who are computing stuff. And then another chapter,
link |
wait, people are computers, which is about computational neuroscience. So there's this
link |
whole circle here. And that computation is it. And, you know, I have talked to people about,
link |
well, maybe it's not computation that goes on in the head. Of course it is. Yeah. Okay. Well,
link |
well, when Elon Musk's rocket goes up, is it computing? Is that how it gets into orbit?
link |
By computing? But we've got this idea. If you want to build an AI system, you'll write a computer
link |
program. Yeah. In a sense, so the word computation very quickly starts doing a lot of work that
link |
was not initially intended to do. It's just like going to say, if you talk about the universe
link |
as essentially performing a computation. Yeah, right. Wolfram does this. He turns it into
link |
computation. You don't turn rockets into computation. Yeah. By the way, when you say
link |
computation in our conversation, do you tend to think of computation narrowly in the way
link |
touring thought of computation? It's gotten very, you know, squishy. Yeah. Squishy. But
link |
computation in the way touring thinks about it and the way most people think about it actually
link |
fits very well with thinking like a hunter gatherer. There are places and there can be stuff in
link |
places and the stuff in places can change and it stays there until someone changes it. And it's
link |
this metaphor of place and container, which, you know, is a combination of our place cells in our
link |
hippocampus and cortex. But this is how we use metaphors for mostly to think about. And when
link |
we get outside of our metaphor range, we have to invent tools, which we can sort of switch on to
link |
use. So calculus is an example of a tool. It can do stuff that our raw reasoning can't do and we've
link |
got conventions of when you can use it or not. But sometimes, you know, people try to, all the
link |
time, we always try to get physical metaphors for things, which is why quantum mechanics has
link |
been such a problem for 100 years because it's a particle. No, it's a wave. It's got to be something
link |
we understand. And I say, no, it's some weird mathematical logic that's different from those.
link |
But we want that metaphor. Well, you know, I suspect that, you know, 100 years or 200 years
link |
from now, neither quantum mechanics nor dark matter will be talked about in the same terms,
link |
you know, in the same way that Flodgerton's theory eventually went away because it just wasn't an
link |
adequate explanatory metaphor. You know, that metaphor was the stuff. There is stuff in the
link |
burning. The burning is in the matter. It turns out the burning was outside the matter. It was
link |
the oxygen. So our desire for metaphor and combined with our limited cognitive capabilities gets us
link |
into trouble. That's my argument in this book. Now, and people say, well, what is it then? And I
link |
say, well, I wish I knew that right about that. But I, you know, I give some ideas. But so, so
link |
there's the three things. Computation is sort of a particular thing we use. Oh, can I tell you
link |
one beautiful thing? One beautiful thing. So, you know, I used an example of a thing that's
link |
different from computation. You hit a drum and it vibrates. And there are some, some stationary
link |
points on the drum surface, you know, because the wave is going up and down the stationary points.
link |
Now, you could compute them to arbitrary precision. But the drum just knows them. The
link |
drum doesn't have to compute. What was the very first computer program ever written by Adel
link |
Lovelace to compute Bernoulli numbers? And Bernoulli numbers are exactly what you need to find those
link |
stable points in the drum surface. Wow. Anyway, and there was a bug in the program. The arguments
link |
to divide were reversed in one place. And it still worked. Well, she's never got to run it. They
link |
never built the analytical engine. She wrote the program without, without it, you know.
link |
So, so computation. Computation is sort of, you know, a thing that's become dominant as a metaphor.
link |
But yeah, is it the right metaphor? All three of these four fields adopted computation. And,
link |
you know, the, a lot of it swirls around Warren McCulloch and his, all his students. And he
link |
funded a lot of people. And, and our human metaphors, our limitations to human thinking
link |
will play into this. The three themes of the book. So I have a little to say about computation. So
link |
you're saying that there is a gap between the computer or the, the, the machine that performs
link |
computation and this machine that appears to have consciousness and intelligence. Yeah. Can we
link |
that piece of meat in your head piece of meat, and maybe it's not just the meat in your head.
link |
It's the rest of you too. I mean, you have, you have, you actually have a neural system in your
link |
gut. I tend to also believe, not believe, but we're now dancing around things we don't know. But
link |
I tend to believe other humans are important. Like, so we're almost like, I just don't think
link |
we would ever have achieved the level of intelligence we have with other humans. I'm not
link |
saying so confidently, but I have an intuition that some of the intelligence is in the interaction.
link |
Yeah. And, and I think, you know, I think it seems to me very likely again, we, you know,
link |
this speculation, but we, our species, and probably, probably in the end, those to some
link |
extent, because you can find old bones where they seem to be counting on them by putting notches
link |
that were near them in the anticles had done. We're able to put some of our stuff outside
link |
our body into the world, and then other people can share it. And then we get these tools that
link |
become shared tools. And so there's a whole coupling that would not occur in, you know,
link |
the single deep learning network, which was fed, you know, all of literature or something.
link |
Yeah. The neural network can't step outside of itself. But is there, is there some,
link |
can we explore this dark room a little bit and try to get at something? What is the magic? Where
link |
does the magic come from in the human brain that creates the mind? What's your sense as scientists
link |
that try to understand it and try to build it? What are the directions if followed might be
link |
productive? Is it creative interactive robots? Is it creating large deep neural networks that
link |
do like self supervised learning? And just like wolf will, will discover that when you
link |
make something large enough, some interesting things will emerge. Is it through physics and
link |
chemistry and biology, like artificial life angle, like we'll sneak up in this four quadrant
link |
matrix that you mentioned? Is there anything you're most, if you had to bet all your money,
link |
financial? I wouldn't. Okay. So every intelligence we know, who's, you know,
link |
animal intelligence, dog intelligence, you know, octopus intelligence, which is a very different
link |
sort of architecture from us. All the intelligences we know perceive the world in some way and then
link |
have action in the world, but they're able to perceive objects in a way which is actually pretty
link |
damn phenomenal and surprising. You know, we tend to think, you know, that the box over here
link |
between us, which is a sound box, I think, is a blue box. But blueness is something that we
link |
construct with color constancy. It's not a, it's not a, it's not, the blueness is not a direct
link |
function of the photons we're receiving. It's actually context, you know, which is why you can
link |
turn, you know, maybe seen the examples where someone turns a stop sign into a, some other sort
link |
of sign by just putting a couple of marks on them and the deep learning system gets it wrong.
link |
And everyone says, but the stop sign's red. You know, why is it, why is it thinking it's the
link |
other sort of sign? Because redness is not intrinsic in just the photons. It's actually
link |
a construction of an understanding of the whole world and the relationship between objects
link |
to get color constancy. But our tendency in order that we get an archive paper really quickly is
link |
to just show a lot of data and give the labels and hope it figures it out. But it's not figuring
link |
it out in the same way we do. We have a very complex perceptual understanding of the world.
link |
Dogs have a very different perceptual understanding based on smell. They go smell, smell a post,
link |
they can tell how many, you know, different dogs have visited it in the last 10 hours and how
link |
long ago there's all sorts of stuff that we just don't perceive about the world. And just taking
link |
a single snapshot is not perceiving about the world. It's not perceiving the registration between
link |
us and the object. And registration is a philosophical concept. Brian Cantrell Smith
link |
talks about a lot, very difficult, squirmy thing to understand. But I think none of our systems
link |
do that. We've always talked in AI about the symbol grounding problem, how our symbols that we
link |
talk about are grounded in the world. And when deep learning came along and started labeling
link |
images, people said, ah, the grounding problem has been solved. No, the labeling problem was solved
link |
with some percentage accuracy, which is different from the grounding problem. So you agree with Hans
link |
Marwick and what's called the Marwick's paradox that highlights this counterintuitive notion that
link |
reasoning is easy, but perception and mobility are hard. Yeah, we shared an office when I was
link |
working on computer vision and he was working on his first mobile robot. What were those conversations
link |
like? That were great. Do you still kind of maybe you can elaborate and do you still believe this
link |
kind of notion that perception is really hard? Can you make sense of why we humans have this
link |
poor intuition about what's hard or not? Well, let me give us sort of another story.
link |
Sure. If you go back to, you know, the original, you know, teams working on AI
link |
from the late 50s into the 60s, you know, and you go to the AI lab at MIT,
link |
who was it that was doing that? Was it a bunch of really smart kids who got into MIT
link |
and they were intelligent? So what's intelligence about? Well, the stuff they were good at, playing
link |
chess, doing integrals, that was that was hard stuff. But, you know, a baby could see stuff.
link |
That wasn't that wasn't intelligent. I mean, anyone could do that. It's not intelligence.
link |
And so it, you know, this, there was this intuition that the hard stuff is the things they were good
link |
at. And the easy stuff was the stuff that everyone could do. Yeah. And maybe I'm overplaying it a
link |
little bit. And I think there's an element of that. Yeah. I mean, there, I don't know how much
link |
truth there is to like chess, for example, has was for the longest time seen as the highest
link |
level of intellect, right? Until we got computers that were better at it than people. And then
link |
we realized, you know, if you go back to the 90s, you'll see, you know, the stories and the press
link |
around when, when Kasparov was beaten by Deep Blue. Oh, this is the end of all sorts of things.
link |
Computers are going to be able to do anything from now on. And we saw exactly the same stories
link |
with AlphaZero, the go playing program. Yeah. But still, to me, reasoning is a special thing.
link |
And perhaps, no, we actually, we're really bad at reasoning. We just use these analogies based
link |
on our hunter gatherer intuitions. But why is that not, don't you think the ability to construct
link |
metaphor is a really powerful thing? Oh, yeah, it is stories. It is. It's the construction of the
link |
metaphor and registering that something constant in our brains. Like, isn't that what we're doing
link |
with with vision to and what we're telling our stories, we're constructing good models of the
link |
world? Yeah, yeah. But I think we jumped between what we're capable of and how we're doing it
link |
right there. There was a little confusion that went on. Sure. As we were telling each other
link |
stories. Yes, exactly. Trying to delude each other. No, I just think I'm not exactly, so I'm
link |
trying to pull apart this Morvex paradox. I don't view it as a paradox. What did evolution,
link |
what did evolution spend its time on? Yes, it spent its time on getting us to perceive and
link |
move in the world. That was, you know, 600 million years as multi subtle creatures doing that. And
link |
then it was, you know, relatively recent that we that we, you know, were able to hunt or gather
link |
or, you know, even, even animals hunting. That's much more recent. And then, and then anything
link |
that we, you know, speech, language, those things are, you know, a couple of hundred thousand years
link |
probably, if that long, and then agriculture, 10,000 years, you know, all that stuff was built
link |
on top of those earlier things, which took a long time to develop. So if you then look at the
link |
engineering of these things, so building it into robots, what's the hardest part of robotics,
link |
do you think, as the decades that you worked on robots, in the context of what we're talking
link |
about vision, like a perception, the actual sort of the, the biomechanics of movement,
link |
I'm kind of drawing parallels here between humans and machines always, like, what do you think is
link |
the hardest part of robotics? I sort of think all of them. There are no easy parts to do well.
