Josh Bongard: What are Biological Robots? How AI is Reshaping Life, Consciousness & Matter!

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
So I am interested in the philosophical side of things.
So one of the things that I see in the Xenobots and some of the
other kind of exotic robots thatwe work with is a conflation of
thought and action. So if you take a Roomba, the the
robot vacuum cleaner, you can point to the wheels and you can
say action. You can open it up and point to

(00:21):
the central, the CPU, and say, you know, thought or cognition
or processing traditional robots, there is a Cartesian cut
you can make that separates the body from the brain.
But what I'm excited about is ifyou look at a lot of cutting
edge technologies, that separation is becoming less and
less obvious, which I think is, you know, finally the beginnings

(00:46):
of an acid that's dissolving this distinction that at least
in the West has been around for over 300 years.
So I think that's great from a purely intellectual point of
view. But again, it, it also is
important for us as a species tounderstand that, you know,
brains are not everything and everything else is expendable.
You know, it's, the story is much more complicated than that.

(01:09):
And I think that finally overturning Cartesian dualism
will actually be a positive thing for for society.
Josh, I've been following you work for years and you guys have
done some incredible stuff in the field.
It's biology, robotics, artificial intelligence, so many

(01:30):
different branches we can go into.
But I think the best place to start would be let's start with
definitions. How do you define life,
consciousness and intelligence? Life, consciousness and
intelligence. So we're diving in the deep end
to start. OK, so to me I am very much in
favour, not surprisingly, of an embodied approach to life.

(01:54):
So life in many ways is a process of self construction.
This has been articulated I think, best in the idea of
autopoiesis, that you can construct your own components.
Those may be parts of your own body, parts of your own brain,
tools. They may be offspring, let you

(02:16):
know that it's almost a cliche, but life perpetuates itself,
which seems obvious and almost tautological.
But it is not a trivial thing toperpetuate oneself and
constantly self construct. And so that is to me the
definition of life. Not, none of our technologies,
no inert materials that we know of are capable of this.

(02:39):
And that's part of what interests me about robotics.
I believe in future we will create machines that are living
in the sense that maybe they're made of living materials, but
even if they're not, that they're able to constantly
construct the self and propagatethe self in various ways into
the future. So that that's my definition of

(03:00):
life. You asked about consciousness.
All, all cards on the table. I'm an illusionist.
I do not believe consciousness is a thing.
We've been looking for it for a long time, haven't found it.
I think it's just one of those things that's the best, the most
convenient term that humans havecome up with for something

(03:22):
that's epiphenomenal, that's notreally there.
I would follow in the footsteps of Daniel Dennett in this way,
that it is a form of illusion and we can point to optical
illusions, auditory illusions, motor illusions.
Our brains are very good at fooling themselves and I think
fooling themselves into believing they are conscious is

(03:44):
one of those things. And what, what was your third
dissideratum? Intelligence.
Intelligence. Aha, yes, OK, that one again,
I'm sort of a very pragmatist about intelligence is the
ability to make sure that you donot get painted into a corner in
the future. That's it.

(04:07):
I love the fact that you broughtup the fact that you think
consciousness is an illusion. Well I I guess the word there is
quite important and and difficult to work around because
you often have to backtrack whatthat means.
When I wrote my dissertation, itwas on illusionism as a theory
of consciousness. OK, there you go.
Followed in the footsteps of dand in it, Keith Frankish,
Michael Graziano and I put all of this together to defended

(04:29):
within psychiatry. And I find it interesting
because it's it's a very misunderstood theory of
consciousness. Well, it's not really a theory
of consciousness, but rather a theory that shows how other
theories of consciousness tend to lack substantive evidence of
what they're talking about. So it's almost like intuition
pumps, what Dan and it called for theory of consciousness,

(04:50):
sort of how to think of it. But when I look at your work,
it's almost easy to say that youmight be an illusionist.
And I thought that right up until now.
So you've you've managed to confirm what I've been thinking
all about. OK, great.
Let's start off with I've I've I've scheduled this.
Sorry, I've prepped this podcastwith 10 main questions in mind.
The first one in what you just mentioned embodied intelligence.

(05:12):
In the book you co-authored, Howthe body Shapes the way we
think. You argue that cognition is
deeply rooted within the body's physical form.
So how does this embodied perspective challenge our
traditional abstract models of mind?
I mean, this podcast is called Mind Body Solution, paying
homage to the infamous mind bodyproblem.

(05:32):
But therein might be a problem, the fact that we're
differentiating between mind andbody, separating these two
entities. Does this problem exist?
What are your thoughts of this? Yeah, OK, great.
So yeah, embodied cognition, this term, it means a lot of
different things to a lot of people.
I've worked in embodied cognition for many years now.

(05:54):
So my my particular perspective in it on it is exactly that,
that there is no distinction between what we conveniently
refer to as mind and what we maybe less problem problematically
refer to as the body. There is no separation.
There are some additional kind of add-ons, evolutionarily

(06:17):
recent add-ons like neurons and brains and central nervous
systems and prefrontal cortices,but they're the icing on the
cake. They're not the cake the nervous
system facilitates, things that organisms without brains were
able to do for a very, very longtime.

(06:37):
That when you get down to brass tacks, there isn't anything
really qualitatively new that brains have brought to the
table, although from a human perspective, it often feels like
there is. Again, illusions are creeping
back in. When when you think of embodied
cognition, how familiar or how much do you also consider what

(07:01):
psychologists call the 4E cogsaw?
I mean inactivism and embodied cognition.
And then you have the others like embedded cognition and is
the other E? And there's one more.
Yeah, I'm trying to remember now.
What it is? It's embedded.
OK, yeah, I, I, I think you knowthis.
My belief comports with a lot ofthose, you know, and now putting

(07:23):
on my roboticist hat, you know, the devil is in the details.
What exactly does it mean to be,you know, all the ease,
empowered, enacted, enabled, youknow, you name it, embodied that
it's at the surface, it's kind of intuitive.
You know, something or someone that has a body that's a
physical tool with which to interact with the environment

(07:46):
somehow, you know, intuitively has more options and ability to
prevent getting painted into a corner in the future, right?
It just kind of makes sense. But what I find fascinating in
the the research that I do and some of my colleagues do is when
you get down into the into the weeds, things become very non

(08:07):
obvious. So I co-authored how the body
shapes the way we think with my PhD advisor Roll Pfeiffer.
And when I was a PhD student with Rolf, I would always ask
him, you know, how exactly does the body shape the way we think?
What exactly is it about the, the body that neural networks
can't do? And we would have, you know, a
lot of heated discussions about that.

(08:29):
And and that's what I've dedicated my career to is to
trying to quantify and concretize, you know, exactly
what this means because we can build machines to try out some
of these theories. Whereas in the, you know, the
cognitive sciences and the psychological sciences, you
know, there's only so much we can do to humans to try and

(08:49):
understand how ease exist in humans.
But with robots, you can do anything.
So I think that's the differencebetween the fields, but I think
their long term goal is the sameas to really understand what all
these ES mean. You you actually touched on the
one. I think I might have made a
mistake, but the other E was extended and.

