Kind: captions
Language: en
the following is a conversation with
Jeff Hawkins he's the founder of the
redwood centre for theoretical
neuroscience in 2002 and Numenta in 2005
in this 2004 book titled on intelligence
and in the research before and after he
and his team have worked to
reverse-engineer the neocortex and
proposed artificial intelligence
architectures approaches and ideas that
are inspired by the human brain these
ideas include hierarchical temporal
memory
htm' from 2004 and new work the
thousands brains theory of intelligence
from 2000 17 18 and 19
Jeff's ideas have been an inspiration to
many who have looked for progress beyond
the current machine learning approaches
but they have also received criticism
for lacking a body of empirical evidence
supporting the models this is always a
challenge when seeking more than small
incremental steps forward in AI Jeff was
a brilliant mind and many of the ideas
he has developed and aggregated from
your science are worth understanding and
thinking about there are limits to deep
learning as it is currently defined
forward progress in AI is shrouded in
mystery my hope is that conversations
like this can help provide an inspiring
spark for new ideas this is the
artificial intelligence podcast if you
enjoy it subscribe on youtube itunes or
simply connect with me on twitter at lux
friedman spelled fri d and now here's my
conversation with Jeff Hawkins
are you more interested in understanding
the human brain or in creating
artificial systems that have many of the
same qualities but don't necessarily
require that you actually understand the
underpinning workings of our mind so
there's a clear answer to that question
my primary interest is understanding the
human brain no question about it but I
also firmly believe that we will not be
able to create fully intelligent
machines until we understand how the
human brain works so I don't see those
as separate problems
I think there's limits so what can be
done with machine intelligence if you
don't understand the principles by which
the brain works and so I actually
believe that studying the brain is
actually the fast the fastest way to get
to machine intelligence and within that
let me ask the impossible question how
do you not define but at least think
about what it means to be intelligent so
I didn't try to answer that question
first we said let's just talk about how
the brain works let's figure out how
certain parts of the brain mostly the
new your cortex but some other parts to
the parts of the very most associated
intelligence and let's discover the
principles about how they work because
intelligence isn't just like some
mechanism and it's not just some
capabilities it's like okay we don't
even have know where to begin on this
stuff and so now that we've made a lot
of progress on this after we've made a
lot of progress on how the neocortex
works and we can talk about that I now
have a very good idea what's going to be
required to make intelligent machines I
can tell you today you know some of the
things are gonna be necessary I believe
to create intelligent machines well so
we'll get there we'll get to the
neocortex and some of the theories of
how the whole thing works and you're
saying as we understand more and more
about the neocortex about our own human
mind we'll be able to start to more
specifically define what it means to be
intelligent it's not useful to really
talk about that until I don't know if
it's not useful
look there's a long history of AI as you
know right and there's been different
approaches taken to it and who knows
maybe they're all useful right so you
know the good old fashioned AI the
expert systems current convolution
neural networks they all have their
utility
they all have a value in the world but I
would think almost everyone agree that
none of them are really intelligent in a
set of a deep way that that humans are
and so it's it's just the question is
how do you get from where those systems
were or are today to where a lot of
people think we're going to go and just
big big gap there a huge gap and I think
the quickest way of bridging that gap is
to figure out how the brain does that
and then we can sit back and look and
say oh what do these principles that the
brain works on are necessary and which
ones or not
kula we don't have to build this in and
telogen machines aren't going to be
built out of you know organic living
cells but there's a lot of stuff that
goes on the brain it's going to be
necessary so let me ask me B before we
get into the fun details
let me ask me to get depressing or a
difficult question do you think it's
possible that we will never be able to
understand how our brain works that
maybe there's aspects to the human mind
like we ourselves cannot introspectively
get to the core that there's a wall you
eventually hit yeah I don't believe
that's the case
I have never believed that's the case
there's not have been a single thing
we've ever humans have ever put their
minds to so we've said oh we reached the
wall we can't go any further it just
people keep saying that people used to
believe that about life you know Ilan's
Vittal right there's like what's the
difference in living matter and
nonliving matter something special you
never understand we no longer think that
so there's there's no historical
evidence to suggest is the case and I
just never even considered that's a
possibility I would also say today we
understand so much about the neocortex
we've made tremendous progress in the
last few years that I no longer think of
it as an open question the answers are
very clear to me and the pieces we know
we don't know I are clearly me but the
framework is all there and it's like oh
okay we're gonna be able to do this this
is not a problem anymore it just takes
time and effort but there's no mystery a
big mystery anymore so then let's get it
into it for people like myself we're not
very well versed in the human brain
except my own can you describe to me at
the highest level what are the different
parts of the human brain and then
zooming in on the neocortex the parts of
the neocortex and so on a quick overview
yeah sure human brain we can divide it
roughly into two parts there's the old
parts lots of pieces and then there's a
new part the new part is the neocortex
it's new because it didn't exist before
mammals the only mammals have a
neocortex and in humans it's in primates
it's very large in the human brain the
neocortex occupies about seventy to
seventy-five percent of the volume of
the brain it's huge and the old parts of
the brain are there's lots of pieces
there there's a spinal cord and there's
the brain stem and the cerebellum and
the different parts of the basal ganglia
and so on in the old parts of the brain
you have the autonomic regulation like
breathing and heart rate you have basic
behaviors so like walking and running or
controlled by the old parts of the brain
all the emotional centers of the brain
are in the old part of the brains when
you feel anger or hungry lust with
things like that those are all in the
old parts of the brain
and and we associate with the neocortex
all the things we think about as sort of
high-level perception and cognitive
functions anything from seeing and
hearing and touching things to language
to mathematics and engineering and
science and so on those are all
associative the neocortex and they're