link |
We sort of go reductionist and we reduce it. If only we had all the, the location of all the
link |
points in 3D, things would be great. You know, if only we had labels on the, on the images,
link |
you know, things would be great. But, you know, as, as we see, that's not good enough. Some deeper
link |
understanding. But if you, if I came to you and I could solve one category of problems in robotics
link |
instantly, what would give you the greatest pleasure?
link |
I mean, is it, you know, you look at robots that manipulate objects.
link |
What's hard about that? You know, is it the perception? Is it the, the reasoning about the
link |
world, like common sense reasoning? Is it the actual building a robot that's able to interact
link |
with the world? Is it like human aspects of a robot that's interacting with humans and that,
link |
that game theory of how they work well together?
link |
Well, let's talk about manipulation for a second, because I had this really blinding moment.
link |
You know, I'm a grandfather, so grandfathers have blinding moments. Just three or four miles from
link |
here. Last year, my 16 month old grandson was in his new house, first time, right? First time in
link |
this house. And he'd never been able to get to a window before, but this had some low windows.
link |
And he goes up to this window with a handle on it that he's never seen before. And he's got one
link |
hand pushing the window and the other hand turning the handle to open the window. He, he, he knew
link |
two different hands, two different things he knew how to, how to put together. And he's 16 months
link |
old. And there you are watching an awe.
link |
In an environment, environment he'd never seen before.
link |
How did he do that? How did he do that? Yes, that's a good question. How did he do that?
link |
That's why it's like, okay, like, you could see the leap of genius from using one hand to perform
link |
a task to combining doing, I mean, first of all, in manipulation, that's really difficult. It's
link |
like two hands, both necessary to complete the action and completely different. And he'd never
link |
seen a window open before. But he inferred somehow a handle open something.
link |
There may have been a lot of slightly different failure cases that you didn't see.
link |
Not with a window, but with other objects of turning and twisting and handles.
link |
There's a great counter to reinforcement learning. We'll just give the robot,
link |
we'll give the robot plenty of time to try everything. Can I tell a little side story here?
link |
So I'm in DeepMind in London, four years ago, where there's a big Google building,
link |
and then you go inside and you go through this more security, and then you get to DeepMind,
link |
where the other Google employees can't go. And I'm in a conference room,
link |
bare conference room with some of the people. And they tell me about their reinforcement
link |
learning experiment with robots, which are just trying stuff out. And they're my robots.
link |
They're Soyuz that we sold them. And they really like them because Soyuz are compliant
link |
and can sense forces, so they don't break when they're bashing into walls. They stop and they
link |
do stuff. So you just let the robot do stuff and eventually it figures stuff out.
link |
By the way, we're talking about robot manipulation, so robot arms and so on.
link |
Yeah, Soyuz is a robot arm. Just go, what's Soyuz?
link |
Soyuz is a robot arm that my company, Rethink Robotics, built.
link |
Thank you for the context.
link |
Yeah, sorry. Okay, cool. So we're in DeepMind.
link |
And it's in the next room, these robots are just bashing around to try and use reinforcement
link |
learning to learn how to act. Can I go see them? Oh, no, they're secret. They're all
link |
my robots that were secret. That's hilarious. Okay.
link |
Anyway, the point is, you know, this idea that you just let reinforcement learning figure
link |
everything out is so counter to how a kid does stuff. So again, story about my grandson, I gave
link |
him this box that had lots of different lock mechanisms. He didn't randomly, you know,
link |
and he was 18 months old. He didn't randomly try to touch every surface or push everything.
link |
He found, he could see where the mechanism was, and he started exploring the mechanism
link |
for each of these different lock mechanisms. And there was reinforcement, no doubt of some
link |
sort going on there. But he applied a prefilter, which cut down the search space dramatically.
link |
I wonder to what level we're able to introspect what's going on,
link |
because what's also possible is you have something like reinforcement learning
link |
going on in the mind in the space of imagination. So like you have a good model of the world you're
link |
predicting, and you may be running those tens of thousands of like loops, but you're like as a human,
link |
you're just looking at yourself trying to tell a story of what happened. And it might seem simple,
link |
but maybe there's a lot of computation going on. Whatever it is, but there's also a mechanism
link |
that's being built up. It's not just random search. That mechanism prunes it dramatically.
link |
Yeah, that pruning, that pruning step, but it doesn't, it's possible that that's,
link |
so you don't think that's akin to a neural network inside a reinforcement learning algorithm?
link |
Yeah, until it's possible. I'll be incredibly surprised if that happens. I'll also be incredibly
link |
surprised that after all the decades that I've been doing this where every few years someone
link |
thinks, now we've got it. Now we've got it. Four or five years ago, I was saying, I don't think
link |
we've got it yet. And everyone was saying, you don't understand how powerful AI is. I had people
link |
tell me, you don't understand how powerful it is. I sort of had a track record of what the world
link |
had done to think, well, this is no different from before. Well, we have bigger computers. We had
link |
bigger computers in the 90s and we could do more stuff. But okay, so let me push back. I'm
link |
generally sort of optimistic and tried to find the beauty in things. I think there's a lot of
link |
surprising and beautiful things that neural networks, this new generation of deep learning
link |
revolution has revealed to me, has continually been very surprising, the kind of things it's
link |
able to do. Now, generalizing that over saying like this, we've solved intelligence, that's another
link |
big leap. But is there something surprising and beautiful to you about neural networks that
link |
where actually you said back and said, I did not expect this?
link |
Oh, I think their performance on ImageNet was shocking. So computer vision in those early
link |
days was just very like, wow, okay. That doesn't mean that they're solving everything in computer
link |
vision. We need to solve in vision for robots. What about AlphaZero and self play mechanisms
link |
and reinforcement learning? Yeah, that was all in Donald Mickey's 1961 paper. Everything there
link |
was there, which introduced reinforcement learning. No, but come on. So you're talking
link |
about the actual techniques. But isn't this surprising to you the level it's able to achieve
link |
with no human supervision of chess play? Like, to me, there's a big, big difference in deep blue.
link |
And maybe what that's saying is how overblown our view of ourselves is.
link |
You know, we had that chess is easy. Yeah, I mean, I came across this 1946 report that,
link |
and I'd seen this as a kid in one of those books that my mother had given me actually,
link |
1946 report, which pitted someone with an abacus against an electronic calculator.
link |
And he beat the electronic calculator. So there at that point was, well, humans are still better
link |
than machines are calculating. Are you surprised today that a machine can do a billion floating
link |
point operations a second and you're puzzling for minutes through one? So I don't know,
link |
but I am certainly surprised. There's something to me different about learning.
link |
So system that's able to learn learning. Now you see, now you're getting into one of the deadly
link |
sins because of using terms overly broadly. Yeah, I mean, there's so many different forms of learning.
link |
Yeah. And so many different forms. You know, I learned my way around the city. I learned to
link |
play chess. I learned Latin. I learned to ride a bicycle. All of those are, you know,
link |
a very different capabilities. Yeah. And if someone has a, you know, in the old days,
link |
people would write a paper about learning something. Now the corporate press office
link |
puts out a press release about how company X has leading the world because they have a system that
link |
can. Yeah, but here's the thing. Okay. So what is learning? What I'm referring to learning is many
link |
things, but I suitcase word. It's a suitcase word. But loosely, there's a dumb system. And over time,
link |
it becomes smart. Well, it becomes less dumb at the thing that it's doing. Yeah. Smart is a
link |
is a loaded word. Yes, less, less dumb at the thing. It gets better performance under some measure.
link |
Yeah. And some set of conditions at that thing. And most of these learning algorithms,
link |
learning systems fail when you change the conditions just a little bit in a way that
link |
humans don't. So, right, I was at DeepMind. The AlphaGo had just come out. And I said,
link |
what would have happened if you'd given it a 21 by 21 board instead of a 19 by 19 board? They said,
link |
fail totally. But a human player would actually, you know, well, would actually be able to play
link |
a game. And actually, funny enough, if you look at DeepMind's work, since then, they are presenting
link |
a lot of algorithms that would do well at the bigger board. So they're slowly expanding this
link |
generalization. I mean, to me, there's a core element there. It is very surprising to me that
link |
even in a constrained game of chess or Go that through self play by system playing itself,
link |
that can, it can achieve super human level performance through learning alone. So like.
link |
Okay. So, so, you know, you didn't still fund them as you did in a search of that.
link |
You didn't, you didn't like it when I referred to Donald Mickey's 1961 paper. There in the second
link |
part of it, which came a year later, they had self play on an electronic computer at tic,
link |
tac, toe. Okay. It's not as, but it learned to play tic, tac, toe through self play. That's
link |
not what learned to play optimally. What I'm saying is I, okay, I have a little bit of a bias,
link |
but I find ideas beautiful, but only when they actually realize the promise that's another level
link |
of beauty. Like, for example, with Bezos and Elon Musk are doing with rockets, rockets for a
link |
long time, but doing reusable, cheap rockets, it's very impressive. In the same way, I, okay,
link |
yeah, I would have not predicted. First of all, when I was started and fell in love with AI,
link |
the game of Go was seen to be impossible to solve. Okay. So I thought maybe, you know,
link |
I, maybe it'd be possible to maybe have big leaps in a Moore's law style of way in computation,
link |
I'll be able to solve it. But I would never have guessed that you could learn your way. However,
link |
I mean, in the narrow sense of learning, learn your way to, to, to beat the best people in
link |
the world at the game of Go without human supervision, not studying the game of experts.
link |
Okay. So, so using a different learning technique, Arthur Samuel in the early 60s,
link |
and he was the first person to use machine learning, got, had a program that could beat
link |
the world champion at checkers. Now, so, and that time was considered amazing. By the way,
link |
Arthur Samuel had some fantastic advantages. Do you want to hear Arthur Samuel's advantages?
link |
Two things. One, he was at the 1956 AI conference. I knew Arthur later in life.
link |
He was at Stanford when I was grad student there. He wore a tie and a jacket every day.
link |
The rest of us didn't. He's a delightful man, delightful man.
link |
It turns out, Claude Shannon, in a 1950 scientific American article, outlined on chess
link |
playing, outlined the learning mechanism that Arthur Samuel used and they had met in 1956.
link |
I assume there was some communication, but I don't know that for sure. But Arthur Samuel
link |
has been a vacuum tube engineer on getting reliability of vacuum tubes and then had
link |
overseen the first transistorized computers at IBM. And in those days, before you shipped a
link |
computer, you ran it for a week to seek to get early failures. So here you had this whole farm
link |
of computers running random code for hours and hours a week for each computer. He had a whole
link |
bunch of them. So he ran his chess learning program with self play on IBM's production line.
link |
He had more computation available to him than anyone else in the world. And then he was able
link |
to produce a chess playing program. I mean, a checkers playing program that could beat the
link |
world champion. So that's amazing. The question is, I mean, I'm surprised, I don't just mean
link |
it's nice to have that accomplishment. Is there is a stepping towards something that feels
link |
more intelligent than before? Yeah, but that's in your view of the world.
link |
Okay, okay. Well, I mean, then it doesn't mean I'm wrong.
link |
No, no, no. So the question is, if we keep taking steps like that, how far that takes us?
link |
Are we going to build a better recommender systems? Are we going to build a better robot?