(09:11):
Yes. Those tools that we do use,
yeah. So like our cell phones,
everything else that we use to that form bottle cognition, that
in terms of evolutionary design of intelligence.
Your work in evolutionary robotics explores how machines
evolve adaptive behaviours and it's.
Can you explain how these evolutionary algorithms guide

(09:31):
this process and with such evolved systems could ever cross
into the realm of, let's say, consciousness or sentience?
Sure, sure. So, yeah.
So for those of your listeners that aren't familiar with the
field of evolutionary robotics, the approach is more or less
what the name implies, which is that we try and we create AI

(09:53):
that evolves bodies and brains of robots in simulation.
So I'm sitting here in my officeand on the far side of the quad
I can see the building where a supercomputer is housed.
And inside that supercomputer, in the GPU's, there's 10,000
virtual worlds running right now.
And inside each of those virtualworlds, there's a swarm of

(10:15):
virtual robots. And every once in a while, some
of those robots disappear because the AI is deleting the
ones that aren't doing such a good job at whatever we want
them to do. And the AI makes randomly
modified copies of the survivingvirtual robots.
So that that's the methodological approach.
This AI that I mentioned, it's, it's arguably the oldest form of

(10:39):
AI. It's called an evolutionary
algorithm, and you can trace it back almost to the 1940s.
Some argue that the very first computer program that was ever
written was something that looked kind of like an
evolutionary algorithm. And it's kind of, you know, an
intuitive idea. If you don't know how to solve
your problem, why don't you justcreate a population of random
solutions, measure how good those solutions are, delete the

(11:02):
bad ones, make randomly modifiedcopies of the survivors, and off
you go. The I think the second part of
your question was, well, how might this process lead to, you
know, more abstract forms of cognition like sentience and
consciousness? It's a very good question.
I've been working in evolutionary robotics for over
20 years now, and we tend to focus on, you know, sensor motor

(11:26):
tasks, behaviors rooted in sensory motor things, so
locomotion, object manipulation,swarm behavior, collective
intelligence, that sort of thing.
But we have dabbled in the more abstract, you know, aspects of
cognition. In some work I did in my post
doc with Hod Lipson at Cornell, we evolved a robot that evolves

(11:51):
understandings of itself. So, so it gets a little like in
the movie Inception. So you have a virtual robot and
it's trying to survive, and thatvirtual robot is running
simulations of itself in its ownvirtual head, and it is using
those guesses about itself and its current situation to guide

(12:14):
its behavior. So there's an old theory, you
know, in the biological and cognitive sciences that that
self-awareness, I don't know about consciousness, but
self-awareness evolved for very good evolutionary reasons.
If you are near the edge of a Cliff and you are able to
simulate what would happen to yourself, you can mentally
rehearse what would happen if you take three more steps

(12:36):
forward. Obviously you're in a much
better evolutionary situation than someone who cannot mentally
rehearse and has to physically take those three steps to see
what happens. So we have seen evidence in some
of our simulations where our robots start to evolve what we
call self models, a model of self, and they use it to avoid

(12:58):
risky behaviors, practice thingsbefore trying them in reality.
And you can see that how for many people, that's a stepping
stone towards very abstract things like consciousness.
And I think it, it's, it remindsme of some work done well.
Josh about this talks about something similar in his work.

(13:19):
But Nicholas Humphrey, he wrote a book called Soul Dust.
And he talks about one of the functions of consciousness also
having this, this self-awarenessor this awareness of being alive
in itself, the beauty of being conscious and, and, and
surviving in this world, which can be another force and driving
force with keeping us alive. So if that, if the AI eventually

(13:40):
discovers that the fact that it's there in the 1st place is
important in itself because it now has this ability to
experience and therefore will not take that further step if it
does process those thoughts of if I take a few more steps, I'm
going to fall off the Cliff. Is that something that you've
seen presenting itself in any form or any way?
Sorry, I'm sidetracking now. No, that's OK.
I, I, I haven't, but that's something that we certainly are

(14:03):
on the lookout. So if you are able to simulate
self, if you have a, you know, some sort of abstraction that
you can form of self, you know, obviously it's also useful for
social creatures. So if I can form a model of
self, I, you know, I can adapt that to form a model of other
and a good model of other like you is the assumption that

(14:24):
you're also running a simulationof yourself in your head, which
probably includes a simulation of me, which includes a
simulation of you. And so you know this, this,
these evolutionary pressures towards self-awareness very
quickly can lead to recursion and reflexivity, which is the
awareness of being aware. And for many people that is a

(14:48):
definition of consciousness. So.
That sort of meta cognition which most people talk about is
their own definition of what consciousness is.
There you go. No, it doesn't quite handle
qualia very well, but but it, you know, captures a lot of what
people think of when they think of consciousness.
I think to a lot of philosophersand psychologists, just knowing
that some artificial intelligences have this ability

(15:10):
to self reflect and develop these self models is probably
enough for them to think that this this is conscious.
I again, as an illusionist I'd say sure, absolutely.
You know, using the term consciousness to describe this,
you know this. This recursive self-awareness
seems useful as it is also useful for describing our

(15:32):
ability to be recursively self aware.
Let let's expand a little bit. I mean an illusionist when they
think about consciousness. I mean, we know that the brain
plays tricks on us all the time.Whether it's optical illusions,
auditory, doesn't really matter.Hallucinations and, and
experiences occur all the time. Just like AI confabulates, we
confabulate, we make up stories.When something happens, we

(15:55):
remember it incorrectly and we create and draft new stories to
explain this phenomenon. Are these all the very varying
reasons that you've come to thisconclusion that we, we
fundamentally are flawed species?
We're we're limited by biological processes, heuristic
adaptations and and and therefore we can't assume when
we ask what is consciousness in an AI that we even are conscious

(16:17):
in the 1st place. It's like we know well.
Again, I, you know, I, I'm, I'm perfectly fine with that non
problematic side of it, which again for most people captures
part of what they mean by consciousness.
As an illusionist. What I have a problem with is
that taking the further step to things like qualia and the
redness of red and you know thatthere's there are these things

(16:38):
called qualia and they're in there somewhere.
To me, that's a step too far. You know it, it feels as if we
have them. And I think it's an
epiphenomenal, you know, part ofsomething to do with
self-awareness. So that that's that part of it.
I don't, I don't buy. When you, you guys work, there's

(16:59):
the four of you with when it comes to biological robots,
which we're going to touch on soon.
But on the other side, I mean, you've got someone like Mike who
Mike Levin for all those who know, but but he's obviously
within the Pancychus realm in that he sees this cognition
happening in layers going upward.
And, and then you were on the other end saying actually now
these are just phenomenal phenomena occurring, but you're
both producing such amazing workas a collaboration in a unit.

(17:23):
How do you guys go about discussing these?
It doesn't, surprisingly. It doesn't come up too much.
I think we kind of nerd out on the nuts and bolts, you know,
experimental side of it. And then how each of us
interprets the implications of the work is sort of a private,
private affair, I guess. OK, well, you guys have done

(17:44):
something incredible and I always think about it and I
always wonder why this isn't spoken about more.
I mean, when you think of firstly, what are biological
robots, which we we call now Xenobots and thanks to you guys,
and I mean these living machinesthat you help design, they blur
the line between life and and machine.
And how do they force us to rethink our definitions of life,

(18:07):
intelligence, and even agency? Sure.
Yeah. So these biological robots, this
was a collaboration as you mentioned, between myself and
Mike Levin at Tufts and two of our collaborators, Sam Krigman
and and Doug Blackiston. And so over five years ago now,
the four of us were together on a funded project.
And myself and Sam, as the roboticist, showed what we could

(18:30):
do at the beginning of this collaboration, that we could
teach an AI to reconfigure virtual robot parts, you know,
in a supercomputer to make new virtual robots.
And we demonstrated all of this.And then a week later, Mike and
Doug joined us on the Zoom call.And Doug was a very talented
microsurgeon, showed us that he built a version of one of our

(18:53):
four legged robots under the microscope from frog cells.
And I'll never forget this moment.
There was complete silence on the call because this thing
looked like, you know, our robots, traditional robots, but
it was clearly some something biological.
And we found out it was made from frog cells.

(19:14):
And so Sam and I immediately asked Doug, we said, can you
take this statue and can we makeit move?
And it took Doug a few more months, but he figured out how
to put some muscle cells in there.
And then that was sort of the first xenobot, this little 4
legged creature walking along the bottom of a, a Petri dish.
So in, you know, the, the simplest explanation of a

(19:35):
biological robot is it's a robot.
It's something that's been builtby us, in this case, us being an
AI plus humans. It's been built by us to do
something we wanted to do, whichin this first experiment was
just walk along the bottom of a Petri dish.
A lot of folks don't realize it,but that's actually the original
definition of a robot. It's something built by humans

(20:00):
that runs around, is capable of moving about and does stuff on
our behalf and. In the original, the original
definition of robots, it comes from a Czech play over 100 years
ago. And in that play the robota were
built from protoplasm. It was some sort of biological

(20:21):
mass. But of course, then into the
20th century, along comes metal and plastics and ceramics and
eventually electronics. And so now, and thanks to
Hollywood, most people think of robots as things built from 20th
century materials. But we're in a way as
roboticists. This is very satisfying.
We're going back to our roots and we're tasking AI with

(20:44):
building robots from living materials.
And I think there's sorry, continue the no.
No, please go ahead, I. Was going to say, and I think
it, it would be more comforting to most people to see this type
of architecture within a system that's robotic in that it's not
scary, it's not dangerously, it's not metal, it's not going
to harm you or hurt you. It's softer, it's slightly, it's

(21:06):
easier to look at. Do you think that
psychologically this is an easy approach as well?
Just. For the my experience has been
the exact opposite of what you just described.
People are absolutely terrified of what have been been known as
the Zenobots. We, so we, we finished this

(21:26):
work, we published it in a very visible journal and it triggered
this, you know, huge media interest in obviously, you know,
frog bots built by AI. What, what's not to love?
So, so psychologically, it's kind of interesting because I
think the Xenobots have discovered one of the deepest
parts of the uncanny valley. So for folks who haven't heard,

(21:48):
the uncanny valley is something where there's something out
there that's kind of like us, but also in an unexpected way,
not like us. And that seems to press on some
very deep buttons in humans. So, you know, zombie, all the
zombie movies and TV shows that just never seem to end, you
know, zombies are a great example of something that's
that's in this, you know, uncanny valley.