certainly correlated our abilities in
those regards are correlated with the
relative size of our neocortex compared
to other mammals so that's like the
rough division and you obviously can't
understand the new your cortex is
completely isolated but you can
understand a lot of it with just a few
interfaces so the all parts of the brain
and so it it gives you a system to study
the other remarkable thing about the
neocortex compared to the old parts of
the brain is the neocortex it's
extremely uniform it's not visually or
anatomically or it's very sucky
I always like to say it's like the size
of a dinner napkin about two and a half
millimeters thick and it looks
remarkably the same everywhere
everywhere you look and that children
have millimeters is this detailed
architecture and it looks remarkably the
same everywhere and that's a cross
species the mouse versus a cat and a dog
and a human or if you look at the old
parts of the brain there's lots of
little pieces do specific things so it's
like the old parts of a brain evolved
look this is the part that controls
heart rate and this is the part that
controls this and this is this the kind
of thing and that's this kind of thing
and he's evolved for eons a long long
time and they have their specific
functions and all sudden mammals come
along and they got this thing called the
neocortex and it got large by just
replicating the same thing over and over
and over again this is like wow this is
incredible so all the evidence we have
and this is an idea that was first
articulated in a very cogent and
beautiful argument by a guy named Vernon
mal Castle in 1978 was that the
neocortex all works on the same
principle so language hearing touch
vision engineering all these things are
basically underlying or all built in the
same computational substrate they're
really all the same problem
all over the building blocks all look
similar yeah and they're not even that
low-level we're not talking about like
like neurons we're talking about this
very complex circuit that exists
throughout the neocortex is remarkably
similar it is it's like yes did you see
variations of it here and there more of
the cell uh so that's not all so on but
what now encruster argued was it says
you know if you take a section on your
cortex why is one a visual area and one
is a auditory area or why 'since and his
answer was it's because one is connected
to eyes and one is connected ears
literally you mean just its most closest
in terms of number of connections to
listen sir literally if you took the
optic nerve and it attached it to a
different part of the neocortex that
part would become a visual region this
actually this experiment was actually
done by Mercosur oh boy
and uh in in developing I think it was
lemurs I can't remember there was some
animal and and there's a lot of evidence
to this you know if you take a blind
person the person is born blind at Birth
they they're born with a visual
neocortex it doesn't may not get any
input from the eyes because of some
congenital defect or something and that
region become does something else it
picks up another task so and it's it's
so it's just it's this very complex
thing it's not like oh they're all built
on neurons no they're all built in this
very complex circuit and and somehow
that circuit underlies everything and so
this is the it it's called the common
cortical algorithm if you will some
scientists just find it hard to believe
and they decide can't really that's true
but the evidence is overwhelming in this
case and so a large part of what it
means to figure out how the brain
creates intelligence and what is
intelligence in the brain is to
understand what that circuit does if you
can figure out what that circuit does as
amazing as it is then you can then you
then you understand what all these other
cognitive functions are so a few words
to sort of put neural cortex outside of
your book on intelligence you look if
you wrote a giant tome a textbook on the
neocortex and you look maybe a couple
centuries from now how much of what we
know now would still be
two centuries from now so how close are
we in terms of understand I have to
speak from my own particular experience
here so I run a small research lab here
it's like yeah it's like I need other
research lab I'm the sort of the
principal investigator there was
actually two of us and there's a bunch
of other people and this is what we do
we started the neocortex and we
published our results and so on so about
three years ago we had a real
breakthrough in this in this film just
tremendous spectrum we started we've now
published I think three papers on it and
so I have I have a pretty good
understanding of all the pieces and what
we're missing I would say that almost
all the empirical data we've collected
about the brain which is enormous if you
don't know the neuroscience literature
it's just incredibly big and it's it's
the most part all correct its facts and
and experimental results and
measurements and all kinds of stuff but
it none none of that has been really
assimilated into a theoretical framework
it's it's data without it's in the
language of Thomas Kuhns a historian it
would be a sort of a pre paradigm
science lots of data but no way to fit
in together I think almost all of that's
correct it's gonna be some mistakes in
there and for the most part there aren't
really good cogent theories about it how
to put it together it's not like we have
two or three competing good theories
which ones are right and which ones are
wrong it's like yeah people just like
scratching their heads wrong things you
know some people given up on trying to
like figure out what the whole thing
does in fact is very very few labs that
we that we do that focus really on
theory and all this unassimilated data
and trying to explain it so it's not
like we have we've got it wrong it's
just that we haven't got it at all so
it's really I would say pretty early
days in terms of understanding the
fundamental theories forces of the way
our mind works I don't think so that
what I would have said that's true five
years ago so I
we have some really big breakthroughs on
this recently and we started publishing
papers on this so look it but so I don't
think it's I you know I'm an optimist
and from where I sit today most people
would disagree with this but from where
I sit city from what I know uh it's not
super early days anymore we are it's
it's you know the way these things go is
it's not a linear path right you don't
just start accumulating and get better
and better better no you okay all the
stuff you've collected none of it makes
sense all these different things we just
turn around and then you're gonna have
some breaking points or all sudden oh my
god now we got it right so that's how it
goes and science and I feel like we
passed that little thing about a couple
years ago all that big thing a couple
years ago so we can talk about that time
will tell if I'm right but I feel very
confident about it that's my moment to
say it on tape like this at least very
optimistic so let's before those few
years ago
let's take take step back to HTM the
hierarchical temporal memory theory
which you first proposed on intelligence
and went through a few different
generations can you describe what it is
how it evolved through the three
generations yes you first put it on
paper yeah so one of the things that