link |
Or will we solve intelligence? So, you know, I'm putting my bet on
link |
but still missing a whole lot a lot. And why would I say that? Well, in these games, they're all,
link |
you know, 100% information games. But again, but each of these systems is a very short
link |
description of the current state, which is different from registering and perception
link |
in the world, which gets back to Maurevec's paradox.
link |
I'm definitely not saying that chess is somehow harder than perception,
link |
or any kind of robotics in the physical world. I definitely think it's way harder than the game
link |
of chess. So I was always much more impressed by the workings of the human mind. It's incredible.
link |
The human mind is incredible. I believe that from the very beginning. I wanted to be a psychiatrist
link |
for the longest time. I always thought that's way more incredible in the game of chess. I think
link |
the game of chess is, I love the Olympics. It's just another example of us humans picking a task
link |
and then agreeing that a million humans will dedicate their whole life to that task. And that's
link |
the cool thing that the human mind is able to focus on one task and then compete against each
link |
other and achieve like weirdly incredible levels of performance. That's the aspect of chess that's
link |
super cool. Not that chess in itself is really difficult. It's like the Fermat's last theorem
link |
is not in itself to me that interesting. The fact that thousands of people have been struggling
link |
to solve that particular problem is fascinating. So can I tell you my disease in this way? Sure.
link |
Which actually is closer to what you're saying. So as a child, you know, I was building various,
link |
I called them computers. They weren't general purpose computers. Ice cube tray. The ice cube
link |
tray was one. But I built other machines. And what I liked to build was machines that could beat
link |
adults at a game. And they couldn't, the adults couldn't beat my machine. Yeah. So that was,
link |
you were like, that's powerful. Like that's a way to rebel. Yeah. By the way,
link |
when was the first time you built something that outperformed you? Do you remember?
link |
Well, I knew how it worked. I was probably nine years old. And I built a thing that
link |
was a game where you take turns in taking matches from a pile. And either the one who
link |
takes the last one or the one who doesn't take the last one wins, I forget. And so it was pretty
link |
easy to build that out of wires and nails and little coils that were like plugging in the number
link |
and a few light bulbs. The one that I was proud of, I was 12 when I built a thing out of old
link |
telephone switchboard switches that could always win at tic tac toe. And that was a
link |
much harder circuit to design. But again, it was just, it was no active components. It was just
link |
three positions, which is empty x zero, and nine of them and light bulb one, which,
link |
which move it wanted next. And then the human would go and move that.
link |
See, there's magic in that creation. I tend to, I tend to see magic in robots that
link |
like I also think that intelligence is a little bit overrated. I think we can have deep
link |
connections with robots very soon. And we'll come back to connections. Sure. But, but I do want to
link |
say I don't, I think too many people make the mistake of seeing that magic and thinking,
link |
well, we'll just continue, you know, but each, each one of those is a hard fought battle for
link |
the next step, the next step. Yes. The open question here is, and this is why I'm playing
link |
devil's advocate, but I often do when I read your blog post in my mind, because I have
link |
like this eternal optimism is it's not clear to me. So I don't do what obviously the journalists do
link |
or like give into the hype, but it's not obvious to me how many steps away we are from,
link |
from a truly transformational understanding of what it means to build intelligence systems,
link |
like, or how to build intelligence systems. I'm also aware of the whole history of artificial
link |
intelligence, which is where you're deep grounding of this is, is there has been an optimism for
link |
decades. And that optimism, just like reading old optimism is absurd, because people were like,
link |
this is, they were saying things are trivial for decades since the sixties. They're saying
link |
everything is true computer vision is trivial. But I think my mind is working crisply enough to
link |
where I mean, we can dig into if you want. I'm really surprised by the things deep mind has
link |
done. I don't think they're so they're yet close to solving intelligence, but I'm not sure it's
link |
not 10, 10 years away. What I'm referring to is interesting to see when the engineering
link |
it takes that idea to scale and this and the idea works and no, it fools people.
link |
Oh, okay. Honestly, Ronnie, if it was you, me and Demis inside a room, forget the press,
link |
forget all those things. Just as a scientist as a roboticist, you know, that wasn't surprising to
link |
you that at scale. So we're talking about a very large now. Okay, let's pick one that's the most
link |
surprising to you. Okay, please don't yell at me. GPT three. Okay. Well, I was gonna bring that out.
link |
Okay, Alpha zero, Alpha go Alpha go zero, Alpha zero, and then Alpha fold one and two. So aren't
link |
any of do any of these kind of have this core of not forget usefulness or application and so on,
link |
which you could argue for Alpha fold, like as a scientist was doors surprising to you that it
link |
worked as well as it did. Okay, so if we're going to make the distinction between surprise and
link |
usefulness, and I'll have to explain this, I would say Alpha fold. And one of the problems
link |
at the moment with Alpha fold is, you know, it gets a lot of them right, which is a surprise to me
link |
because they're a really complex thing. But you don't know which ones it gets right, which then
link |
is a bit of a problem. Now they've come out with a reason. You mean the structure of the protein
link |
gets a lot of those right? Yeah, it's a surprising number of them right. It's been a really hard
link |
problem. So that was a surprise how many it gets right. So far, the usefulness is limited because
link |
you don't know which ones are right or not. And now they've come out with a thing in the last few
link |
weeks, which is trying to get a useful tool out of it. And they may well do it. In that sense,
link |
the least Alpha fold is different, because your Alpha fold two is different. Because now it's
link |
producing data sets that are actually, you know, potentially revolutionizing competition
link |
biology, like they will actually help a lot of people. But you would say potentially
link |
revolutionizing. We don't know yet. But yeah, that's true. Yeah, but they're, you know, but I
link |
got your, I mean, this is okay. So you know what, this is going to be so fun. So let's go
link |
right into it. Speaking of robots that operate in the real world. Let's talk about
link |
self driving cars. Oh, okay. Because you do you have built robotics companies, you're one of the
link |
greatest roboticists in history. And that's not in space of just in the space of ideas.
link |
We'll also probably talk about that. But in the actual building and execution of businesses that
link |
make robots that are useful for people and that actually work in the real world and make money.
link |
You also sometimes are critical of Mr. Elon Musk, or let's more specifically focus on this
link |
particular technology, which is autopilot inside Tesla's. What are your thoughts about Tesla autopilot
link |
or more generally vision based machine learning approach to semi autonomous driving?
link |
These are robots that are being used in the real world by hundreds of thousands of people.
link |
And if you want to go there, I can go there, but that's not too much, which there is, let's say
link |
they're on par safety wise as humans currently, meaning human alone versus human plus robot.
link |
Okay. So first, let me say I really like the car I came here in here today, which is
link |
a 2021 model Mercedes E 450. I am impressed by the machine vision. So now other things,
link |
I'm impressed by what it can do. I'm really impressed with many aspects of it. And it's able
link |
to stay in lane. It does the lane stuff. It's looking on either side of me. It's telling me
link |
about nearby cars, blind spots and so on. Yeah. When I'm going in close to something in the park,
link |
I get this beautiful, gorgeous, top down view of the world. I am impressed
link |
up the wazoo of how registered and metrical that is. So it's like multiple cameras and
link |
it's all very good together to produce a 360 view kind of 360 view, you know, synthesized
link |
so it's above the car. And it is unbelievable. I got this car in January. It's the longest
link |
I've ever owned a car without digging it. So it's better than me. Me and it together better.
link |
So I'm not saying technologies are bad or not useful. But here's my point. Yes.
link |
It's a replay of the same movie. Okay. So maybe you've seen me ask this question before.
link |
But when did the first car go over 55 miles an hour for over 10 miles
link |
on a public freeway with other traffic around driving completely autonomously? When did that
link |
happen? Was it in the 80s or something? It was a long time ago. It was actually in 1987
link |
in Munich at the Bundeswehr. So they had it running in 1987. When do you think,
link |
and Elon has said he's going to do this, when do you think we'll have the first car
link |
drive coast to coast in the US hands off the wheel, hands off the wheel, feet off the pedals,
link |
coast to coast? As far as I know, a few people have claimed to do it. 1995. That was comedy.
link |
I didn't know. But oh, that was the code. Yeah. They didn't claim, did they claim 100%
link |
they claim? Not 100%. Not 100%. And then there's a few marketing people who have claimed 100%.
link |
But my point is that what I see happening again is someone sees a demo and they over generalize
link |
and say we must be almost there. But we've been working on it for 35 years. So that's demos.
link |
But this is going to take us back to the same conversation with AlphaZero. Are you not?
link |
Okay. I'll just say what I am. Because when I first started interacting with the
link |
with the Mobileye implementation at Tesla Autopilot, I've driven a lot of cars. You know,
link |
I've been in Google stuff driving cars since the beginning. I thought there was no way before I sat
link |
and use Mobileye, I thought they're just knowing computer vision. I thought there's no way it could
link |
work as well as it was working. So my model of the limits of computer vision was way more limited
link |
than the actual implementation of Mobileye. So that's one example. I was really surprised.
link |
I was like, wow, that was incredible. The second surprise came when Tesla threw away Mobileye
link |
and started from scratch. I thought there's no way they can catch up to Mobileye. I thought what
link |
Mobileye was doing was kind of incredible, like the amount of work and the annotation.
link |
Yeah. Well, Mobileye was started by Amnon Shresher and used a lot of traditional, you know,
link |
hard fought computer vision techniques. But they also did a lot of good, sort of,
link |
like non research stuff, like actual, like just good, like what you do to make a successful
link |
product, right? At scale, all that kind of stuff. And so I was very surprised when they
link |
from scratch were able to catch up to that. That's very impressive. And I've talked to a
link |
lot of engineers that was involved. That was impressive. And the recent progress, especially
link |
under, well, with the involvement under Capati, what they were, what they're doing with the data
link |
engine, which is converting into the driving tasks into these multiple tasks, and then doing this
link |
edge case discovery when they're pulling back, like the level of engineering made me rethink
link |
what's possible. I don't, I still, you know, I don't know to that intensity, but I always thought
link |
it was very difficult to solve the time I was driving with all the sensors, with all the computation.
link |
I just thought it was a very difficult problem. But I've been continuously surprised how much
link |
you can engineer. First of all, the data acquisition problem, because I thought, you know,
link |
just because I worked with a lot of car companies, they're, they're so a little,
link |
a little bit old school to where I didn't think they could do this at scale, like AWS style,
link |
data collection. So when Tesla was able to do that, I started to think, okay, so what are the limits
link |
of this? I still believe that a driver like sensing and the interaction with a driver and like
link |
studying the human factors psychology problem is essential. It's, it's always going to be there.
link |
It's always going to be there, even with fully autonomous driving. But I've been surprised,
link |
what is the limit, especially a vision based alone, how far that can take us? So that's my
link |
levels of surprise. Now, okay, can you explain in the same way you said, like Alpha zero,
link |
that's a homework problem that's scaled large in his chest, like who cares, go with, here's
link |
actual people using an actual car and driving, many of them drive more than half their miles
link |
using the system. Right. So, yeah, they're doing well with, with pure vision.