(22:11):
And I think Chachi PT and how from 2001, these disembodied
authoritative voices from above,you know, that somehow also
presses our buttons. And now, and maybe in
retrospect, not so surprisingly,AI designed frogbots also
frightens a lot of people. I will say that it's mostly

(22:33):
older people, I think that are afraid at the xenobots.
Mike and I now, you know, on a daily basis get emails from
young folks saying how do I train to become a xenoboticist?
And you know, they, they think it's the coolest thing under the
sun. So it depends on who we're
talking about. How did the name Xenobot come
about? Who?
Who said that the first time? Yep.

(22:54):
So that was Mike, and I'm, I'm probably going to botch this
story, but I think he was sitting in on someone's PhD
defense and he saw a little bit of an animal cat.
This is one part of a frog egg, and it was moving about on its
own. And from Mike's perspective, it
looked like a little robot running around doing something.

(23:14):
And I think if I'm getting the story right, that's where the
name came from. But I think it fits quite well
for this technology because it has the bot part in these are
living machines that have been built by us for a purpose.
The Zeno comes from Xenopus Levus, which is the Latin name

(23:34):
of the the particular frog that we draw these cells from.
But Zeno is also the Greek cognate for like stranger or
newcomer, which was just coincidental, but it was a nice
a nice add to the term I think. Yeah, I think it's such a cool
name. I think I might have asked Mike
this many years ago and I can't,I can't remember if that if the

(23:55):
story matches me, I need to. I'll double.
S you can, you can compare, yeah.
See if you got it right. But these xenobots have
demonstrated self replication and something typically that's
associated with life only. And what does this ability tell
us about the relationship between reproduction,
intelligence and autonomy in living systems?
So this comes back to the top ofour interview where you were

(24:19):
asking me about intelligence andand life.
And to me these things are entwined.
So life is capable of perpetuating itself, making
stuff including parts of itself or copies of itself or
offspring. And the ability to self
construct is a very good tool ifyour goal is to not get painted

(24:39):
into a corner in the future, which for me is the working
definition of intelligence. It gives you options.
So one way of viewing the self replicating xenobots is that in
essence, we took individual frogskin cells and just like pulled
them apart. So this is like the a maximally

(25:00):
deconstructed frog incapable of its tree, of its usual way of
reproducing or perpetuating itself.
And these skin cells seem like they want to come back together
and they sort of somehow figure out how to make copies of
themselves. And I am, I'm adopting the

(25:21):
intentional stance here. I'm talking about frog cells
wanting things and having goals.I'm not sure whether that's
true, but again, it kind of makes sense.
If you're an Organism made-up ofa whole bunch of parts, a way to
be intelligent is to make sure you can reconstitute yourself no
matter what happens and continueon.

(25:41):
So one of the interesting thingsabout the self replication study
was, you know, life finds a way.And this was certainly an
unexpected way that had previously been unknown.
And I, I, I can sort of understand when you, when you
talk about the, the backlash, let's say when you guys

(26:03):
published the first batch of these papers, because when you
think about the ethics within biology and with the way people
perceived it years ago, any, anything relating to sort of
tinkering with an animal doing this or doing that with a cell
to them is sort of playing God. And it's, it's their, it, it
comes with a lot of backlash in general.

(26:23):
I was watching one of your lectures and I find it quite
funny when you spoke about the tadpole that now has an eye on
its, but I just said to myself, that's a joke.
I mean, today people watching it, it, it is something we, we
noticed that this thing was not harmed.
It actually still evolved tradesthat could do things.
The frog was fully functional, still able to turn around and
then still catch its prey, so these systems adapted

(26:45):
appropriately and technically noharm was done.
So do you think the ethics of just generally tinkering with
biology is what's coming into play when it comes to this,
rather than just being used to metal robots?
I think, yeah, you know, obviously this is one of the an
old theme, possibly one of the oldest themes in humanity.
Like what, what starts to happenwhen you tinker with things

(27:08):
you're not supposed to tinker with, right.
All major religions, their origin stories have to do with
some version of that. And I would say this is just the
latest chapter in that. And you know, that that's where
we as a global society are. We are now a technological
civilization. We are doing things far beyond
what our ancestors could do. And guess what?

(27:30):
It's starting to cause some problems.
So I, I can understand people, alot of people's perspective here
that, that we are tinkering withthings that are very powerful
that can deflect in directions we didn't think of and that can
really cause problems for us. On the other hand, you know,
that is a reality. So the most dangerous thing we

(27:52):
could do is just to stop innovating.
You know, if we, you know, we'reflying up an airplane, if we
turn off the engine, it not a good solution.
If if we are going to as a species or as part of this
ecosystem come back to a steady state, whatever that means, it's
going to be by going forward andinnovating rather than trying to
turn things off and go back. Do you think what a big fear

(28:16):
would be is sort of creating newforms of life that could outpace
or control us in a in a way thatwe did not anticipate?
I think so. So you mentioned outpace, right?
So you look at, you know, fear around AI technologies.
Again, it boils down to the speed at which things are
changing. And again, I think this is a

(28:38):
reality that we are only now just waking up to.
So 25 years ago, Bill Joy, the CEO of Sun Microsystems, an old
type of computer, wrote a very influential article back then
about exponential technologies. These are things that, you know,
grow at an exponential rate. And you know, we are now living

(29:01):
in the exponential age. You know, COVID is an example of
computer viruses. You know, the, the rapid
improvement of AI and it's exponentially growing powers.
We, it's just where we are. So for me, the xenobots are even
more important because they are sometimes an exponential
technology. They can grow exponentially if

(29:22):
self replication is involved. So we need to understand how how
exponential technologies work. You know, it was a great human
achievement that we developed the COVID vaccine.
You know, the, the, the Black Death lasted for centuries.
COVID is still here, but the worst of it, you know, was over.
In a few years, there will be things that spread exponentially

(29:46):
around the planet. And we are AI have to learn how
to create other exponential technologies that can bring the
bad exponential technologies down as quickly as possible.
And that requires understanding exponential technologies.
How was your? How was your experience with
encountering these xenobots whenhe showed you what he had done

(30:08):
with your work the first time you saw it?
How did this fundamentally either change the way you saw
the mind body problem or or or just impact you in general?
How did it change your philosophical or psychological
views on the concept? Yeah, sure, sure.
So it was a very visceral experience.
So as a scientist, you know, 99 things you try don't work.
It's failure after failure afterfailure.

(30:29):
And then every once in a while, if you're lucky, something like
the Xenobots comes along. So I remember I was here in my
office and my then PhD student Sam came in and showed me some
preliminary data that the xenobots that the AI had dreamed
up in our supercomputer. Doug was able to build them
under the microscope. So he, so the AI, in effect, was

(30:51):
right. It knew it knew how to rearrange
living materials to create a robot for us.
And that robot did what it was supposed to do.
And it was, it was Thanksgiving time here.
So there was no one in the research building here.
And I remember just getting up and going for a walk in the hall
and, and my legs just felt like jello.