neuroscientists just sort of missed for
many many years and ice and especially
people were thinking about theory was
the nature of time in the brain brains
process information through time the
information coming into the brain is
constantly changing the patterns from my
speech right now if you're listening to
it at normal speed
we'd be changing on IRA's about every 10
milliseconds or so you'd have it change
this constant flow when you look at the
world your eyes are moving constantly
three to five times a second and the
inputs complete completely if I were to
touch something like a coffee cup as I
move my fingers that input changes so
this idea that the brain works on time
changing patterns is almost completely
or was almost completely missing from a
lot of the basic theories like fears of
vision and so it's like oh no we're
going to put this image in front of you
and flash it and say what is it a
convolutional neural networks work that
way today right you know classify this
picture but that's not what visions like
vision is this sort of crazy time-based
pattern that's going all over the place
and
was touched and so is hearing so the
first part of a hierarchal temporal
memory was the temporal part it's it's
the same you you won't understand the
brain orally understand intelligent
machines unless you're dealing with
time-based patterns the second thing was
the memory component of it was is to say
that we aren't just processing input we
learn a model of the world that's the
memory stands for that model we have to
the point of the brain part of the New
York white chest it learns a model of
the world we have to store things that
our experience is in a form that leads
to a model the world so we can move
around the world we can pick things up
and do things and navigate know how it's
going on so that's that's what the
memory referred to and many people just
they were thinking about like certain
processes without memory at all it just
like processing things and finally the
hierarchical component was reflection to
that the New York or check so though
it's just uniform sheet of cells
different parts of it project to other
parts which project to other parts and
there is this sort of rough hierarchy in
terms of them so the hyperbole temporal
memory is just saying look we should be
thinking about the brain as time-based
you know model memory based and
hierarchical processing and and that was
a placeholder for a bunch of components
that we would then plug into that we
still believe all those things I just
said but we now know so much more that
I'm stopping to use the word hierarchal
thumper memory yeah because it's it's
insufficient to capture the stuff we
know so again it's not incorrect but
it's I now know more and I would rather
describe it more accurately yeah so
you're basically we can think of HTM as
emphasizing that there's three aspects
of intelligence that important to think
about whatever the whatever the eventual
theory it converges to yeah so in terms
of time how do you think of nature of
time across different time scales so you
mentioned things changing a sensory
inputs changing every 10 being myself
what about it every few minutes every
few yeah Montse well if you think about
a neuroscience problem the brain problem
neurons themselves can stay active for
certain perks of time they parts of the
brain with this doctor 4-minute
you know so you could hold up a certain
perception or an activity for a certain
period of time but not most of them
don't last that long
and so if you think about your thoughts
are the activity neurons if you're going
to want to involve something that
happened a long time ago I'm even just
this morning for example the neurons
haven't been active throughout that time
so you have to store that so if I asked
you what did you have for breakfast
today that is memory that is you've
built into your model of the world now
you remember that and that memory is in
the in the synapses it's basically in
the formation of synapses and so it's it
you're sliding into what you know is two
different time scales there's time
scales of which we are like
understanding my language and moving
about and seeing things rapidly and over
time that's the time scales of
activities of neurons but if you want to
get longer time scales then it's more
memory and we have to invoke those
memories to say oh yes well now I can
remember what I had for breakfast
because I stored that someplace I may
forget it tomorrow but I'd stored for
afor now so this is memory also need to
have so the hierarchical aspect of
reality is not just about concepts it's
also about time do you think of it that
way yeah time is infused in everything
it's like you really can't separate it
out if I ask you what is the what is
your you know how's the brain learning a
model of this coffee cup here I have a
coffee cup and I'm at the coffee cup I
said well time is not an inherent
property of this of this of the model I
have of this cup whether it's a visual
model or attack the model I can sense it
through time but if the model self
doesn't really much time if I asked you
if I said well what is the model of my
cell phone my brain has learned a model
of the cell phone so if you have a smart
phone like this and I said well this has
time aspects to it I have expectations
when I turn it on what's gonna happen
what water how long it's going to take
to do certain things if I bring up an
app what sequences and so I have instant
it's all like melodies in the world you
know yeah melody has a sense of time so
many things in the world move and act
and there's a sense of time related to
them some don't but most things do
really so it's it's sort of infused
throughout the models of the world you
build a model of the world you're
learning the structure of the objects in
the world and you're also learning how
those things change through time okay so
it's it's it really is just a fourth
dimension that's infused deeply and they
have to make sure that your models have
been intelligence incorporated so like
you mentioned the state of neuroscience
is deeply empirical a lot of data
collection it's uh you know that's
that's where it is using meshing Thomas
Kuhn right yeah and then you're
proposing a theory of intelligence and
which is really the next step the really
important stuff to take but why why is
HTM or what we'll talk about soon
the right theory so is it more in this
it what is it backed by intuition is it
backed by evidence is it backed by a
mixture of both is it kind of closer to
or string theories in physics where this
mathematical components would show that
you know what it seems that this it fits
together too well for not to be true
which is what we're string theory is is
that where your fix of all those things
although definitely where we are right
now it's definitely much more on the
empirical side than let's say string
theory the way this goes about we're
theorists right so we look at all this
data and we're trying to come up with
some sort of model that explains it
basically and there's yeah unlike string
theory there's this vast more amounts of
empirical data here that I think than
most physicists deal with
and so our challenge is to sort through
that and figure out what kind of
constructs would explain this and when
we have an idea you come up with a
theory of some sort you have lots of
ways of testing it first of all I am you