link |
Pure vision, yeah. And, you know, they, and now no radar, which is, I suspect that can't go all the
link |
way. And one reason is without, without new cameras that have a dynamic range closer to the human eye,
link |
because human eye has incredible dynamic range. And we make use of that dynamic range in, in,
link |
it's a, we have an autism magnitude or some crazy number like that. The cameras don't have that,
link |
which is why you see the, the, the bad cases where the sun on a white thing and it blinds it
link |
in a way, it wouldn't blind a person. I think there's a bunch of things to think about before
link |
you say, this is so good, it's just going to work. Okay. And I'll come at it from multiple angles.
link |
And I know you've got a lot of time. Yeah. Okay, let's, let's do this. I have thought about these
link |
things. Yeah. I know. You've been writing a lot of great blog posts about it for a while before
link |
Tesla had autopilot, right? So you've been thinking about autonomous driving for a while from every
link |
angle. So, so a few things, you know, in the US, I think that the death rate from motor vehicle
link |
accidents is about 35,000 a year, which is an outrageous number, not outrageous compared to
link |
COVID deaths, but you know, there is no rationality. And that's part of the thing people have said,
link |
engineers say to me, well, if we cut down the number of deaths by 10% by having autonomous
link |
driving, that's going to be great. Everyone will love it. And my prediction is that if
link |
autonomous vehicles kill more than 10 people a year, they'll be screaming and hollering,
link |
even though 35,000 people a year have been killed by human drivers. It's not rational.
link |
It's a different set of expectations. And that will probably continue.
link |
So there's that aspect of it. The other aspect of it is that
link |
when we introduce new technology, we often change the rules of the game. So when we introduced cars,
link |
first, you know, into our daily lives, we completely rebuilt our cities and we changed
link |
all the laws. J walking was not an offense. That was pushed by the car companies so that
link |
people would stay off the road so there wouldn't be deaths from pedestrians getting hit.
link |
We completely changed the structure of our cities and had these foul smelling things,
link |
you know, everywhere around us. And now you see pushback in cities like Barcelona is really
link |
trying to exclude cars, etc. So I think that to get to self driving, we will
link |
large adoption. It's not going to be just take the current situation, take out the driver and
link |
put the same car doing the same stuff because the end case is too many. Here's an interesting question.
link |
How many fully autonomous train systems do we have in the US?
link |
I mean, do you count them as fully autonomous? I don't know because they're usually as a driver,
link |
but they're kind of autonomous, right? No, let's get rid of the driver.
link |
Okay, I don't know. It's either 15 or 16. Most of them are in airports.
link |
Okay. There's a few that go about five, two that go about five kilometers out of airports.
link |
When is the first fully autonomous train system for mass transit expected to operate fully
link |
autonomously with no driver in the US city? It's expected to operate in 2017 in Honolulu.
link |
Oh, wow. It's delayed, but they will get there. But by the way,
link |
it was originally going to be autonomous here in the Bay Area.
link |
I mean, they're all very close to fully autonomous, right?
link |
Yeah, but getting the closest to things. And I have often gone on a fully autonomous train
link |
in Japan, one that goes out to that fake island in the middle of Tokyo Bay. I forget the name of
link |
that. And what do you see when you look at that? What do you see when you go to a fully autonomous
link |
train in an airport? It's not like regular trains. At every station, there's a double
link |
set of doors. So there's a door of the train and there's a door off the platform.
link |
Yeah. And it's really visible in this Japanese one because it goes out in amongst buildings.
link |
The whole track is built so that people can't climb onto it. Yeah.
link |
So there's an engineering that then makes the system safe and makes them acceptable.
link |
I think we'll see similar sorts of things happen in the US. What surprised me, I thought,
link |
wrongly, that we would have special purpose lanes on 101 in the Bay Area, the leftmost lane,
link |
so that it would be normal for Teslas or other cars to move into that lane and then say,
link |
okay, now it's autonomous and have that dedicated lane. I was expecting movement to that. Five years
link |
ago, I was expecting we'd have a lot more movement towards that. We haven't. And it may be because
link |
Teslas has been overpromising by calling their system fully self driving. I think they may have
link |
been gotten there quicker by collaborating to change the infrastructure. This is one of the
link |
problems with long hold trucking being autonomous. I think it makes sense on freeways at night
link |
for the trucks to go autonomously. But then how do you get on to and off of the freeway?
link |
What sort of infrastructure do you need for that? Do you need to have the human in there to do that?
link |
Can you get rid of the human? So I think there's ways to get there, but it's an infrastructure
link |
argument because the long tail of cases is very long and the acceptance of it will not be at the
link |
same level as human drivers. So I'm with you still and I was with you for a long time, but I am
link |
surprised how well, how many edge cases of machine learning and vision based methods can cover.
link |
This is what I'm trying to get at. I think there's something fundamentally different
link |
with vision based methods and Tesla autopilot and any company that's trying to do the same.
link |
I'm not going to argue with it because we're speculating. My gut feeling tells me it's going to
link |
be things will speed up when there is engineering of the environment because that's what happened
link |
with every other technology. I don't know about you, but I'm a bit cynical that infrastructure
link |
which relies on government to help out in these cases. If you just look at infrastructure in all
link |
domains, government always drags behind on infrastructure. There's so many just...
link |
Well, in this country, in this country, and of course, there's many, many countries that are
link |
actually much worse on infrastructure. Oh, yes, many of them are much worse in the
link |
somewhat high speed rail that other countries have done much better.
link |
I guess my question is at the core of what I was trying to think through here and ask is,
link |
how hard is the driving problem as it currently stands? You mentioned we don't want to just take
link |
the human out and duplicate whatever the human was doing, but if we were to try to do that,
link |
how hard is that problem? Because I used to think it's way harder. I used to think it's,
link |
with vision alone, it would be three decades, four decades.
link |
Okay, so I don't know the answer to this thing I'm about to pose, but I do notice that on Highway 280
link |
here in the Bay Area, which largely has concrete surface rather than blacktop surface,
link |
the white lines that are painted there now have black boundaries around them.
link |
And my lane drift system in my car would not work without those black boundaries.
link |
Interesting. So I don't know whether they started doing it to help the lane drift,
link |
whether it is an instance of infrastructure following the technology, but my car would not
link |
perform as well without that change in the way they paint the lane.
link |
Unfortunately, really good lane keeping is not as valuable. It's orders of magnitude more
link |
valuable to have a fully autonomous system. But for me, lane keeping is really helpful because
link |
I'm always at it. But you wouldn't pay 10 times. The problem is there's not financial,
link |
like it doesn't make sense to revamp the infrastructure to make lane keeping easier.
link |
It does make sense to revamp the infrastructure. If you have a large fleet of autonomous vehicles,
link |
now you change what it means to own cars, you change the nature of transportation,
link |
but for that you need autonomous vehicles. Let me ask you about Waymo then. I've gotten
link |
a bunch of chances to ride in a Waymo self driving car. I don't know if you'd call them
link |
self driving. Well, I mean, I rode in one before they were called Waymo at X.
link |
So there's currently another surprisingly, but I didn't think it would happen,
link |
which is they have no driver currently. Yeah, in Chandler.
link |
In Chandler, Arizona. And I think they're thinking of doing that in Austin as well,
link |
but they're expanding. Although, you know, I do an annual checkup on this.
link |
So as of late last year, they were aiming for hundreds of rides a week, not thousands.
link |
And there is no one in the car, but there's certainly safety people in the loop.
link |
And it's not clear how many, you know, what the ratio of cars to safety people is.
link |
It wasn't, obviously, they're not 100% transparent about this.
link |
No, none of them are 100% transparent. They're very untransparent.
link |
But at least the way they're, I don't want to make definitively, but they're saying
link |
there's no teleoperation. And that sort of fits with YouTube videos I've seen of people being
link |
trapped in the car by a red cone on the street. And they do have rescue vehicles that come,
link |
and then a person gets in and drives it. But isn't it incredible to you,
link |
it was to me to get in a car with no driver and watch the steering wheel turn,
link |
like for somebody who has been studying, at least certainly the human side of autonomous vehicles
link |
for many years, and you've been doing it for way longer. It was incredible to me that this
link |
was actually could happen. I don't care if that scales 100 cars. This is not a demo.
link |
This is not, this is me as a regular. The argument I have is that people make
link |
interpolations from that. Interpolations. That, you know, it's here, it's done.
link |
You know, it's just, you know, we've solved it. No, we haven't yet. And that's my argument.
link |
Okay, so I'd like to go to, you keep a list of predictions on your amazing blog posts.
link |
It'd be fun to go through them. But before that, let me ask you about this. You have,
link |
you have a harshness to you sometimes in your criticism of what is perceived as hype.
link |
And so like, because people extrapolate, like you said, and they kind of buy into the hype,
link |
and then they, they kind of start to think that the technology is way better than it is.
link |
But let me ask you maybe a difficult question. Do you think if you look at history of progress,
link |
don't you think to achieve the quote impossible, you have to believe that it's possible?
link |
Absolutely. Yeah. Look, his, his, his, his, his two great runs. Great. Unbelievable.
link |
1903, first human power, human, you know, heavier than air flight.
link |
Yeah. 1969, we land on the moon. That's 66 years. I'm 66 years old in my lifetime,
link |
that span of my lifetime. Barely get, you know, flying, I don't know what it was, 50 feet or
link |
the length of the first flight or something to landing on the moon. Unbelievable. Fantastic.
link |
But that requires, by the way, one of the Wright brothers, both of them, but one of them didn't
link |
believe it's even possible like a year before, right? So like, not just possible soon, but like
link |
ever. So, so, so, you know, how important is it to believe and be optimistic is what I guess?
link |
Oh, yeah, it is important. It's when it goes crazy. When, when, when I, you know, you said,
link |
what was the word you used for my bad harshness? Harshness. Yes. I just get so frustrated. Yes.
link |
When, when people make these leaps and tell me that I'm, that I don't understand. Right. I, you
link |
know, yeah. There's just from iRobot, which I was co founder of. Yeah. I don't know the exact
link |
numbers now because I haven't, it's 10 years since I stepped off the board. But I believe it's well
link |
over 30 million robots cleaning houses from that one company. And now there's lots of other companies.
link |
Was that a crazy idea that we had to believe in 2002 when we released it? Yeah. That was, we,
link |
we had, we had to, you know, believe that it could be done. Let me ask you about this. So iRobot,
link |
one of the greatest robotics companies ever in terms of many, creating a robot that actually
link |
works in the real world is probably the greatest robotics company ever. You were the co founder of
link |
it. If, if the Rodney Brooks of today talked to the Rodney of back then, what would you tell him?
link |
Because I have a sense that would you pet him on the back and say, well, you're doing is going to
link |
fail, but go at it anyway. That's what I'm referring to with the harshness. You've accomplished an
link |
incredible thing there. One of the several things we'll talk about. Well, like that's what I'm trying
link |
to get at that line. No, it's, it's when my harshness is reserved for people who are not doing it,
link |
who claim it's just, well, this shows that it's just going to happen. But here, here's the thing.
link |
This show, but you have that harshness for Elon too. And no, no, it's a different harshness.
link |
No, it's a different argument with Elon. You know, I, I think SpaceX is an amazing company.
link |
On the other hand, you know, I, in one of my blog posts, I said, what's easy and what's hard? I said,
link |
SpaceX vertical landing rockets, it had been done before grid fins had been done since the 60s.