(31:12):
Just realizing that the first realization was it's possible.
You can do this. We can make machines from
genetically unmodified living materials.
This is not GMO. This is not something else.
It's just, it's just possible. And for a scientist or an
engineer, you know, that's, that's the best part.
Just suddenly this whole new Vista opens up and then all the

(31:34):
implications since then, you know, are exciting and fun.
But that was that was the big moment for me at least.
Do you guys ever sit back and and just reminisce on that day?
Do you Do you talk about the? Not yet, but I'm hoping we're so
busy, you know, because it really has started to snowball
since then, you know, and then typical nerds and scientists.

(31:55):
We could just get caught up in the work and then, you know,
it's I'm sure one of these days we'll get, we'll sit back with a
beer and say remember when, but haven't haven't quite found that
moment yet. Well, you, you've touched on
this at the moment AI's role in shaping biology because I mean,
this clearly showed us what could be possible.
So it played a critical role in designing these innerbots.

(32:16):
Could you walk us through how AIis reshaping the way we design
biological systems currently, and what this might mean for,
let's say future artificial in life or evolution in general?
Sure, sure. So stepping back from the
xenobots for a moment, you know,AI and biology in general are
biological engineering, genetic engineering.

(32:37):
They're merging. So it's, you know, it's
impossible to say. There's so many exciting things
happening every week. So the first most obvious one is
genetic engineering, that that AI is being fed vast amounts of
genetic information. And so AI is learning.
If you tweak this gene, this will happen.
If you tweak that gene, this will happen.

(32:59):
And that's, you know, immensely interesting and exciting.
It's, you know, bringing us, it's bringing us a certain
grains, strains of rice that arevery helpful in developing
countries. You know, it's lifting people
out of poverty and and misery, dealing with some huge health
issues, Immensely, immensely powerful and helpful.

(33:23):
Then there is the generative AI side, you know, Gen.
AI where it's not looking at data and learning how to tweak
things, it's learning how to build things.
And so at the moment, the big successes there have been at the
sub cellular level. So there are now AI design
proteins and these proteins lookand act like no other protein

(33:44):
that humans have ever seen, right?
And that's revolutionary revolutionizing, you know, drug
design and you know, therapeutics in general.
So the the xenobots then is sortof Gen.
I Gen. AI at the Super cellular level.
So the AI is taking cells as thebuilding block, not peptides,

(34:06):
and putting those cells togetherto make things that are, you
know, a millimeter in size or, or a little bit larger to create
little tiny machines. And so stepping back, you know,
we're basically developing AI that is learning how to tinker
with life at all levels, sub cellular, super cellular, you

(34:26):
know, I don't know how how high up the scale we can go, but you
know, the trend suggests there'sthere doesn't seem to be a limit
insight yet. Have they has it?
Well, I'm not too familiar with the with the data within this
field though. But when it comes to these
proteins and these, does it evergive us insight into certain
amino acids? Like maybe we missed something,

(34:47):
maybe there's something we haven't yet figured this out.
Have there been cases like this where people are figuring things
out, where biologists just neverwould have never considered or
expected aside? Yeah, absolutely.
Now it's a much harder, it's a much harder lift.
It does happen. It's rarer because if you think
about it, you know the we've learned now that so AI is black

(35:10):
box. It's able to do some amazing
things and not bother to explainto us poor humans how it did it.
But you can make things harder on the AI by saying, please come
up with new things we never thought of and also come up with
explanations that we as pure poor humans will understand and
that, you know, that's a harder ask.

(35:33):
We joke in the lab, you know, from the AI perspective says,
oh, you want me to also have to try and explain this to you?
You know, that's really hard. But but it is possible.
And I, I've dabbled with some ofthat in a different branch of AI
where the AI takes in raw data and spits out equations.
It basically generates math to explain the data.

(35:54):
And you can read this AI generated math term by term.
And when this AI is working, it should generate the terms that
you as an expert in that field are familiar.
What with you would expect to see this term, that term.
Hey, wait a second. Why?
I don't recognize this term? That's where things get really
interesting. So yeah, it does.

(36:16):
It does happen. But we are the weakest link in
that chain, right? It's up to us to struggle to
actually understand what the AI is trying to tell us it.
Kind of reminds me of the film Her with Joaquin Phoenix.
We're at the very end. I mean this artificial
intelligence is just way too complex for him to even
understand. Yes.
I mean, we're there already. We're there already.

(36:38):
It's yeah, yeah, we're, we're living in that reality for sure.
The outpacing is a problem, I would say that that a lot of us
fear in general. I mean, even I, I myself as a
doctor, when I see how many of my colleagues would use AI to
draft motivation letters or evenjust to diagnose or try and
interact with patients, they send an e-mail reply.
I mean, there's a lot of legalities around the privacy

(37:00):
and obviously trying to not giveout personal information, but
but people are using this everywhere all the time.
It's in, it's almost inescapable, and we have to
consider this to be part of the extended cognition we were
talking about earlier. AI has now become fundamentally
a part of this extended cognition.
What are your thoughts about? Yeah.
I mean, the great, the greatest extension of our ability so far,

(37:24):
right? And it's hard to argue that it
isn't. It's Yeah, it is.
Yeah, I'm slowly struggling to see a day that goes by where I
don't use some sort of artificial intelligence.
To assist. And you probably used it 32
times today in ways you didn't even realize that you were using
it right. There's also that, there's the
there's the dark matter of AI that now exists in our world as

(37:45):
well. I can't remember who the author
was who mentioned the technological singularity, but
you. Think Ray Kurzweil.
Yes, yes. Do you, do you think we're close
to that? Do you think it's?
Yeah, I got him sounding like a cynic.
I don't believe in the singularity.
Things are moving fast and even possibly exponentially, but that
doesn't mean that they will continue to you.

(38:07):
You see, you see a lot of sigmoid curves in the world.
Things suddenly get really, really fast and then you start
to get diminishing returns. I don't, I've never seen any
phenomenon that just keeps goinglike this and I don't see why AI
would. Well, let's Josh, let's bring it
back. So this podcast obviously is

(38:27):
focused on mind body solution, the mind body problem,
consciousness, trying to understand this fundamental
question. And I can't help but think, but
I know you love sci-fi and and growing up, I mean, this was
something you were always fascinated by.
Have you watched Ex Machina? Ex Machina, Yes, yes, yes.
I think I remember the. Alex Garland So the the science

(38:50):
advisor was Maurice Shanahan andhe he spoke about how the, when
he was one of the best scenes inthe film was when, when Oscar
Isaac asks the I'm sorry, the guy who asks Oscar Isaac about
his artificial intelligence, he says, did he pass the Turing
test? And he says no, he doesn't need
to. So she doesn't need to because I

(39:11):
want you to know this is a robot.
I want you to know this is artificial intelligence.
The question is, will you still believe it's consciousness
thereafter? So so here the the cards out on
the table. We know this is not a conscious
biological being. You know, it's physically
artificial intelligence. But are you still fooled?
And he called us the Garland test since that form.

(39:31):
Do you think that's a bit it's more of an interesting approach
rather than having the normal Turing test, but rather know
this is a non, well a non biological being, but yet he's
super convinced that it is conscious.
And I think I don't know that ithelps much.
I would mention Thomas Nagel's essay.
You know what it is like to be abat?
So a bat is also not human. It is biological.

(39:55):
But, you know, humans have a long, you know, we've got a very
long history of looking at things that are not us.
And making decisions that's conscious, that's not conscious,
that maybe is conscious. I'll never know whether that is
conscious. So, you know, in terms of from
that perspective, suddenly having, you know, super
realistic humanoids or you know,how level type chat GPTS, what's

(40:21):
the difference? I I don't know whether a bat is
conscious. I also don't know whether
ChatGPT is conscious. I don't know whether it's that's
a well formed question. I don't know that it's that
different in that that way. Well, it's, it's, it's one of
those things where the philosophical zombie will always
come into play. It's, there's, there's
absolutely no way for me to lookat you and just assume that you

(40:42):
are actually a conscious being. It is, I only have access to
this mind and it, and, and all we do is taking cues and, and
sort of make relative guesses asto what we think is conscious.
Do you think we anthropomorphizethings way too much as a
species? Way too much.
Oh, that's a that's a good one. We certainly do it.
You know, an obscene amount of time.