know there are hundred years of
assimilated unassimilated empirical data
from neuroscience so we go back and read
papers we said oh did someone find this
already with you we can predict x y&z
and maybe no one's even talked about it
since 1972 or something but we go back
and find out we say Oh either it can
support the theory or it can invalidate
the theory and we said okay we have to
start over again oh no it's the poor
let's keep going with that one so the
way I kind of view it when we do our
work we come up we we look at all this
empirical data and it's it's what I call
is a set of constraints we're not
interested in something that's
biologically inspired we're trying to
figure out how the actual brain works so
every piece of empirical data is a
constraint on a theory in theory if you
have the correct theory it needs to
explain every pin right so we have this
huge number of constraints on the
problem which initially makes it very
very difficult if you don't have any
constraints you can make up stuff all
the day you know here's an answer how
you can do this you can do that you can
do this but if you consider all biology
as a set of constraints all neuroscience
instead of constraints and even if
you're working on one little part of the
neocortex for example there are hundreds
and hundreds of constraints these are
empirical constraints that it's very
very difficult initially to come up with
a radical framework for that but when
you do and it solves all those
constraints at once you have a high
confidence that you got something close
to correct it's just in mathematically
almost impossible not to be so it that's
the the curse and the advantage of what
we have the curse is we have to solve we
have to meet all these constraints which
is really hard but when you do meet them
then you have a great confidence that
you discover something in addition then
we work with scientific labs so we'll
say oh there's something we can't find
we can predict something but we can't
find it anywhere in the literature so we
will then we have people we collaborated
with
say that sometimes they'll say you know
I have some collected data which I
didn't publish but we can go back and
look in it and see if we can find that
which is much easier than designing in
your experiment you know new
neuroscience experiments take a long
time years so although some people are
doing that now too so but between all of
these things I think it's reasonable
it's actually a very very good approach
we we are blessed with the fact that we
can test our theories out the ying-yang
here because there's so much on a
similar data and we can also falsify our
theories very easily which we do often
it's kind of reminiscent to whenever
whenever that was with Copernicus you
know when you figure out that the sun's
at the center of the the solar system as
opposed to earth the pieces just fall
into place yeah I think that's the
general nature of aha moments is in
history Copernicus it could be you could
say the same thing about Darwin you
could say same thing about you know
about the double helix that that people
have been working on a problem for so
long and I have all this data and they
can't make sense of it they can't make
sense of it but when the answer comes to
you and everything falls into place it's
like oh my gosh that's it
that's got to be right I asked both Jim
Watson and Francis Crick about this I
asked him you know when you were working
on trying to discover the structure of
the double helix and when you came up
with the the sort of the structure that
ended up being correct but it was sort
of a guess you know I wasn't really
verified yeah I said did you know that
it was right and they both said
absolutely so we absolutely knew it was
right and it doesn't matter if other
people didn't believe it or not we knew
it was right they get around the thing
agree with it eventually anyway
and that's the kind of thing you hear a
lot with scientists who who really are
studying a difficult problem and I feel
that way too about our work if you talk
to Kirk or Watson about the the problem
you're trying to solve the of finding
the DNA of the brain yeah in fact
Francis Crick was very interested in
this in the latter part of his
and in fact I got interested in brains
by reading an essay he wrote in 1979
called thinking about the brain and that
is when I decided I'm gonna leave my
profession of computers and engineering
and become a neuroscientist just reading
that one essay from Francis Crick I got
to meet him later in life
I got I spoke at the Salk Institute and
he was in the audience and then I had a
tea with him afterwards you know he was
interested in a different problem and he
was he was focused on consciousness yeah
and the easy problem right well I I
think it's the red herring and and so we
weren't really overlapping a lot there
Jim Watson who's still alive is is also
interested in this problem and he was
when he was director of the coast of
Harbor laboratories he was really sort
of behind moving in the direction of
neuroscience there and so he had a
personal interest in this field and I
have met with him numerous times and in
fact the last time was a little bit over
a year ago I gave a talk close to me
Harbor labs about the progress we were
making in in our work and it was a lot
of fun because he said well you you
wouldn't be coming here unless you had
something important to say so I'm gonna
go change our talk so he sat in the very
front row next to most next to him was
the director of the lab was Stillman so
these guys are in the front row of this
auditorium right so nobody else in the
auditorium wants to sit in the front row
because Jim Watson is detective and and
I gave a talk and I had dinner with Jim
afterwards but it's I there's a great
picture of my colleague sue Battaglia
mahad took where I'm up there sort of
like screaming the basics of this new
framework we have and Jim Watson is on
the edge of his chair he's literally on
the edge of his chair like intently
staring up at the screen and when he
discovered the structure of DNA the
first public talk he gave was that Cold
Spring Harbor labs so and there's a
picture those famous picture Jim Watson
standing at the whiteboard was where the
overrated thing pointing at something
was holding a double helix at this point
it actually looks a lot like the picture
of me so there was funny I got talking
about the brain and there's Jim Watson
staring intently I didn't course there
was you know whatever sixty years
earlier he was standing you know
pointing at the double helix and it's
one of the great discoveries and and all
of you know whatever by all the science
all science yeah yeah hey so this is the
funny that there's echoes of that in
your presentation do you think in terms
of evolutionary timeline in history the
development of the neocortex was a big
leap or is it just a small step so like
if we ran the whole thing over again
from the from the birth of life on Earth
how likely develop the mechanism and you
okay well those are two separate
questions one it was it a big leap and
one was how like it is okay they're not
necessarily related maybe correlated we
don't really have enough data to make a
judgment about that I would say
definitely was a big league and leap and
I can tell you why I think I don't think
it was just another incremental step at