link |
Every Soyuz has them reusable space DCX reuse those rockets that landed vertically.
link |
There's a whole insurance industry in place for rocket launches. So all sorts of infrastructure
link |
that was doable. It took a great entrepreneur, a great personal expense. He almost drove himself,
link |
you know, bankrupt doing it. A great belief to do it. Whereas Hyperloop has a whole bunch
link |
more stuff that's never been thought about and never been demonstrated. So my estimation is
link |
Hyperloop is a long, long, long further off. And if I've got a criticism of, of, of Elon,
link |
it's that he doesn't make distinctions between when the technology's coming along and ready.
link |
And then he'll go off and, and mouth off about other things, which then people go and compete
link |
about and try and do. And so this is where I, I understand what you're saying. I tend to draw
link |
a different distinction. I, I have a similar kind of harshness towards people who are not telling the
link |
truth, who are basically fabricating stuff to make money or to... Oh, he believes what he says.
link |
I just think he's wrong. To me, that's a very important difference. Yeah, I'm not, I'm not.
link |
Because I think in order to fly, in order to get to the moon, you have to believe,
link |
even when most people tell you you're wrong and most likely you're wrong, but sometimes you're
link |
right. I mean, that's the same thing I have with Tesla autopilot. I think that's an interesting
link |
one. I was, especially when I was, you know, at MIT and just the entire human factors in the
link |
robotics community were very negative towards Elon. It was very interesting for me to observe
link |
colleagues at MIT. I wasn't sure what to make of that. That was very upsetting to me because I
link |
understood where that, where that's coming from. And I agreed with them and I kind of almost felt
link |
the same thing in the beginning until I kind of opened my eyes and realized there's a lot of
link |
interesting ideas here that might be overhyped. You know, if you focus yourself on the idea that
link |
you shouldn't call a system full self driving when it's obviously not
link |
autonomous, fully autonomous, you're going to miss the magic. Oh, yeah, you are going to miss the
link |
magic. But at the same time, there are people who buy it, literally pay money for it and take those
link |
words as given. So that's, but I haven't, so that I take words as given as one thing. I haven't
link |
actually seen people that use autopilot that believe that the behavior is really important,
link |
like the actual action. So like this is like to push back on the very thing that you're frustrated
link |
about, which is like journalists and general people buying all the hype and going on in the
link |
same way. I think there's a lot of hype about the negatives of this too, that people are buying
link |
without using people use the way this is what this was this open my eyes. Actually, the way
link |
people use a product is very different than the way they talk about it. This is true with robotics
link |
with everything. Everybody has dreams of how a particular product might be used or so on this.
link |
And then when it meets reality, there's a lot of fear of robotics, for example,
link |
that robots are somehow dangerous and all those kinds of things. But when you actually have
link |
robots in your life, whether it's in the factory or in the home, making your life better, that's
link |
going to be, that's way different. The your perceptions of it are going to be way different.
link |
And so my just tension was like, here's an innovator.
link |
What is it? Sorry, super cruise from Cadillac was super interesting too. That's a really
link |
interesting system. We should like be excited by those innovations.
link |
Okay, so can I tell you something that's really annoyed me recently? It's really annoyed me
link |
that the press and friends of mine on Facebook are going, these billionaires and their space games,
link |
you know, why are they doing that? Yeah, that's been very frustrating.
link |
It really pisses me off. I must say, I applaud that. I applaud it. It's the taking and not
link |
necessarily the people who are doing the things, but, you know, that I keep having to push back
link |
against unrealistic expectations when these things can become real.
link |
Yeah, I, this was interesting on the, because there's been a particular focus for me is
link |
autonomous driving, Elon's prediction of when he's going to be able to do that.
link |
When certain milestones will be hit.
link |
There's several things to be said there that I always, I thought about because
link |
whenever you said them, it was obvious that's not going to me as a person that kind of
link |
not inside the system is obviously unlikely to hit those. There's two comments I want to make.
link |
One, he legitimately believes it. And two, much more importantly, I think that having ambitious
link |
deadlines drives people to do the best work of their life, even when the odds of those deadlines
link |
are very low. To a point, and I'm not, I'm not talking about anyone here. I'm just saying.
link |
So there's a line there, right? You have to have a line because you overextend and it's demoralizing.
link |
But I will say that there's an additional thing here that those words also drive the stock market.
link |
And, you know, we have because of the way that rich people in the past have manipulated the
link |
rubes through investment, we have developed laws about what you're allowed to say.
link |
There's an area here which is... I tend to be, maybe I'm naive, but I tend to believe
link |
that engineers, innovators, people like that, they're not, they don't think like that,
link |
like manipulating the stock price. But it's possible that I'm, I'm certain it's possible
link |
that I'm wrong. It's a very cynical view of the world because I think most people that run companies
link |
and build, like, especially original founders, they... Yeah, I'm not saying that's the intent.
link |
I'm saying it's a... Eventually it's kind of, you fall into that kind of behavior pattern. I don't
link |
know. I tend to... I wasn't saying it's falling into that intent. It's just you also have to protect
link |
investors in this market. Yeah. Okay, so you have, first of all, you have an amazing blog
link |
that people should check out, but you also have this in that blog, a set of predictions.
link |
Such a cool idea. I don't know how long ago you started, like three, four years ago. It was
link |
January 1st, 2018. Yeah. And I made these predictions and I said that every January
link |
1st, I was going to check back on how my predictions... That's such a great thought.
link |
For 32 years. Oh, you said 32 years. I said 32 years because I thought that'll be January 1st,
link |
2050. I'll be... I will just turn 95. Nice. And so people know that your predictions,
link |
at least for now, are in the space of artificial intelligence. Yeah. I didn't say I was going
link |
to make new predictions. I was just going to measure this set of predictions that I made
link |
because I was sort of annoyed that everyone could make predictions. They didn't come true
link |
and everyone forgot. So I should hope myself to a higher standard. Yeah. But also just putting
link |
years and date rangers on things, it's a good thought exercise and reasoning your thoughts out.
link |
And so the topics are artificial intelligence, autonomous vehicles, and space. I was wondering
link |
if we could just go through some that stand out. Maybe from memory, I can just mention to you some,
link |
let's talk about self driving cars, some predictions that you're particularly proud of
link |
or are particularly interesting from flying cars to the other element here is like how
link |
widespread the location where the deployment of the autonomous vehicles is. And there's also
link |
just a few fun ones. Is there something that jumps to mind that you remember from the predictions?
link |
Well, I think I did put in there that there would be a dedicated self driving lane on 101
link |
by some year. And I think I was over optimistic on that one. Yeah, I actually do remember that.
link |
But I think you were mentioning like difficulties at different cities. Yeah. So Cambridge,
link |
Massachusetts, I think was an example. Yeah, like in Cambridgeport. I lived in
link |
Cambridgeport for a number of years and the roads are narrow and getting anywhere as a human
link |
driver is incredibly frustrating when you start to put. And people drive the wrong way on one way
link |
streets there. It's just. So your prediction was driverless taxi services operating on all
link |
streets in Cambridgeport, Massachusetts in 2035. Yeah. And that may have been too optimistic.
link |
You think so. You know, I've gotten a little more pessimistic since I made these internally
link |
on some of these things. So what can you put a year to a major milestone of deployment of a
link |
taxi service in a few major cities? Like something where you feel like autonomous vehicles are here.
link |
So let's take the grid streets of San Francisco north of market. Relatively benign environment.
link |
The streets are wide. The major problem is delivery trucks stopping everywhere,
link |
which has made things more complicated. A taxi system there with somewhat designated pickup
link |
and drop offs, unlike with Uber and Lyft, where you can sort of get to any place and the drivers
link |
will figure out how to get in there. We're still a few years away. I live in that area. So I see
link |
the, you know, the self driving car companies, cars, multiple, multiple ones every day out
link |
there by the cruise. Zooks less often, Waymo all the time, different ones come and go.
link |
And there's always a driver. There's always a driver at the moment, although I have noticed
link |
that sometimes the driver does not have the authority to take over without talking to the
link |
home office because they will sit there waiting for a long time. And clearly something's going on
link |
where the home office is making a decision. So there, you know, and so you can see whether
link |
they've got their hands on the wheel or not. And it's the incident resolution time that
link |
tells you gives you some clues. So what year do you think? What's your intuition? What date range
link |
are you currently thinking San Francisco would be autonomous taxi service from any point A to
link |
any point B without a driver? Are you still thinking 10 years from now, 20 years from now,
link |
30 years from now? Certainly not 10 years from now. It's going to be longer. If you're allowed
link |
to go south of the market way longer, unless there's reengineering of roads.
link |
By the way, what's the biggest challenge? You mentioned a few. Is it the delivery trucks?
link |
Is it the edge cases, the computer perception? Well, it is a case that I saw outside my house
link |
a few weeks ago, about 8 p.m. on a Friday night. It was getting dark before the solstice.
link |
It was a cruise vehicle come down the hill, turned right, and stopped dead covering the
link |
crosswalk. Why did it stop dead? Because there was a human just two feet from it.
link |
Now, I just glanced. I knew what was happening. The human was a woman was at the door of her car
link |
trying to unlock it with one of those things that, you know, when you don't have a key.
link |
Yes. The car thought, oh, she could jump out in front of me any second.
link |
As a human, I could tell, no, she's not going to jump out. She's busy trying to unlock her.
link |
She's lost her keys. She's trying to get in the car. And it stayed there until I got bored.
link |
And so the human driver in there did not take over. But here's the kicker to me.
link |
A guy comes down the hill with a stroller. I assume there's a baby in there. And now the
link |
crosswalk is blocked by this cruise vehicle. What's he going to do? Cleverly, I think he
link |
decided not to go in front of the car. But he had to go behind it. He had to get off the crosswalk
link |
out into the intersection to push his baby around this car, which was stopped there.
link |
And no human driver would have stopped there for that length of time. They would have gotten out
link |
of the way. And that's another one of my pet peeves, that safety is being compromised
link |
for individuals who didn't sign up for having this happen in their neighborhood.
link |
Yeah. But now you can say that's an edge case, but...
link |
Yeah. Well, I'm in general not a fan of anecdotal evidence for stuff like this is
link |
one of my biggest problems with the discussion of autonomous vehicles and in general,
link |
people that criticize them or support them are using anecdotal evidence.
link |
So let me... But I got you.
link |
You know, your question is when is it going to happen in San Francisco? I say not soon,
link |
but it's going to be one of the... But where it is going to happen is in limited domains,
link |
campuses of various sorts, gated communities, where the other drivers are not arbitrary people.
link |
They're people who know about these things, they've been warned about them, and at velocities
link |
where it's always safe to stop dead, you can't do that on the freeway.
link |
That, I think, we're going to start to see. And they may not be shaped like
link |
current cars, they may be things like main mobility has those things and various companies have
link |
these. Yeah, I wonder if that's a compelling experience. To me, it's not just about automations,
link |
it's about creating a product that makes your... It's not just cheaper, but makes your... It's fun
link |
to ride. One of the most... One of the least fun things is for a car that stops and waits.