(41:04):
I, I have AI, have a robot lawnmower at home and my 4 year
old, you know, this, this is, he's basically views it as his
sibling and he will get upset with us if we work the, the
lawnmower too much. You know, it's, it's everywhere
and and robots are just going toexacerbate it.
Now, whether it's too much, you know, you know, suggests that

(41:26):
maybe it's a bad thing. I think, you know, the main
thing that humans struggle with is increasing our circle of
empathy. So if we mistakenly, you know,
empathize with things that maybedon't there's nothing at home,
there's there's zombie or not. I think better to air on that
side than to air on the side of this is not something that's

(41:46):
worthy of our consideration, ourmoral consideration.
When you look back at someone like Isaac Asimov when he came
up with his laws of robotics andAI, and then you look back to
today with xenobots present, howdo you think that this field has
fundamentally shifted and changed throughout its course?
Yeah, I mean, Isaac Asimov, brilliant, creative and also

(42:09):
like all of us, a creature of his time.
So, you know, it was a lot of, you know, androids and, and that
sort of thing and, and metal andplastic and, but with the laws
of robotics, you know, I think he got that exactly right, which
is human hubris that we think wecan put guardrails around
complicated things. And so in in AI and robotics,

(42:32):
there's a famous term called perverse instantiation, which
means that the machine somehow instantiates the behavior we
want, but instantiates it in a perverse way.
If you ask an autonomous vehicleto get the human occupant from
point A to point B as quickly aspossible, fastest way to do that
is to drive in a straight line, you know, through parks and over

(42:55):
sidewalks and it's everywhere, right?
So, and it's a very difficult problem to solve.
We, we end up having to tell allour AI technologies, please do
X, but don't do it in this way and don't do it in that way and
don't do we keep tacking on these, these things, which
reminds me of, you know, all thethings that went wrong with the
three laws of robotics in Asimov's book.

(43:18):
So I would say actually he was very prescient and he got at the
root of what the problem would be, which is exactly the problem
We're all now wrestling. It's pretty incredible how these
sci-fi writers back in the day just guessed so many things so
accurately and well informed within their within their time.

(43:39):
It's so cool to see it all come to fruition today.
When you look at the field and and the work you does you guys
have done and that you've done over time, what do you feel most
proud about when you look back? What's the most exciting you you
think's been happening so far? Oh gosh, that's very hard to
say. A lot of the things we've done I

(43:59):
found very personally satisfyingin that for me, they were good
pieces of the puzzle in answering this question of how
the body shapes the way we think.
But some of those results are kind of, you know, abstract or
abstruse and and, you know, they're going to have probably
limited impact. So I, I guess I would say the
Xenobots for the simple fact that they've given, they've

(44:22):
given society a new way to make helpful tools, you know, if we
want to clean up microplastics, not leave additional metal and
stuff, and then when we clean up, great.
So I'm happy about that. And then there's also something
about the Xenobots, as I mentioned, that seems
inspirational to younger people.And that's something I care a

(44:43):
lot about. I consider myself an old person
or, you know, we we've made a lot of problems that
unfortunately younger folks are going to have to try and deal
with somehow. And as a scientist and engineer,
I feel morally responsible for trying to at least leave them
some tools with which to try andsolve some of these problems.
And I hope that Xenobots will beone of those tools.

(45:05):
When you look back when you weregrowing up, who who were the
scientists or writers Sign for writers, whatever that really
inspired you and got you to to like look at this field and
think about it so seriously too,to a point where you guys create
new organisms that have. Yeah, Yeah, It's a good
question. So I didn't, I there were no
scientists in my family. It wasn't something that, you
know, was a part of my childhood.

(45:26):
So I didn't actually really think about scientists or
engineers or what we now called STEM.
That wasn't a thing. Being a professor also wasn't,
you know, wasn't really a thing in my family, but it, it was the
sci-fi. And you know, what I took away
from it is, you know, all these things sound wonderful.

(45:46):
Where are they? Like if, and, you know, as a
young person, I could see back then, you know, there are also
problems. Like wouldn't it be great?
Like this literature seems to suggest there's other paths
forward. There's other ways to do things.
And then I would, from my limited vantage point, look
around and just, I remember justbeing confused, like, why not?

(46:07):
Why aren't we trying to? And then of course, I learned as
I got older that we were trying and failing miserably.
You know, I, I came of age in, in academia during the AI winter
when nothing was working. It seemed that, you know, robots
and AI was 1000 years off. And I, for me, I guess my
personality that the challenge appealed to me, the fact that

(46:31):
there were so many failures thateven if I entered the field and
was slightly less of a failure, that would be progress that, you
know, being able to contribute asmall bit.
And then suddenly the AI summer happened and, you know, all
these other things started to happen.
But but anyways, that's how I got to this point.
You you mentioned something likeSTEM, being a professor.

(46:51):
When you think about you, your work and how it bridges biology,
computer science, robotics, philosophy, so many different
fields, this interdisciplinary work that you guys are doing, so
many different people coming together.
How do you think that this has changed the game for you guys?
Because it wasn't a thing back in the day where, I mean, we're
all stuck in these niche fields and everybody's just doing their

(47:14):
own work. Even within biology, you'll have
a biologist who knows nothing about a specific molecule and
there's so little that everyone knows about everyone else's
work. But but I see you guys doing
this in such an interdisciplinary way that it's
it's pretty epic to watch from from the outside.
And how's that been for you? Well, first of all, it's been
very difficult and it's been very slow going.

(47:36):
So we've been talking about the Xenobots, which is kind of the
end, you know, or it's a, it's astopping point on a very long
journey. I was very fortunate and again,
just by chance to have fallen into various programs at
different universities in, in four different countries that
taught me how to be an interdisciplinary researcher.

(47:58):
There are ways, you know, to do it very badly, right?
So, you know, Jack of all trades, master of none, you
know, there, it's not an easy thing to do that The, the reason
that most people are specialists, and I'm glad there
are specialists, you know, is because it's easier to drill
down and really grasp a branch of human knowledge and extend it

(48:22):
than it is to try and connect two branches together.
It takes at least twice as long.So in the case of my
collaboration with Mike, I told you about, you know, the, the
Xeno sculpture and then the xenobot.
But you know, it took a long time to get to that point.
We, we don't even speak the samelanguages.
You have to learn, you know how what each are capable of, what

(48:44):
takes a week, what takes five years, you know all those sorts
of things. And then you can proceed
carefully. And of course, in some cases all
of this effort is worth it because you get lucky with
something like the Xenobots, butit is definitely not the norm.
Yeah, and then you get stuck in a podcast that's philosophical
based on the mind body problem being asked.

(49:06):
What is consciousness? That's right, that's right.
You got to roll with the roll with the punches.
Tell me, Josh, when you, when you think about this field
moving forward, what excites youmost about it?
Where, what do you think in terms of the the future of the
industry, people that are working perhaps under you or
work that you've seen what what's what's been most exciting

(49:27):
and even within your own work? Yeah, so, so I am interested in
the philosophical side of things.
So one of the things that I see in the Xenobots and some of the
other kind of exotic robots thatwe work with is a, a conflation
of thought and action. So in a, if you take a Roomba,
the the robot vacuum cleaner, you can point to the wheels and

(49:50):
you can say action. You can open it up and point to
the central, the CPU and say, you know, thought or cognition
or processing, you know, traditional robots, there is a
Cartesian cut you can make that separates the body from the
brain. But what I'm excited about is if
you look at a lot of cutting edge technologies, that

(50:11):
separation is becoming less and less obvious, which I think is,
you know, finally the beginningsof an acid that's dissolving
this distinction that at least in the West has been around for
over 300 years. And so I think that's great from
a purely intellectual point of view.

(50:32):
But again, it, it also is important for us as a species to
understand that, you know, brains are not everything and
everything else is expendable. You know, it's the story is much
more complicated than that. And I think that's finally
overturning Cartesian dualism will actually be a positive

(50:52):
thing for for society, assuming it actually happens.
You know, the Keith Frankish, the one who termed it, coined
the term illusionism, He's writing a book called Escaping
Descartes Prison and it's fundamentally premised on that,
on that whole idea that this particular dualism is, is
definitely on a decline, however.