that moment I don't really have any idea
how likely it is if we look at evolution
we have one data point which is earth
right life formed on earth billions of
years ago whether it was introduced here
or it created it here or someone
introduced it we don't really know but
it was here early it took a long long
time to get to multicellular life and
then from multi to other started life it
took a long long time to get his
neocortex and we've only had the New
York Texas for a few hundred thousand
years so that's like nothing okay so is
it likely well certainly isn't something
that happened right away on earth and
there were multiple steps to get there
so I would say it's probably not get
something what happened instantaneous on
other planets that might have life it
might take several billion years on
average um is it likely I don't know but
you'd have to survive for several
billion years to find out probably is it
a big leap yeah I think it's it is a
qualitative difference than all other
evolutionary steps I can try to describe
that if you'd like sure you know which
way uh yeah I can tell you how pretty
much I'll start a little press
many of the things that humans are able
to do do not have obvious survival
advantages precedent yeah you know we
create music is that is there a really
survival advantage to that maybe maybe
not
what about mathematics is there a real
survival advantage to mathematics it's
stretchy you can try to figure these
things out right but up but mostly
evolutionary history everything had
immediate survival advantages too right
so I'll tell you a story which I like me
may not be true but the story goes as
follows organisms have been evolving
first since the beginning of life here
on earth anything this sort of
complexity on to that just sort of
complexity and the brain itself is
evolved this way in fact there's an old
parts and older parts and older older
parts of the brain that kind of just
keeps calling on new things and we keep
adding capabilities and we got for the
neocortex initially it had a very clear
survival advantage and that it produced
better vision and better hearing and
better thoughts and maybe a new place so
on but what what I think happens is that
evolution just kept it took it took a
mechanism and this is in our recent
theories but it took a mechanism evolved
a long time ago for navigating in the
world for knowing who you are these are
the so called grid cells and place cells
of an old part of the brain and it took
that mechanism for building maps of the
world and knowing we are in those maps
and how to navigate those maps and turns
it into a sort of a slimmed-down
idealized version of it mm-hmm and that
ideally this version could now apply to
building maps of other things maps of
coffee cups and maps the phone's maps of
these concepts yes and not just almost
exactly and and so you and it just
started replicating this stuff right you
just think more and more more bits so we
went from being sort of dedicated
purpose neural hardware to solve certain
problems that are important to survival
to a general purpose neural hardware
that could be applied to all problems
and now it's just it's
the orbit of survival it's we are now
able to apply it to things which we find
enjoyment you know but aren't really
clearly survival characteristics and
that it seems to only have happened in
humans to the large extent and so that's
what's going on where we sort of have
we've sort of escape the gravity of
evolutionary pressure in some sense in
the neocortex and it now does things
which but not that are really
interesting discovery models of the
universe which may not really help us
doesn't matter how is it help of
surviving knowing that there might be
multiple no there might be you know the
age of the universe or what how do you
know various stellar things occur it
doesn't really help us survive at all
but we enjoy it and that's what happened
or at least not in the obvious way
perhaps it is required if you look at
the entire universe in an evolutionary
way it's required for us to do
interplanetary travel and therefore
survive past our own Sun but you know
let's not get too but you know evolution
works at one time frame it's it's
survival if you think of a survival of
the phenotype survival of the individual
it is that what you're talking about
there is spans well beyond that so
there's no genetic I'm not transferring
any genetic traits to my children that
are gonna help them survive better on
Mars right it's totally different
mechanism let's yeah so let's get into
the the new as you've mentioned the idea
that I don't know if you have a nice
name thousand you call it a thousand
brain theory often told I like it so can
you talk about the this idea of spatial
view of concepts and so on yeah so can I
just describe sort of the there's an
underlying core discovery which then
everything comes from that that's a very
simple this is really what happened we
were deep into problems about
understanding how we build models of
stuff in the world and how we make
predictions about things and I was
holding a coffee cup just like this in
my hand and I had my finger was touching
the side my index finger and I moved it
to the top and I was going to feel the
the rim at the top of the cover and I
asked myself a very simple question I
said well first of all I have to say I
know that my brain predicts what its
gonna feel before it touches it you can
just think about it and imagine it and
so we know that the brain is making
predictions all the time so the question
is what does it take to predict that
right and there's a very interesting
answer that first of all it says the
brain has to know it's touching a coffee
cup and I said a model or a coffee cup
and needs to know where the finger
currently is on the cup relative to the
cup because when I make a movement and
used to know where it's going to be on
the cup after the movement is completed
relative to the cup and then it can make
a prediction about what's going to sense
so this told me that Dean your cortex
which is making this prediction needs to
know that it's sensing it's touching a
cup and it needs to know the location of
my finger relative to that cup in a
reference frame of the cup it doesn't
matter where the cup is relative my body
it doesn't matter its orientation none
of that matters it's where my finger is
relative to the cup which tells me then
that the neocortex is has a reference
frame that's anchored to the cup because
otherwise I wouldn't be able to say the
location and I wouldn't be able to
predict my new location and then we
quickly vary installation instantly you
can say well every part of my skin could
touch this cup and therefore every part
of my skin is making predictions and
every part my skin must have a reference
frame that it's using to make
predictions so the the big idea is that
throughout the neocortex
there are everything as being is being
stored and referenced in reference
frames you can think of them like XYZ
reference things but they're not like
that we know a lot about the neural
mechanisms for this but the brain thinks
in reference frames and it's an engineer
if you're an engineer this is not
surprising you'd say if I wanted to
build a a CAD model of the coffee cup
well I would bring it up in some CAD
software and I would assign some
reference frame and say this features at
this locations and so on but the fact