link |
There's something deeply frustrating for us humans, for the rest of the world to take
link |
advantage of us as we wait. But think about not you as the customer, but someone who's in their
link |
80s in a retirement village whose kids have said, you're not driving anymore.
link |
And this gives you the freedom to go to the market. That's a hugely beneficial thing, but
link |
it's a very few orders of magnitude less impact on the world. It's not just a few people in a
link |
small community using cars as opposed to the entirety of the world. I like that the first
link |
time that a car equipped with some version of a solution to the trolley problem is what's NIML
link |
stand for? Not in my life. I define my lifetime as 2050. I ask you, when have you had to decide
link |
which person shall I kill? No, you put the brakes on and you brake as hard as you can.
link |
You're not making that decision. I do think autonomous vehicles or semi autonomous vehicles
link |
do need to solve the whole pedestrian problem that has elements of the trolley problem within
link |
it. And I talk about it in one of the articles or blog posts that I wrote.
link |
One of my coworkers has told me he does this. He tortures autonomously driven vehicles and
link |
pedestrians. We'll torture them. Now, once they realize that putting one foot off the curb
link |
makes the car think that they might walk into the road, kids, teenagers will be doing that
link |
all the time. By the way, this is a whole other discussion because my main issue with robotics
link |
is HRI, human and robot interaction. I believe that robots that interact with humans will have to
link |
push back. They can't just be bullied because that creates a very uncompelling experience for
link |
the humans. Waymo, before it was called Waymo, discovered that they had to do that at four way
link |
intersections. They had to nudge forward to give the cue that they were going to go because otherwise
link |
the other drivers would just eat them all the time. You cofounded iRobot, as we mentioned,
link |
one of the most successful robotics companies ever. What are you most proud of with that company
link |
and the approach you took to robotics? Well, there's something I'm quite proud of there,
link |
which may be a surprise, but I was still on the board when this happened. It was March 2011,
link |
and we sent robots to Japan and they were used to help shut down the Fukushima Daiichi nuclear
link |
power plant, which was everything was up in there since I was there in 2014. Some of the
link |
robots were still there. I was proud that we were able to do that. Why were we able to do that?
link |
People have said, well, Japan is so good at robotics. It was because we had had
link |
about 6,500 robots deployed in Iraq and Afghanistan, teleopped, but with intelligence,
link |
dealing with roadside bombs. We had, I think it was at that time, nine years of in field experience
link |
with the robots in harsh conditions, whereas the Japanese robots, which goes back to what
link |
annoys me so much, getting all the hype, look at that. Look at that Honda robot. It can walk. Well,
link |
the future's here. Couldn't do a thing because they weren't deployed, but we had deployed in
link |
really harsh conditions for a long time, and so we're able to do something very positive
link |
in a very bad situation. What about just the simple, and for people who don't know, one of the
link |
things that iRobot has created is the Roomba vacuum cleaner. What about the simple robot
link |
that is the Roomba, quote unquote, simple, that's deployed in tens of millions of homes?
link |
What do you think about that? Well, I make the joke that I started out life as a pure
link |
mathematician and turned into a vacuum cleaner salesman, so if you're going to be an entrepreneur,
link |
be ready to do anything. But I was, you know, there was a wacky lawsuit that I got
link |
opposed for not too many years ago, and I was the only one who had emailed from the 1990s,
link |
and no one in the company had it, so I went and went through my email, and it reminded me of,
link |
you know, the joy of what we were doing, and what was I doing? What was I doing at the time we were
link |
building the Roomba. One of the things was we had this incredibly tight budget, because we
link |
wanted to put it on the shelves at $200. There was another home cleaning robot at the time. It was
link |
the Electrolux Trilobite, which sold for 2,000 euros, and to us that was not going to be a
link |
consumer product. So we had reason to believe that $200 was a thing that people would buy at,
link |
that was our aim. But that meant we had, you know, that's on the shelf making profit.
link |
That means the cost of goods has to be minimal. So I found all these emails of me going, you know,
link |
I'd be in Taipei for a MIT meeting, and I'd stay a few extra days. I'd go down to Shinshu and talk to
link |
these little tiny companies, lots of little tiny companies outside of TSMC, Taiwan Semiconductor,
link |
Taiwan Semiconductor Manufacturing Corporation, which let all these little companies be fabulous.
link |
They didn't have to have their own fab, so they could innovate. And they were building, their
link |
innovations were to build stripped down 6802s. 6802 was what was in an Apple One. Get rid of
link |
half the silicon and still have it be viable. And I'd previously got some of those for some
link |
earlier failed products of iRobot. And then that was in Hong Kong, going to all these companies
link |
that built, you know, they weren't gaming in the current sense. There were these handheld games
link |
that you would play, or birthday cards, because we had about a 50 cent budget for computation.
link |
And so I'm tracking from place to place, looking at their chips, looking at what they'd removed.
link |
Oh, they're interrupt, they're interrupt handling is too weak for a general purpose. So I was going
link |
deep technical detail. And then I found this one from a company called Winbond, which had,
link |
and I'd forgotten that had this much RAM, it had 512 bytes of RAM. And it was in our budget,
link |
and it had all the capabilities we needed. Yeah. So you're excited. Yeah. And I was reading all
link |
these emails, Colin, I found this. So did you think, did you ever think that you guys could be
link |
so successful? Like, eventually, this company would be so successful. Did you could you possibly
link |
have imagined? And no, we never did think that we had 14 failed business models up to 2002.
link |
And then we had two winners same year. No, and then, you know, we I remember the board,
link |
because by this time we had some venture capital in the board went along with us building
link |
some robots for, you know, aiming at the Christmas 2002 market. And we went three times over what
link |
they authorized and built 70,000 of them and sold them all in that first because we released on
link |
September 18. And I was all sold by Christmas. So it was so we were gutsy. But, but yeah,
link |
you didn't think this will take over the world. Well, this is a
link |
so a lot of amazing robotics companies have gone under over the past few decades.
link |
Why do you think it's so damn hard to run a successful robotics company?
link |
Well, there's a few things. One is expectations of capabilities by the
link |
founders that are off base. The founders, not the consumer, the founders.
link |
Yeah, expectations of what what can be delivered. Sure. Miss pricing. And what a customer thinks
link |
is a valid price is not rational necessarily. Yeah. And expectations of customers. And
link |
just the sheer hardness of getting people to adopt a new technology. And I've suffered from all
link |
three of these. You know, I've had more failures and successes in terms of companies. I've suffered
link |
from all three. So do you think one day there will be a robotics company? And by robotics company,
link |
I mean, where your primary source of income is from robots, that will be a trillion plus dollar
link |
company. And so what would that company do? I can't, you know, because I'm still starting robot
link |
companies. Yeah. I'm not making any such predictions in my own mind. I'm not thinking
link |
about a trillion dollar company. And by the way, I don't think, you know, in the 90s, anyone was
link |
thinking that Apple would ever be a trillion dollar company. So these are these are very hard to
link |
to predict. But sorry to interrupt. But don't you because I kind of have a vision in a small way
link |
and it's a big vision in a small way that I see that there would be robots in the home
link |
at scale like Roomba, but more. And that's trillion dollar. Right. And I think there's a
link |
real market pull for them because of the demographic inversion, you know, who's who's
link |
going to do the stuff for the older people? There's too many, you know, I'm leading here.
link |
There's gonna be too many of us. And but we don't have capable enough robots to
link |
to make that economic argument at this point. Do I expect that that will happen? Yes, I expect
link |
it will happen. But I got to tell you, we introduced the Roomba in 2002, and I stayed another
link |
nine years. We were always trying to find what the next home robot would be. And
link |
still today, the primary product of 20 years late, almost 20 years later, 19 years later,
link |
the primary product is still the Roomba. So I robot hasn't found the next one. Do you think it's
link |
possible for one person in the garage to build it versus like Google launching Google self driving
link |
car that turns into Waymo? Do you think this is almost like what it takes to build a successful
link |
robotics company? Do you think it's possible to go from the ground up? Or is it just too much
link |
capital investment? Yeah, so it's very hard to get there without a lot of capital. And we started
link |
to see, you know, fair chunks of capital for some robotics companies, you know, Series Bs.
link |
Because I saw one yesterday for $80 million, I think it was for covariant.
link |
But it can take real money to get into these things, and you may fail along the way. I
link |
certainly failed at Rethink Robotics. And we lost $150 million in capital there.
link |
So okay, so Rethink Robotics is another amazing robotics company you co founded.
link |
So what was the vision there? What was the dream? And what are you most proud of with
link |
Rethink Robotics? I'm most proud of the fact that we got robots out of the cage in factories
link |
that were safe, absolutely safe for people and robots to be next to each other.
link |
So these are robotic arms? Robotic arms for me to pick up stuff and interact with humans.
link |
Yeah, and that the humans could retask them without writing code. And now that's sort of
link |
become an expectation for a lot of other little companies and big companies are advertising
link |
they're doing. That's both an interface problem and also a safety problem. Yeah. So I'm most proud
link |
of that. I completely, I let myself be talked out of what I wanted to do. And you know, you've
link |
always got, you know, I can't replay the tape. You know, I can't replay it. Maybe,
link |
maybe, you know, if I've been stronger on, and I remember the day, I remember the exact meeting.
link |
Can you take me through that meeting? Yeah. So I'd said that I'd set as a target for the company
link |
that we were going to build $3,000 robots with force feedback that were safe for people to be
link |
around. Wow. That was my goal. And we built, so we started in 2008. And we had prototypes built
link |
of plastic, plastic gearboxes. And at a $3,000, you know, lifetime or $3,000, I was saying,
link |
we're going to go after not the people who already have robot arms in factories, the people who would
link |
never have a robot arm, we're going to go after a different market. So we don't have to meet their
link |
expectations. And so we're going to build it out of plastic. It doesn't have to have a 35,000 hour
link |
lifetime. It's going to be so cheap that it's OPEX, not CAPEX. And so we had, we had a prototype
link |
that worked reasonably well. But the control engineers were complaining about these plastic
link |
gearboxes with a beautiful little planetary gearbox. But we could use something called
link |
serious elastic actuators, we embedded them in there, we can measure forces, we knew when we
link |
hit something, etc. The control engineers were saying, yeah, but this is torque ripple, because
link |
these plastic gears, they're not great gears. And there's this ripple and trying to do force
link |
control around this ripple is so hard. And I'm not going to name names, but I remember
link |
one of the mechanical engineers saying, we'll just build a metal gearbox with spur gears.
link |
And it'll take six weeks, we'll be done, problem solved. Two years later, we got the
link |
gear, the spur gearbox working. We cost reduced at every possible way we could.
link |
But now the price went up to, and then the CEO at the time said, well,
link |
we have to have two arms, not one arm. So our first robot product Baxter now cost $25,000.
link |
And the only people who were going to look at that were people who had arms in factories,
link |
because that was somewhat cheaper for two arms than arms and factories. But they were used to
link |
0.1 millimeter reproducibility of motion and certain velocities. And I kept thinking, but
link |
that's not what we're giving you. You don't need position repeatability, use force control like
link |
a human does. No, but we want that repeatability. We want that repeatability. All the other robots
link |
have that repeatability. Why don't you have that repeatability? So can you clarify force controls
link |
you can grab the arm and you can move it? Yeah, you can move it around. But suppose you,
link |
can you see that? Yes. Suppose you want to... Yes.