(51:12):
I know, I know Keith, and I'm glad to know he's also like,
wielding an axe from a differentdirection.
Well, we're doing our empiricistbest, and yeah, hopefully the
tree will come down. But but on the flip side, if you
think about it, Josh, while thatCartesian dualism is breaking
down, you've got a rise in Ben psychicism and idealism as well.
So you've got a whole group of thinkers that think that either

(51:34):
everything is conscious or everything is consciousness.
What are your thoughts on that? I know this is not your field.
Now we're just going beyond. You know what, I'm OK with it.
I'm I'm OK with that. It's actually the the the stance
in the middle that I find problematic.
I am conscious and that thing over there isn't.
So it doesn't matter what I do to it.

(51:56):
Like that's the problem. So if there are a greater number
of folks that believe everythingis conscious and we should be
careful about how we interact with other things and other
other entities, great. If the illusionists say, listen,
we're all on the same level playing field, you know, for
different reasons, that's also good.

(52:18):
I think it's yeah. My problem is the actually the
intermediate stance. Yeah, I agree.
I think ethically and, and sort of the the the moral philosophy
behind both of those, whether it's illusionism or panpsychism,
pretty much lead to the same thing and that we're all the
same thing. We're all the same thing,
exactly. And I think that's all the
lines. And even within an idealist

(52:39):
philosophy where they think thateverything is consciousness and
we're just a part of it, that also tends to always lead to
some sort of a this consciousness trying to love
itself, enjoy itself. It always leads in some sort of
an empathetic positive worldview, which I think in
essence would be a better optionanyway, instead of drawing these
fundamental lines between one thing or another.

(52:59):
But when when you think about your work currently, what are
you guys doing that you think weshould be looking forward to the
most and that you're most excited about?
Yeah. So this is true of the xenobots
in general, but we are also doing a lot of work in my lab
with what are called meta materials, and these are human

(53:21):
engineered materials that act very differently from natural
materials. And it turns out that these meta
materials also are capable of thought or cognition and action
simultaneously. Imagine that I take a sheet of
this material and I start vibrating it, but I vibrate it

(53:41):
at different frequencies. It turns out you can get this
material to do different things at these different frequencies
and actually use these differentfrequencies to carry
information. You can you can design A sheet
of this material that when you shake it, it performs logical
AND at one frequency and logicalOR at the other frequency

(54:04):
simultaneously. And while computing that
function, the material is movingand if you put it on the ground,
it will actually move along the ground.
So where is the distinction between thought and action?
So in answer to your question, I'm excited about, you know, all
of these exotic materials, whichincludes living materials that

(54:27):
conflate thought and action. And then a is ability to exploit
the potential of these new materials to make useful
machines in ways, you know, Hollywood hasn't even begun to
explore. They will, I guess, but not not
yet. They always eventually catch up
and but but that's incredible. I mean that the, there's there's

(54:49):
so much ground breaking stuff occurring that I mean, you're
right. The the pace even for me when I
watch what you guys do, sometimes when I look at it, I
think like this is crazy how fast this is happening.
Like you see five years ago one Xenobots made and now suddenly
you guys are making materials that can vibrate at different I
think. Yeah, I mean, the the reason I
would say is again, it Mike and myself and some others, you

(55:11):
know, have been on a very long journey to try and see things
differently, to escape from the Cartesian worldview, right.
So if you don't escape, you know, you do things like ChatGPT
and you try and make bigger and bigger brains, which I would say
is also useful. Nothing against, you know, open
AI and ChatGPT and Stable Diffusion.
They're they're great. But the reason I think that that

(55:34):
we are discovering a lot of things is because we're not
looking under the lamppost anymore, right?
We're out in the dark and it turns out there's a lot of stuff
out there. It's just that people haven't,
haven't thought to look there, been able to get there because
you have to think in a non Cartesian way to get there.
You have to ignore a distinctionbetween body and brain, and then

(55:56):
you start finding these weird materials and finding ways to
bend them to new purposes. Something I found particularly
fascinating when I listen to youspeak was you mentioned
something about cause and effectand, and until an artificial
intelligence is able to actuallycause an effect, that's when you
know that this thing is has become a fundamentally part of

(56:18):
reality. Do you want to expand on that
idea? Sure, sure.
So this is something that current AI technology struggle
with cause and effect. You know, the best, the best a
non embodied AI can do is sort of read about all cause effect
relationships that humans have written about on the Internet or
in books or whatever. That's that's all those AIS have

(56:41):
access to and they can do OK with cause and effect.
Not great, but again, you know how the body shapes the way we
think. Like I said, I have a four year
old at home and he pushes a coffee mug off the table and it
shatters like he caused an effect.
And not only did it shatter, buthe sees how his two parents
react and how they react differently to what's there's.

(57:03):
When you can push against the world, literally, you know,
there is a massive amount of rich effectual data that comes
back. And so I say this a lot.
Embodied cognition is about pushing against the world and
observing how the world pushes back.
And that is a cause effect loop.You can cause them in all sorts
of ways from, you know, pushing a coffee cup off a table to, you

(57:27):
know, changing a society, landing someone on the moon, all
sorts of things. And that that gives embodied
learners, not just human learners, but non human
learners, you know, front row seat for cause and effect.
If you're not sure, if you're thinking that this effect is
caused by that cause, just go out and verify it be be a

(57:49):
scientist, push against the world and see whether you were
right or not. It's actually very simple when
you think about it. So I think there are some of
some of these examples that makeclear why we can't just ignore
the body. It matters in very important
ways. Yeah.
It runs, I think it was a paper I read.
I can't remember the author's name, but I'll try and find it.
But it's called the nobody problem.

(58:10):
Just that there's a problem in having nobody.
That's the thing. It's it's a fundamental.
That's the real problem. Agreed.
Nobody. Is a big problem.
Yeah. And it's it's it's, yeah.
It goes down to the crux of whatwe're talking about.
That same cause and effect that you're talking about is that
almost it reminds you of that inactivist part, the fact that
we're here trying to consistently and continuously do

(58:32):
something like there's a purposedriven by behind everything we
do. We must find food.
We must. We see it.
We see something red. We know it might be an apple.
There's, there's this interaction with reality that's
always looping back and forth. And do you think fundamentally
with AI we would either have to program it to have some sort of
a personal loss of if it does not exist, we must have some

(58:55):
sort of program within it or does it not need something like
that in order to? Be I, I, I think it absolutely
needs it, but not necessarily inthe literal sense of, you know,
making robots that push against the world and observing how the
world pushes back. All the, you know, ChatGPT and
all the chat bots, they kind of already are doing it because

(59:16):
they say stuff and that affects their human interlocutors, some
of whom go off and do things andthen report back to ChatGPT,
which becomes part of its training set.
So, you know, modern AI, they already have, you know, slaves,
which is us. You know, they can use us to to
try out cause and effect relationships.

(59:37):
I don't know, you know, what thebig AI companies are doing.
A lot of them are very smart. They probably started to figure
this out. So, you know, there are ways
that you can learn about the real world.
You can learn about cause and effect where you have something
or someone else go and, you know, embody that cause and
effect loop for you. So it's already happening.

(59:59):
It's a crazy thought. In other words, we are AI's
extended mind. We we absolutely are.
So it is an extended mind already It's got I don't know
how many of us as its end effectors.
It's it's happening. Do you think that a, a, a bigger
step would to take this even further would be then giving it
this sort of sense of loss? So having such a hard wired

(01:00:21):
program that if it ceases to exist, it's a, that's a
fundamental problem, the same way we are so afraid of death.
You think that's a big factor that we'll have to consider in
the future to make this? It's a it's a really good
question. I don't know whether, like, a
fear of death or extinction is anecessary component for
developing embodied cognition. If it is, that's going to be

(01:00:44):
problematic because this is something Hollywood has explored
in great detail. If you know, if you give
technologies existential dread, you know they'll do everything
they can in their power to avoidit.
And I agree with it. That just makes sense.
So I, I think actually it's a research question.
Like we started talking about self-awareness and how that

(01:01:06):
might have evolved in us to avoid extinction, avoid death.
But it's not necessarily true that technologies can't become
self aware and practice cause and effect internally and verify
it physically that that all needs to be driven by, you know,
fear of extinction. I hope it it's not true.