that this the idea that this is
occurring through out in your cortex
everywhere it was a novel idea and and
then zillion things fell into place
after that it's doing so now we think
about the neocortex as processing
information quite differently than we
used to do it we used to think about the
neural cortex is processing sensory data
and extracting features from that
sensory data and then extracting
features from the features very much
like a deep Learning Network does today
but that's not how the brain works at
all the brain works by assigning
everything every input everything to
reference frames and there are thousands
hundreds and thousands of them active at
once in your neocortex it's a surprising
thing the thing about but once you sort
of internalize this you understand that
it explains almost every all the almost
all the mysteries we've had about this
it's about this structure so one of the
consequences of that is that every small
part of the neocortex so you have a
millimeter square and there's a hundred
and fifty thousand of those so it's
about 150,000 square millimeters if you
take every little square millimeter of
the cortex it's got some input coming
into it and it's going to have reference
frames which assign that input to and
each square millimeter can learn
complete models of objects so what do I
mean by that if I'm touching the coffee
cup well if I just touch it in one place
I can't learn what this coffee cup is
because I'm just feeling one part but if
I move it around the cup it touched you
to different areas I can build up a
complete model the cup because I'm now
filling in that three dimensional map
which is the coffee cup I can say oh
what am I feeling in all these different
locations that's the basic idea it's
more complicated than that but so
through time and we talked about time
earlier through time even a single
column which is only looking at or a
single part of the cortex it's only
looking at a small part of the world can
build up a complete model of an object
and so if you think about the part of
the brain which is getting input from
all my fingers
so there's they're spread across the top
and here this is the somatosensory
cortex there's columns associated all
these from areas of my skin
and what we believe is happening is that
all of them are building models of this
cup every one of them or things not do
not all building all not every column
every part of the cortex builds models
of everything but they're all building
models of something and and so you have
it so when I when I touch this cup with
my hand there are multiple models of the
cup being invoked if I look at it with
my eyes there again many models of the
cup being invoked because each part of
the visual system and the brain doesn't
process an image that's mr. that's a
misleading idea it's just like your
fingers touching
so different parts of my Radnor of
looking at different parts of the cup
and thousands and thousands of models of
the cup are being invoked at once and
they're all voting with each other
trying to figure out what's going on so
that's why we call it the thousand
brains theory of intelligence because
there isn't one model of a cop there are
thousands of models to this Cup there
are thousands of models for your cell
phone and about cameras and microphones
and so on it's a distributed modeling
system which is very different than what
people have thought about it so this is
a really compelling and interesting idea
of f2 first questions - one on the
ensemble part of everything coming
together you have these thousand brains
how do you know which one has done the
best job of forming the great question
let me try Spain there there's a problem
that's known in neuroscience called the
sensor fusion problem yes and so is the
idea of something like oh the image
comes from the eye there's a picture on
the retina and it gets projected to than
your cortex no by now it's all spread
out all over the place and it's kind of
squirrely and distorted and pieces are
all over this you know it doesn't look
like a picture anymore
when does it all come back together
again right or you might say well yes
but I also I also have sounds or touches
associated with a couple so I'm seeing
the cup and touching the cup how do they
get combined together again so this it's
called the sensor fusion problem is if
all these disparate parts have to be
brought together into one model
someplace that's the wrong idea the
right idea is that you get all these
guys voting there's auditory models of
the cup there's visual models the cup
those tactile models of the cup there's
one the individual system there might be
ones that are more focused on black and
white ones fortunate on color it doesn't
really matter there's just thousands and
thousands of models of this Cup and they
vote they don't actually come together
in one spot it just literally think of
it this way I imagine you have these
columns or like about the size of a
little piece of spaghetti okay like a
two and a half millimeters tall and
about a millimeter in mind they're not
physical like but you could think of
them that way and each one's trying to
guess what this thing is they're
touching now they can they can do a
pretty good job if they're allowed to
move over to us so I could reach my hand
into a black box and move my finger
around an object and if I touch enough
spaces like oh okay I don't know what it
is but often we don't do that often I
can just reach and grab something with
my hand all the once and I get it or if
I had to look through the world through
a straw so long
invoking one little column I can only
see part of some things I have to move
the straw around but if I open my eyes
to see the whole thing at once so what
we think is going on it's all these
little pieces of spaghetti if you know
all these little columns in the cortex
or all trying to guess what it is that
they're sensing they'll do a better
guess if they have time and can move
over time so if I move my eyes and with
my fingers but if they don't they have a
they have a poor guest it's a it's a
probabilistic s of what they might be
touching now imagine they can post their
probability at the top of a little piece
of spaghetti each one of them says I
think and it's not really a probability
decision it's more like a set of
possibilities in the brain it doesn't
work as a probability distribution it
works is more like what we call the
Union so you could say and one column
says I think it could be a coffee cup
sort of can or a water bottle and the
other column says I think it could be a
coffee cup or you know telephone or
camera whatever right and and all these
guys are saying what they think might be
and there's these long range connections
in certain layers in the cortex so
there's been some layers in some cell
types in each column send their
projections across the brain and that's
the voting occurs and so there's a
simple associative memory mechanism
we've described this in a recent paper
and we've modeled this that says they
can all quickly settle on the only or
the one best answer for all of them if
there is a single best answer they all
vote and say yeah it's got to be the
coffee cup and at that point they all
know it's a coffee go and at that point
everyone acts as if it's the coffee cup
they yeah we know it's a coffee even
though I've only seen one little piece
of this world