link |
Suppose this thing is a precise thing that's got a fit here in this right angle.
link |
Yeah. Under position control, you have fixed it where this is. You know where this is precisely
link |
and you just move it and it goes there. If force control, you would do something like
link |
slide it over here till we feel that and slide it in there. And that's how a human gets precision.
link |
They use force feedback and get the things to mate rather than just go straight to it.
link |
Yeah. Couldn't convince our customers who are in factories and were used to thinking about
link |
things a certain way. And they wanted it. So then we said, okay, we're going to build
link |
an arm that gives you that. So now we ended up building a $35,000 robot with one arm with...
link |
Oh, what are they called?
link |
A certain sort of gearbox made by a company whose name I can't remember right now,
link |
but it's the name of the gearbox. But it's got torque ripple in it. So now there was an extra
link |
two years of solving the problem of doing the force with the torque ripple. So we had to do the
link |
thing we had avoided. And for the plastic gearbox, as we ended up having to do,
link |
the robot was now overpriced. And that was your intuition from the very beginning,
link |
kind of that this is not... You're opening a door to solve a lot of problems that you're
link |
eventually going to have to solve this problem anyway. Yeah. And also, I was aiming at a low
link |
price to go into a different market that didn't have... $3,000 would be amazing.
link |
Yeah. I think we could have done it for five. But you talked about setting the goal a little too
link |
far for the engineers. Yeah, exactly. So why would you say that company not failed but went under?
link |
We had buyers and there's this thing called the Committee on Foreign Investment in the US,
link |
SIFIUS. And that had previously been invoked twice around where the government could stop
link |
foreign money coming into a US company based on defense requirements. We went through
link |
due diligence multiple times. We were going to get acquired. But every consortium had Chinese money
link |
in it. And all the bankers would say at the last minute, you know, this isn't going to get past SIFIUS
link |
and the investors would go away. And then we had two buyers. We were about to run out of money.
link |
Two buyers. And one used heavy handed legal stuff with the other one.
link |
Said they were going to take it and pay more. Dropped out when we were out of cash and then
link |
bought the assets at one 30th of the price they had offered a week before. That was a tough week.
link |
Do you, does it hurt to think about like an amazing company that didn't, you know,
link |
like iRobot didn't find a way? It was tough. I said I was never going to start another company.
link |
I was pleased that everyone liked what we did so much that the team was hired by
link |
three companies within a week. Everyone had a job in one of these three companies. Some stayed in
link |
their same desks because another company came in and rented the space. So I felt good about people
link |
not being out on the street. So Baxter has a screen with a face. What, that's a revolutionary idea for
link |
a robot manipulation, a robotic arm. How much opposition did you get? Well, first the screen
link |
was also used during codeless programming where you taught by demonstration that showed you what
link |
its understanding of the task was. So it had two roles. Some customers hated it and so we made it
link |
so that when the robot was running it could be showing graphs of what was happening and not show
link |
the eyes. Other people and some of them surprised me who they were saying, well this one doesn't
link |
look as human as the old one. We like the human looking. So there was a mixed bag.
link |
But do you think that's, I don't know, I'm kind of disappointed whenever I talk to
link |
roboticists, like the best robotics people in the world, they seem to not want to do the eyes type
link |
of thing. They seem to see it as a machine as opposed to a machine that can also have a human
link |
connection. I'm not sure what to do with that. It seems like a lost opportunity. I think the
link |
trillion dollar company will have to do the human connection very well no matter what it does.
link |
Yeah, I agree. Can I ask you a ridiculous question? Sure. Can I give a ridiculous answer?
link |
Do you think, well maybe by way of asking the question, let me first mention that
link |
you're kind of critical of the idea of the Turing test as a test of intelligence.
link |
Let me first ask this question. Do you think we'll be able to build an AI system that humans
link |
fall in love with and it falls in love with the human, like romantic love?
link |
Well, we've had that with humans falling in love with cars even back in the 50s.
link |
It's a different love, right? I think there's a lifelong partnership where you can communicate
link |
and grow. I think we're a long way from that. I think Blade Runner had the time scale totally
link |
wrong. To me, honestly, the most difficult part is the thing that you said with the Marvel X
link |
Paradox is to create a human form that interacts and perceives the world. But if we just look at a
link |
voice, like the movie Her or just like an Alexa type voice, I tend to think we're not that far away.
link |
Well, for some people, maybe not. But as humans, as we think about the future, we always try,
link |
and this is the premise of most science fiction movies, you've got the world just as
link |
is today, and you change one thing. But that's the same with the self driving car. You change
link |
one thing. Everything changes. Everything grows together. Surprisingly, I might be surprising
link |
to you or might not, I think the best movie about this stuff was by Centennial Man.
link |
And what was happening there? It was schmaltzy. But what was happening there?
link |
As the robot was trying to become more human, the humans were adopting the technology of the robot
link |
and changing their bodies. So there was a convergence happening in a sense. So we will not
link |
be the same. We're already talking about genetically modifying our babies. There's
link |
more and more stuff happening around that. We will want to modify ourselves even more for all
link |
sorts of things. We put all sorts of technology in our bodies to improve it. I've got things in
link |
my ears so that I can sort of hear you. So we're always modifying our bodies. So I think it's
link |
hard to imagine exactly what it will be like in the future. But on the touring test side,
link |
do you think, so forget about love for a second. Let's talk about just the
link |
elect surprise. Actually, I was invited to be an interviewer for the elect surprise or whatever
link |
that's in two days. Their idea is success looks like a person wanting to talk to an AI system
link |
for a prolonged period of time, like 20 minutes. How far away are we and why is it difficult to
link |
build an AI system with which you'd want to have a beer and talk for an hour or two hours?
link |
Not for to check the weather or to check music, but just to talk as friends.
link |
Yeah. Well, we saw Weisenbaum back in the 60s with his programmer, Liza,
link |
being shocked at how much people would talk to Eliza. And I remember,
link |
in the 70s, typing stuff to Eliza to see what it would come back with.
link |
I think right now, and this is a thing that
link |
Amazon's been trying to improve with elect. So there is no continuity of
link |
topic. You can't refer to what we talked about yesterday. It's not the same as talking to a
link |
person where there seems to be an ongoing existence, right? Changes. We share moments
link |
together and they last in our memory together. Yeah. There's none of that. And there's no
link |
sort of intention of these systems that they have any goal in life, even if it's to be happy.
link |
You know, they don't even have a semblance of that. Now, I'm not saying this can't be done.
link |
I'm just saying, I think this is why we don't feel that way about them.
link |
I'm sort of a minimal requirement. If you want the sort of interaction you're talking about,
link |
it's a minimal requirement. Whether it's going to be sufficient, I don't know.
link |
We haven't seen it yet. We don't know what it feels like.
link |
I tend to think it's not as difficult as solving intelligence, for example,
link |
and I think it's achievable in the near term. But on the Turing test,
link |
why don't you think the Turing test is a good test of intelligence?
link |
Oh, because again, the Turing, if you read the paper, Turing wasn't saying this is
link |
a good test. He was using this as a rhetorical device to argue that if you can't tell the
link |
difference between a computer and a person, you must say that the computer's thinking because
link |
you can't tell the difference when it's thinking. You can't say something different.
link |
What it has become as this sort of weird game of fooling people. So
link |
back at the AI lab in the late 80s, we had this thing that still goes on called the AI Olympics.
link |
And one of the events we had one year was the original imitation game as Turing talked about
link |
because he starts by saying, can you tell whether it's a man or a woman? So we did that at the lab.
link |
We had, you know, you'd go and type and the thing would come back and you had to tell whether it
link |
was a man or a woman. And one man came up with a question that he could ask, which was always a
link |
dead giveaway of whether the other person was really a man or a woman. He would ask them,
link |
did you have green plastic toy soldiers as a kid? Yeah. What do you do with them? And a woman
link |
trying to be a man would say, oh, I lined them up. We had wars. We had battles. And the man just
link |
being a man. I stomped on them. I burned them. So, you know, that's what the Turing test with
link |
computers has become. What's the trick question? That's why it's sort of devolved into this.
link |
Nevertheless, conversation not formulated as a test is a pretty,
link |
it's a fascinatingly challenging dance. That's a really hard problem. To me, conversation when
link |
non poses a test is a more intuitive illustration how far away we are from solving intelligence
link |
than my computer vision. It's hard. Computer vision is harder for me to pull apart. But with
link |
language, with conversation, you could see... Because language is so human. We don't...
link |
It's so human. We can so clearly see it. Shit, you mentioned something I was going to go off on.
link |
Okay. I mean, I have to ask you, because you were the head of CSAIL, AI lab for a long time.
link |
You're, I don't know, to me, when I came to MIT, you're like one of the greats at MIT. So, what
link |
was that time like? And plus, you were friends with, but you knew Minsky and all the folks there,
link |
all the legendary AI people of which you were one. So, what was that time like? What are memories
link |
that send out to you from that time? From your time at MIT, from the AI lab, from the dreams
link |
that the AI lab represented to the actual revolutionary work?
link |
Let me tell you first a disappointment in myself. As I've been researching this book,
link |
and so many of the players were active in the 50s and 60s, I knew many of them when they were older.
link |
And I didn't ask them all the questions. Now, I wish I had asked. I'd sit with them at our
link |
Thursday lunches, which we had a faculty lunch. And I didn't ask them so many questions that now
link |
I wish I had. Can I ask you that question? Because you wrote that. You wrote that you were
link |
fortunate to know and rub shoulders with many of the greats, those who founded AI, robotics,
link |
and computer science, and the World Wide Web. And you wrote that your big regret nowadays is that
link |
often I have questions for those who have passed on. And I didn't think to ask them any of these
link |
questions, even as I saw them and said hello to them on a daily basis. So, maybe also another
link |
question I want to ask. If you could talk to them today, what question would you ask? What
link |
questions would you ask? I would ask him, you know, he had the vision for humans and computers
link |
working together. And he really founded that at DARPA. And he gave the money to MIT, which
link |
started Project MAC in 1963. And I would have talked to him about what the successes were,
link |
what the failures were, what he saw as progress, etc. I would have asked him more questions about
link |
that. Because now I could use it in my book. But I think it's lost. It's lost forever. A lot of
link |
the motivations are lost. I should have asked Marvin why he and Seymour Papert came down so hard
link |
on neural networks in 1968 in their book Perceptrons. Because Marvin's PhD thesis was on
link |
neural networks. How do you make sense of that? That book destroyed the field.
link |
Do you think he knew the effect that book would have?
link |
All the theorems are negative theorems. That's the way of life. But still,
link |
it's kind of tragic that he was both the proponent and the destroyer of neural networks.
link |
Yeah. Is there other memory standouts from the robotics and the AI work at MIT?
link |
Yeah, but you'll be more specific. Well, I mean, it's such a magical place. To me,
link |
it's a little bit also heartbreaking that with Google and Facebook, like DeepMind and so on,
link |
so much of the talent doesn't stay necessarily for prolonged periods of time in these universities.
link |
Oh, yeah. I mean, some of the companies are more guilty than others of paying
link |
fabulous salaries to some of the highest producers. And then just you never hear from them again.