(01:01:28):
I hope that, you know, they don't have to rely on that for
their motivation. But I guess, I guess we'll see.
A lot's going to probably changethe next few years and I think
the, well, as we said, it's veryfast-paced, not to the point of
a singularity, but very fast in general.
When you look back at your viewson consciousness in the mind,

(01:01:49):
how often have you changed your views regarding consciousness
and the mind body problem? So when you were young, let's
say prior to entering the field as a student perhaps, what did
you have a Cartesian mindset on this?
Did you think there was a sort of soul or?
I did not have a religious, yeah, I did not have a religious
upbringing, but yeah, I was convinced there was someone in

(01:02:11):
there. But, and that's just because I
was raised in the West, pretty much everyone, even if you're an
atheist, that's what you're taught indirectly and directly.
So, so my awakening was at the University of Sussex.
I did my masters there and Margaret Bowden and Phil
husbands and Inman Harvey and Anil Seth.

(01:02:33):
You know, they did a very good job of disabusing us of that
notion, or at least, you know, presenting it that it's just one
view. And so that was that was the
start for me. And then, you know, then
developing self aware robots andactually kind of how easy it was
to do that. What wasn't super easy, but

(01:02:53):
surprisingly easy. That was the nail and that was
the final nail in the coffin forme.
I was like, if we can make this $29.00 robot, you know, be able
to create an awareness of self, you know, it's not really
something special. It's probably, you know, it's
not not we should take it off the pedestal, let's put it that
way. Do you think there'll be a

(01:03:15):
fundamental shift or difference when you're able to give these
self aware agents agency or or or allow it to make choices and
and then interact with those cause and effects at some point?
Yeah, absolutely. Because one of the things I love
about AI and robots is like how creative they are.

(01:03:35):
So, you know, there's all these hilarious AI fails, but they're
also, you know, some of them arereally creative.
So it'd be great if they're alsocreative wins.
You know, like, again, when the AI says, I, I'm going to make a
protein like this and all the human experts on the planet say
you can't make a protein like that.
And then of course it does. Like, that's it.
So, you know, humans can only push against the world in so

(01:03:58):
many ways. And wouldn't it be great if we
had allies, you know, here in the real world that could push
on the world in ways that we just can't?
And they're going to learn causeand effect relationships that
would have been impossible or would have taken us a very, very
long time to discover. That's the future that I'm
excited about. And I think it's, I think it's

(01:04:20):
incredible work. You guys are really changing the
game. And the there's, there's so many
different aspects to the work that we haven't really even
touched on because there's so much greater detail that we
could go into. But is there anything particular
that you'd like to touch and that you feel like you haven't
mentioned about your work in general, Josh, that because this
is obviously just the first podcast we've had together, but

(01:04:41):
there's many papers we could dissect in so much detail.
But before we ever do, is there anything particular you'd like
to mention? Well, one thing we, we, I think
we mentioned in passing, but didn't have a chance to address
is how the, the work on xenobotsand even AI design biology is
changing like our understanding of life itself.

(01:05:02):
You know, most of us were taughtthat, you know, frog DNA codes
for frogs and human DNA codes for humans.
And what I've learned in workingwith Mike, and also just looking
at the intersection between AI and biology in general, is that
the species and organisms that exist on the planet, they're
points in an attractor space. So this sort of means that when

(01:05:25):
you're in familiar environments,your genes tend to build humans
or frogs or what have you, but that you can change the
environment in quite drastic ways, including the embryo
itself, or even take it apart into its component cells and put
them together in new ways. And without changing any of the

(01:05:46):
genetics, you get new form and function.
So there there's an old idea in biology of what's called Morpho
space, and it's this imaginary high dimensional space of every
possible living thing that couldsurvive on this planet.
And despite how creative Mother Nature's been over the last 3.5

(01:06:07):
billion years, she's only explored A vanishingly small
part of Morpho space. So, you know, the obvious
question is what else is out there?
You know, everyone you know is curious about aliens, and maybe
we'll find some on another planet someday.
But we can kind of find them like now here by asking an AI to
say, here's frog, here's human, here's axolotl.

(01:06:29):
Now go in the opposite direction, put them at your
back, and drive into morphal space as far from familiar
organisms as you can to find those that can exist here.
And then build them for us, please.
The, the prospect of that, you know, is going to revolutionize
biology, you know, beyond recognition, I think.

(01:06:50):
I, I literally had one of those questions down to to further
elaborate on that. Remember when I told you about
the X, the extra questions I hadplanned for us?
Yes, when when you it brings me back to one of the other
questions I wrote. You've used evolutionary
algorithms to evolve solutions to both software and hardware.
How do these algorithms reflect,or perhaps extend the principles

(01:07:12):
of natural selection? Oh, yeah.
Gosh, I don't know. I mean, I'm not a biologist, so
I'm sort of speaking beyond my professional expertise here.
I don't know that it really is beyond natural evolution.
I mean, we're just, you know, AIis a product of us, and we're a

(01:07:33):
product of natural evolution. And the AI is modifying or
putting pressures on these organisms in ways they haven't
experienced before. So, you know, from a bit of a
distance and if you squint, it is just, you know, natural
selection. It's just entered a new chapter,
like when, you know, the fish crawled out of the Seas or we
came down from the trees. It's just feels like a different

(01:07:56):
phase in a similar process. So I, yeah, I don't know if
we've broken out of the bounds of natural selection yet.
It it is unnatural, of course, because there's things like AI
and robots and a supercomputers involved, but ultimately it's
just challenging genetically unmodified materials to survive

(01:08:19):
and continue, and they're finding a way to do so.
The I I agree with you. I think I had this conversation
with someone recently where we spoke about the fact that it
took us billions of years to getyou to have this sort of
experience. It doesn't mean that if we're
able to do it within 5 minutes in the lab, that that this is
some sort of a miracle or it's just literally all that work

(01:08:41):
that's taken place, as you know,with the cells that you take.
It's already got so much prior information and it's just always
working on itself. So it's not like we're going
beyond the bounds of of nature itself.
And I think, you know, the otherthing is we talked about the
Cartesian dualism. A lot of this work, not just
ours but others, as well as exploding the concept of self.
We tend to view things about like at the self level, if you

(01:09:05):
think about the xenobots and that like the cells that make up
the xenobots from the cells perspective, things might not
actually be that different, you know, or problematic.
When a, when a, when a frog develops or any Organism
develops, the cells that are part of the developmental
process, they're being pulled and pushed and sheared and, you

(01:09:26):
know, shoved through tubes like crazy all the time.
So all of our manipulations for from the perspective of the cell
might just be, you know, business as usual.
It's find your neighbors and connect to them and establish
calcium communication again. And you know, so I agree, like
the xenobots in some way seems miraculous if you think about it

(01:09:49):
from the frog's perspective. But if you think about it from
the frog's cells perspective, maybe it's not so, you know,
surprising or traumatic. And I think that's where I think
Mike at some point, and that's when they see cognition all the
way going all the way down, is that it's really, really hard to
actually differentiate this, this type of process in terms of

(01:10:10):
cognition. When you look at a Xenobot and
you see it exhibiting some sort of autonomous behavior.
Have you ever found yourself ascribing some sort of agency or
free will to this To this? I can feel the temptation to
want to do so. Let's put it that way again.
I'm an illusionist. I know I'm being fooled.

(01:10:31):
I can feel it. And yes, sometimes it takes more
effort than other times to resist.
Have you always? At what point have you actually
said firmly I'm an illusionist? Was this something you one day
read or was? I think I was.
And then I met Keith and Dan at a workshop a while back and and
learned this actual term. I said that's it.

(01:10:52):
That's what I am. Yeah, I think.
I think I felt it before I knew what I was.
That's exactly the way I felt about it.
And and then one day watching Dan and then Keith eventually as
well because he helped me with my dissertation, which is quite
cool. OK yeah, Nicholas Humphrey also
gave some input and Mark solms. Great, which is pretty cool.
And I actually have a chat coming up with Mark and Carl 1st

(01:11:14):
and they're gonna have a nice chat together about some of the
work you can do as well. But that's great.
Then after this podcast, it's weird how my, my firm
illusionist beliefs have slowly like become slightly less
interior. And I thought, I haven't seen, I
don't see myself changing it anyway, but but I'm far less
firm about the belief than I used to be.