I know it's coffee cup I'm touching or
I'm seeing or whatever and so you can
think of all these columns are looking
at different parts in different places
different sensory and put different
locations they're all different but this
layer that's doing the voting that's
it's solidifies it's just like it
crystallizes and says oh we all know
what we're doing and so you don't bring
these models together in one model you
just vote and there's a crystallization
of the vote great that's a at least a
compelling way to think about about the
way you form a model of the world now
you talk about a coffee cup do you see
this as far as I understand you're
proposing this as well that this extends
to much more than coffee cups
it does or at least the physical world
it expands to the world of concepts yeah
it does and well first the primary face
every evidence for that is that the
regions of the neocortex that are
associated with language or high-level
thought or mathematics or things like
that they look like the regions of the
new your cortex that process vision
hearing and touch there they don't look
any different or they look only
marginally different and so one would
say well if Vernon now Castle who
proposed it all that come all the parts
of New York or trees doing the same
thing if he's right then the parts that
during language or mathematics or
physics are working on the same
principle they must be working on the
principle of reference frames so that's
a little odd flawed hmm but of course we
had no eye
we had no prior idea how these things
happen so that's let's go with that
and we in our recent paper we talked a
little bit about that I've been working
on it more since I have better ideas
about it now I'm sitting here very
confident that that's what's happening
and I can give you some examples to help
you think about that
it's not we understand it completely but
I understand it better than I've
described it in any paper so far so but
we did put that idea out there says okay
this is it's it's it's it's a good place
to start you know and the evidence would
suggest this how it's happening and then
we can start tackling that problem one
piece at a time like what does it mean
to do high-level thought what it means a
new language how would that fit into a
reference frame framework yes so there's
a if you could tell me if there's a
connection but there's an app called
Anki that helps you remember different
concepts and they they talk about like a
memory palace that helps you remember a
completely random concepts by so trying
to put them in a physical space in your
mind yeah and putting them next to each
other the method of loci okay yeah for
some reason that seems to work really
well yeah no that's a very narrow kind
of application of just remembering some
facts but that's a very very telling one
yes exactly so it seems like you're
describing a mechanism why this seems
yeah so so basically the way what we
think is going on is all things you know
all concepts all ideas words everything
you know are stored in reference frames
and so if you want to remember something
you have to basically navigate through a
reference frame the same way a rat
navigates to a Maeve in the same way my
finger rat navigates to this coffee cup
you are moving through some space and so
what you if you have a random list of
things you were asked to remember by
assigning him to a reference frame
you've already know very well to see
your house right an idea the method of
loci is you can say okay in my lobby I'm
going to put this thing and then and
then the bedroom I put this one I go
down the hall I put this thing and then
you want to recall those facts so we
call this things you just walk mentally
you walk through your house you're
mentally moving through a reference
frame that you already had and that
tells you there's two things are really
important about it tells us the brain
prefers to store things in reference
frames and that the method of recalling
things or thinking if you will is to
move mentally through those reference
frames you could move physically through
some reference frames like I could
physically move through the reference
name of this coffee cup I can also
mentally move to the reference time the
coffee cup imagining me touching it but
I can also mentally move my house and
and so now we can ask yourself or are
all concepts toward this way there's
some recent research using human
subjects in fMRI and I'm gonna apologize
for not knowing the name of the
scientist that did this but what they
did is they they put humans in this fMRI
machine which was one of these imaging
machines and they they gave the humans
tasks to think about Birds so they had
different types of birds and beverage it
looked big and small and long necks and
long legs things like that and what they
could tell from the fMRI it was a very
clever experiment get to tell when
humans were thinking about the birds
that the birds that the knowledge of
birds was arranged in a reference frame
similar to the ones that are used when
you navigate in a room that these are
called grid cells and there are grid
cell like patterns of activity in the
new your cortex when they do this so
that it's a very clever experiment you
know and what it basically says that
even when you're thinking about
something abstract and you're not really
thinking about it as a reference frame
it tells us the brain is actually using
a reference frame and it's using the
same neural mechanisms these grid cells
are the basic same neural mechanism that
we we propose that grid cells which
in the old part of the brain the entire
cortex that that mechanism is now
similar mechanism is used throughout the
neocortex it's the same nature preserve
this interesting way of creating
reference frames and so now they have
empirical evidence that when you think
about concepts like birds that you're
using reference frames that are built on
grid cells so this that's similar to the
method of loci but in this case the
birds are related so it makes they
create their own reference frame which
is consistent with bird space and when
you think about something you go through
that you can make the same example let's
take a math mathematics all right let's
say you want to prove a conjecture ok
what is a conjecture conjecture is a
statement you believe to be true but you
haven't proven it and so it might be an
equation I I want to show that this is
equal to that and you have a place you
have some places you start with you said
well I know this is true and I know this
is true and I think that maybe to get to
the final proof I need to go through
some intermediate results but I believe
is happening is literally these
equations where these points are
assigned to a reference frame a
mathematical reference frame and when
you do mathematical operations a simple
one might be multiply or divide but you
might be a little applause transform or
something else that is like a movement
in the reference frame of the math and
so you're literally trying to discover a
path from one location to another
location in a space of mathematics and
if you can get to these intermediate
results then you know your map is pretty
good and you know you're using the right
operations much of what we think about
is solving hard problems is designing
the correct reference frame for that
problem figure out how to organize the