link |
They're not allowed to give public talks. It's sort of locked away. And it's sort of like collecting
link |
Hollywood stars or something. And they're not allowed to make movies anymore. I heard them.
link |
Yeah. That's tragic. I mean, there's an openness to the university setting where you do research
link |
to both in the space of ideas and space like publication, all those kinds of things.
link |
Yeah. And there's the publication and all that and often, although these places,
link |
say they publish this pressure. But I think, for instance, net net, I think
link |
Google buying those eight or nine robotics company was bad for the field because it locked
link |
those people away. They didn't have to make the company succeed anymore. Locked them away for years
link |
and then sort of all fiddled away. Yeah. So do you have hope for MIT?
link |
For MIT? Yeah, why shouldn't I? Well, I could be harsh and say that
link |
I'm not sure I would say MIT is leading the world in AI or even Stanford or Berkeley.
link |
I would say DeepMind, Google AI, Facebook AI. I would take a slightly different approach,
link |
a different answer. I'll come back to Facebook in a minute. But I think those other places are
link |
following a dream of one of the founders. And I'm not sure that it's well founded
link |
the dream. And I'm not sure that it's going to have the impact that he believes it is.
link |
You're talking about Facebook and Google and so on.
link |
I'm talking about Google.
link |
Google. But the thing is, those research labs aren't, there's the big dream. And I'm usually a
link |
fan of, no matter what the dream is, a big dream is a unifier. Because what happens is you have a
link |
lot of bright minds working together on a dream. What results is a lot of adjacent ideas. I mean,
link |
there's so much progress is made. Yeah. So I'm not saying they're actually leading. I'm not
link |
saying that the universities are leading. But I don't think those companies are leading in general
link |
because they're, you know, we saw this incredible spike in attendees at NeurIPS. And as I said,
link |
in my January 1st review this year for 2020, 2020 will not be remembered as a watershed year
link |
for machine learning or AI. You know, there was nothing surprising happen. But anyway,
link |
unlike when deep learning hit ImageNet, that was a shake. And there's a lot more people writing
link |
papers, but the papers are fundamentally boring and uninteresting. And incremental work. Yeah.
link |
Is there a particular memories you have with Minsky or somebody else at MIT that stand out?
link |
Funny stories. I mean, unfortunately, he's another one that's passed away.
link |
You've known some of the biggest minds in AI. Yeah. And, you know, they did amazing things.
link |
And sometimes they were grumpy. Well, he was, he was interesting because he was very grumpy.
link |
But that, that was, I remember him saying in an interview that the key to success
link |
or being, to keep being productive is to hate everything you've ever done in the past.
link |
Maybe that explains the Perceptron book. There it was. He told you exactly.
link |
But he, meaning like, just like, I mean, maybe that's the way to not treat yourself too seriously,
link |
just always be moving forward. That was his idea. I mean, that crinkiness, I mean,
link |
that's the character. So let me, let me, let me tell you what really,
link |
you know, the joy memories are about having access to technology before anyone else has seen it.
link |
And so, so, you know, I got to Stanford in 1977 and we had, you know, we had terminals that could
link |
show live video on them, digital, digital sound system. We had a Xerox graphics printer. We could
link |
print, it wasn't, you know, it wasn't like a typewriter ball hitting characters. It could print
link |
arbitrary things, only in, you know, one bit, you know, black or white, but you could arbitrary
link |
pictures. This was science fiction sort of stuff. At MIT, the list machines, which, you know, they
link |
were the first personal computers and, you know, they were cost $100,000 each. And I could, you
link |
know, I got there early enough in the day, I got one for the day, couldn't stand up, had to keep
link |
working. So having that, like, direct glimpse into the future. Yeah. And, you know, I've had email
link |
every day since 1977. And, you know, the host field was only eight bits, you know, that many places,
link |
but I could send email to other people at a few places. So that was, that was pretty exciting
link |
to be in that world so different from what the rest of the world knew. And
link |
let me ask you, I probably edited this out, but just in case you have a story.
link |
I'm hanging out with Don Knuth for a while tomorrow. Did you ever get a chance at such
link |
a different world than yours? He's a very kind of theoretical computer science, the puzzle of
link |
computer science and mathematics. And you're so much about the magic of robotics, like the
link |
practice of it. Did you mention him earlier for like, not, you know, about computation? Did your
link |
worlds cross? They did in a, you know, I know him now, we talked, you know. But let me tell you
link |
my Donald Knuth story. So, you know, besides, you know, analysis of algorithms, he's well
link |
known for writing tech, which is in latex, which is the academic publishing system.
link |
So he did that at the AI lab. And he would do it, he would work overnight at the AI lab.
link |
And one, one day, one night, the mainframe computer went down. And a guy named Robert
link |
Paul was there. He led his PhD at the Media Lab at MIT. And he was, you know, engineer. And so
link |
he and I, you know, tracked down what were the problem was. It was one of this big refrigerator
link |
size or washing machine size disk drives had failed. And that's what brought the whole system
link |
down. So we got panels pulled off. And we're pulling, you know, circuit cards out. And Donald
link |
Knuth, who's a really tall guy, walks in and he's looking down and says, when will it be fixed?
link |
Because he wanted to get back to writing his tech system. Well, Donald Knuth. And so we figured
link |
out, you know, it was a particular chip, 7400 series chip, which was socketed, we popped it out,
link |
we put a replacement in, put it back in, smoke comes out, because we put it in backwards,
link |
because we were so nervous that Donald Knuth was standing over us. Anyway, we eventually got
link |
it fixed and got the mainframe running again. So that was your little, when was that again?
link |
Well, that must have been before October 79, because we moved out of that building then. So
link |
sometime, probably 78, sometime or early 79. Yeah, those, all those figures are just fascinating.
link |
All the people who've passed through MIT is really fascinating. Is there a, let me ask you to put on
link |
your big wise man hat. Is there advice that you can give to young people today, whether in high
link |
school or college, who are thinking about their career, or thinking about life? How to live
link |
a life they're proud of, a successful life? Yeah, so, so many people ask me for advice and have
link |
asked for, and I give, I talk to a lot of people all the time. And there is no one way.
link |
You know, there's a lot of pressure to produce papers
link |
that will be acceptable and be published. Maybe I was, maybe I come from an age where I
link |
would, I could be a rebel against that and still succeed. Maybe it's harder today.
link |
But I think it's important not to get too caught up with what everyone else is doing.
link |
And if you, well, it depends on what you want of life. If you want to have real impact,
link |
you have to be ready to fail a lot of times. So you have to make a lot of unsafe decisions.
link |
And the only way to make that work is to make, keep doing it for a long time. And then one of
link |
them will be work out. And so that, that will make something successful. Or not. Or not.
link |
Yeah. Or you may, or you just may, you know, end up, you know, not having a, you know,
link |
having a lousy career. I mean, it's certainly possible. Taking the risk is the thing.
link |
Yeah. So, but it, but there's no way to, to make all safe decisions and actually
link |
really contribute. Do you think about your death, about your mortality?
link |
I got to say, when COVID hit, I did, because we did, you know, in the early days, we didn't
link |
know how bad it was going to be. And I, that, that made me work on my book harder for a while.
link |
But then I'd started this company and now I'm doing full time, more than full time at the
link |
company. So the book's on hold. But I do want to finish this book. When you think about it,
link |
are you afraid of it? I'm afraid of dribbling.
link |
Yeah. I'm, I'm losing it.
link |
The details of, okay. Yeah. Yeah.
link |
But the fact that the ride ends.
link |
I've known that for a long time. So it's.
link |
Yeah. But there's knowing and knowing. It's such a, yeah. And it really sucks.
link |
It feels, it feels a lot closer. So my, in, in my, my blog with my predictions, my sort of push
link |
back against that was that I said, I'm going to review these every year for 32 years. That puts
link |
me into my mid 90s. So, you know, it's my, every, every time you write the blog posts,
link |
you're getting closer and closer to your own prediction of your, of your death.
link |
Yeah. What do you hope your legacy is?
link |
You're one of the greatest roboticist AI researchers of all time.
link |
Um, what I hope is that I actually finish writing this book and that there's one person
link |
who reads it and sees something about changing the way they're thinking. And that leads to
link |
the next big. And then there'll be on a podcast a hundred years from now saying I once read that
link |
book and that changed everything. What do you think is the meaning of life?
link |
This whole thing, the existence, the, the, the, all the hurried things we do on this
link |
planet. What do you think is the meaning of it all?
link |
Ah, well, you know, I think we're all really bad at it.
link |
Life or finding meaning or both.
link |
Yeah. We get caught up in, in the, it's easier to get, easier to do the stuff that's immediate
link |
and not through the stuff. It's not immediate. So the big picture or bad. Yeah. Yeah.
link |
Do you have a sense of what that big picture is? Like why ever look up to the stars and
link |
ask why the hell are we here?
link |
You know, my, my, my, my atheism tells me it's just random, but, you know, I want to understand
link |
the, the way random in the, in the, that's what I talk about in this book, how order comes from
link |
disorder. Yeah. Um, but it kind of sprung up like most of the whole thing is random, but this little
link |
pocket of complexity they will call earth that like, why the hell does that happen?
link |
And, and what we don't know is how common that those pockets of complexity are or how often,
link |
um, because they may not last forever, which is, uh, more exciting slash sad to you if we're alone
link |
or if there's infinite number of, oh, I think, I think it's impossible for me to believe that
link |
we're alone. Um, that was just too horrible, too cruel. Could be like the sad thing. It could be
link |
like a graveyard of intelligent civilizations. Oh, everywhere. Yeah. That might be the most
link |
likely outcome. And for us too. Yeah, exactly. Yeah. And all of this will be forgotten. Yeah.
link |
Including all the robots you build, everything forgotten. Well, on average,
link |
everyone has been forgotten in history. Yeah. Right. Yeah. Most people are not remembered
link |
beyond the generational too. Um, I mean, yeah. Well, not just on average,
link |
basically very close to a hundred percent of people who've ever lived are forgotten.
link |
Yeah. I mean, no long arc of time. I don't know anyone alive who remembers my great grandparents
link |
because we didn't meet them. So still this fun, this, uh, this, uh, life is pretty fun somehow.
link |
Yeah. Even the immense absurdity and, uh, at times, meaninglessness of it all. It's pretty fun.
link |
And one of the, for me, one of the most fun things is robots. And I've looked up to your work. I've
link |
looked up to you for a long time. That's right. Rod, it's an honor that you would spend your
link |
valuable time with me today talking. It was an amazing conversation. Thank you so much for being
link |
here. Well, thanks for, thanks for talking with me. I enjoyed it. Thanks for listening to this
link |
conversation with Rodney Brooks. To support this podcast, please check out our sponsors in the
link |
description. And now let me leave you with the three laws of robotics from Isaac Asimov.
link |
One, a robot may not injure a human being or through inaction allow human being to come to harm.
link |
Two, a robot must obey the orders given to it by human beings, except when such orders
link |
would conflict with the first law. And three, a robot must protect its own existence as long
link |
as such protection does not conflict with the first or the second laws. Thank you for listening.
link |
I hope to see you next time.