(01:11:36):
But I wrote the dissertation very, very firm illusionism.
And yeah, to the point that it'sjust, to me, it just made the
most sense that, that we're justplaying mind games with
ourselves all the time. It's it's a very difficult thing
to understand. True.
Agreed. Agreed.
It's a strange phenomenon in terms of your work, everything
you've done. How is this overall impacted

(01:11:58):
your entire view of this field of understanding the mind body
connection? I think what it what it's
convinced me of is that whateverthe answer is, I don't think we
have it yet. It's it's going to be very
confusing, very non intuitive. This is what I've learned as a

(01:12:19):
student of AI and robots is thatthings that work, you know, for
Mother Nature, they're very confusing, non intuitive.
So I think there are some answers to, you know, some of
our deeper questions aside who we are, what makes a special, if
anything, you know, how the bodyshapes the way we think.

(01:12:39):
Well, you know, what are aliens going to look like and how are
they going to behave? Whatever the answer is, it's
not, you know what I've learned from the xenobots in these meta
materials? They act in these very, very
strange, surprising ways. And I think we're just
scratching the tip of the the iceberg, you know, the fact that
huge matrix multiplications can give rise to things that look

(01:13:03):
like, you know, brilliant human conversation.
Like it we're being things are moving quickly and everything
that happens is surprising to tothe public in general and also
to the experts. And to me, that's changing my
world view to be like whatever we thought the answer was.
It's much, much stranger. It, it kind of feels like the
quantum revolution for like everything other than physics is

(01:13:27):
now happening. Physics already had its, you
know, brush with the the ineffable and unexplainable.
Now it's our turn. And, and within, I mean, quantum
computers, quantum biology, these are on the rise.
This is this is a thing that's happening.
So people are taking this very seriously, but these meta
materials are really intriguing to me.
And where do you think this is going to lead?

(01:13:47):
Oh, I, I think, you know, meta materials are the future.
Like, so you, you know, all the materials that we build with,
they have problems, you know, you, you can only build a
building so high because the weight ratio, you know, so meta
materials are amazing. Like I, I see a future in which,
you know, we can cut a robot in half and the two halves will

(01:14:10):
form 2 smaller versions of exactly the same robot.
You know, living materials like cells, when they attach to rigid
materials, those cells become more rigid.
When they attach to soft materials, they become softer.
They're cells are like chameleons.
And so I think we're going to make meta materials that are the

(01:14:32):
same way that they, they get thehint when you try and use them
to be something, they become that thing.
They construct what they need orthey change, you know, that's
what, that's what 21st century technologies are gonna look like
as we push further into the 21stcentury.
I think that's one of this is probably one of the most
exciting or or sci-fi ish thing I've heard, and quite some I

(01:14:57):
think it's. Again, we were talking about
surprise. The surprising thing is a lot of
those materials already exist. You know, they're just kind of
in labs at the moment. They're not in general use, but
but the aerospace companies, youknow, they have R&D labs around
these things. And so these, these, these
materials, ironically, you know,they're being, these materials

(01:15:18):
are being designed by AI. And then once they're built,
designed by the AI, the AI can figure out how to exploit them.
You know, how, how can they bestbe incorporated into buildings
and roads and cars and plumbing and you name it.
It's, it's amazing. It hasn't quite reached the
public, you know, eye yet, but it's, it's coming.
It's kind of crazy because it's almost like, I mean, we're

(01:15:40):
making biological robots and then at the same time you're,
well, you are, but at the same time you're making now materials
more biological. It's kind of like you're doing
the both, you're doing like the flip side on both sides.
It's just pretty crazy and cool and.
Yeah, not just my lab. Some of the exciting things that
are going on in these interdisciplinary labs are

(01:16:02):
exactly that, like blurring the line between living and non
living materials, right. So again, there's another
distinction that we just assume is so obvious and so clear cut.
And from the AIS perspective, itsays why I, I don't understand.
I'll put, you know, I'll, I'll embed some proteins in a plastic
to make, you know, a protein infused plastic.

(01:16:25):
So what is that Now? Is that a living thing, a non
living thing? I don't know.
Yeah, just like make the make the metal some sort of an ion
channel to assist with like shifting cells.
I mean, you can. It's exactly that.
When you think about what the potential this has and how much
this can affect medicine and every other field, it's, it's so
widespread that it would literally change humanity.

(01:16:45):
Like it's, it's pretty crazy. I agree.
I thought it's Max. You don't, you don't need, I
don't think we need the the singularity.
You know they'll be sufficientlycrazy stuff that will happen
without us all uploading to the cloud or.
Yeah, no, this is this is probably the most exciting thing
I've heard. Like and trust me, when you
scroll online and you do scroll,you see a lot of crazy things,

(01:17:05):
but this is today, so this is pretty cool.
I mean, this, I can't wait to see what, what you guys come up
with in and I'm really looking forward to it.
Yeah, It's, it's, it's incredible.
Yeah. No.
So, Josh, thank you so much for this wonderful conversation.
Any, any final words from your side?
Anything you'd like to add? Well, just just as we're
wrapping up, you're asking me about some other things that we
might not have talked about. Just one last thing I wanted to

(01:17:26):
mention about connecting AI and embodied cognition is where we
are at the moment. So we now have these immensely,
you know, powerful AIS that say compelling things, but
instinctually we're not sure whether to trust what they say.
So we're doing some work here inthe lab and there are others

(01:17:47):
that are doing so as well, whichis, you know, combining a non
embodied technologies like Chachi BT with embodied
technologies like robots. And So what we see on the
horizon is a human has a problem.
They come to an AI, they ask fora solution and the AI proposes
something that sounds great. And the human says, I don't know

(01:18:09):
that I trust you prove it. And the AI says, connect me to a
3D printer, prints a robot that goes out and, you know,
physically verifies, does some experiments, you know, to show
some version of this solution orsome aspect of it.
So can we teach AI to design experiments that will, that will

(01:18:34):
prove whatever it is that it's proposing?
So I think that's something elsewe haven't talked about that's
often known as like AI science or AI driven science or machine
science that also I think has huge potential for dealing with
at the moment, which seems like an intractable problem that
there's no way to guarantee whata ChatGPT says.

(01:18:54):
And I agree, doesn't matter whatit or anyone says the proof is
in the physical world. And I think we're going to start
to see some of that emerging in the not too distant future.
And so far, Josh, have you seen anything remotely close to that
or how how far off are we? We're close for, for toy

(01:19:16):
problems. I, I guess we're, I would say
we're for in terms of like practical utility, we're still a
ways off. I, I don't even know that we're
ready morally or ethically to allow an AI to build things to
prove its ideas about the world,that there's another one for
Hollywood to play with. But that that.
Actually is quite a good script for a film that is.
Yeah, yeah, exactly, exactly. No, no raw, yeah, no lack of raw

(01:19:40):
material here. But anyways, I just wanted, I
guess I wanted to close just by saying I don't think, you know,
it's non embodiment against embodiment that some of the
really valuable technologies andsolutions in the years to come
are going to be, you know, clever ways of combining them so
you get the best of both worlds.And in general, I mean life and

(01:20:01):
not life. That dichotomy, that false
dichotomy, it seems that at somepoint we're all going to be
chemically bonded in some way. Yeah, Yeah.
I think it's a beautiful. That's a beautiful way to
probably end it there. Josh, thank you so much.
Thanks for your time. Wonderful work.
No, thank you, Tevin. It was a pleasure to be on your
show. Yeah, no, I can't wait to
hopefully have around 2 and and further dissects maybe specific

(01:20:23):
papers and we can go sure some deep dives in the future.
It's it's a privilege to chat toyou and I really appreciate it.
Yeah, likewise. And again, thanks, thanks for
your interest and you clearly put a lot of time and effort
into to preparing this was this was a real pleasure.

All Episodes

Episode Transcript

Popular Podcasts

On Purpose with Jay Shetty

Ruthie's Table 4

The Joe Rogan Experience

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Josh Bongard: What are Biological Robots? How AI is Reshaping Life, Consciousness & Matter!

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}On Purpose with Jay Shetty

Ruthie's Table 4

The Joe Rogan Experience

All Episodes

Josh Bongard: What are Biological Robots? How AI is Reshaping Life, Consciousness & Matter!

On Purpose with Jay Shetty