information and what behaviors I want to
use in that space to get me there yeah
so if you dig in an idea of this
reference frame whether it's the math
you start a set of axioms to try to get
to proving the conjecture can you try to
describe maybe taking step back how you
think of the reference frame in that
context is is it the reference frame
that the axioms are happy in is it the
reference frame that might contain
everything is that a changing thing so
there it is you
any reference frames I mean fact the way
the theory the thousand brain theory of
intelligence says that every single
thing in the world has its own reference
frame so every word has its own
reference names and we can talk about
this the mathematics work out this is no
problem for neurons to do this but how
many reference changes the coffeeCup
have well it's on a table let's say you
asked how many reference names could the
column in my finger that's touching the
coffee cup hat because there are many
many copies there many many models of a
coffee cup so the coffee
there is no walnut model the coffee cup
there are many miles of a coffee cup and
you could say well how many different
things can my finger learn missus it's
just the question you want to ask
imagine I say every concept every idea
everything you've ever know about that
you can say I know that thing it has a
reference frame associated with him and
what we do when we build composite
objects we can we sign reference frames
to point another reference frame so my
coffee cup has multiple components to it
it's got a limb it's got a cylinder it's
got a handle and those things that have
their own reference frames and they're
assigned to a master reference frame
where we just called this cup and now I
have this clementa logo on it well
that's something that exists elsewhere
in the world it's it's own thing so it
has its own reference time so we now
have to say how can I sign the new
mentor bogel reference frame onto the
cylinder or onto the coffee cup so it's
all we talked about this in the paper
that came out in December this last year
the idea of how you can assign reference
names to reference names how neurons
could do this so well my question is
okay even though you mentioned reference
frames a lot I almost feel it's really
useful to dig into how you think of what
a reference frame is I mean I was
already helpful for me to understand
sure you think of reference frames is
something there is a lot of okay so
let's just say that we're gonna have
some neurons in the brain not many
actually 10,000 20,000 are gonna create
a whole bunch of reference frames what
does it mean right what is the reference
in this case first of all these
reference names are different than the
ones you might have be used to let you
know lots of reference in its route for
example we know the Cartesian
coordinates XYZ that's a type of
reference frame we know
longitude and latitude that's a
different type of reference frame if I
look at a printed map you might have
Colin
a through a Monroe's you know one
through twenty that's a different type
of reference frame it's a kind of a
Cartesian coordinate frame though
interesting about the reference frames
in the brain and we know this because
these have been established through
neuroscience studying the anti Rana
cortex so I'm not speculating here okay
this is known neuroscience in an old
part of the brain the way these cells
create reference frames they have no
origin so what it's more like you have
you have a point your appointment in
some space and you give it a particular
movement you can then tell what the next
point should be and you can then tell
what the next point would be and so on
you can use this to to calculate how to
get from one point to another so how do
I get from being around my house to my
home or how do I get my finger from the
side of my cup to the top of the camp
how do I get from the the axioms to the
conjecture
so it's a different type of reference
frame and I can if you want I can
describe in more detail I can paint a
picture how you might want to think
about that so really helpful to think
it's something you can move through yeah
but is there is it is it helpful to
think of it as spatial in some sense or
is there something definitely spatial
its spatial in the mathematical sense we
need to mention can it be crazy numbered
well that's an interesting question in
the old part of the brain the answer I
know cortex
they studied rats and initially it looks
like oh this is just two-dimensional
it's like the rat is in some box and the
maze or whatever and they know where the
rat is using these two-dimensional
reference frames and know where it is
that's right the maze we saw okay but
what about what about bats
that's a mammal and they fly in
three-dimensional space how do they do
that they seem to know where they are
right so there's this is a current area
of active research and it seems like
somehow the rep the neurons in the in
tirana cortex I can learn
three-dimensional space we just to
members of our team along with ela FET
from MIT just released a paper this
little literally last week it's on by
archive where they show that you can if
you the way these things work and I'm
gonna get unless you want to I won't get
into the detail but grid cells can
represent any n-dimensional space
it there's no it's it's not inherently
limited you can think of it this way if
you had two-dimensional is the way it
works is you add as a bunch of
two-dimensional slices that's the way
these things work there's a whole bunch
of two-dimensional models and you can
just you can slice up any n-dimensional
space and with two-dimensional
projections so and you could all have
one dimensional models it does so
there's there's nothing inherent about
the mathematics about the way the
neurons do this which which constrain
the dimensionality of the space which I
think was important and so obviously I
have a three dimensional map of this cup
maybe it's even more than that I don't
know but it's a clearly
three-dimensional map of the cup I don't
just have a projection of the cup and
but when I think about birds or when I
think about mathematics perhaps it's
more than three dimensions or who knows
so in terms of each individual column
building up more and more information
over time do you think that mechanism is
well understood in your mind you've
proposed a lot of architectures there is
that a key piece or is it is the big
piece the thousand brain theory of
intelligence omble at all well I think
they're both big I mean clearly the
concept as a theorist the concept that's
most exciting right we've had a little
con it's a high-level concept is this a
totally new way of thinking about other
new yorker optics work so that is
appealing it has all these ramifications
and with that as a framework for how the
brain works you can make all kinds of
predictions and solve all kinds of
problems now we're trying to work
through many of these details right now
okay how do they neurons actually do
this well turns out if you think about
grid cells and place cells in the old
parts of the brain there's a lot of snow
and about them but there's still some
mysteries there's a lot of debate about
exactly the details how these work and
what are the signs and we have that
still that same level of detail the same
level concern what we spend here most of
our time doing is trying to make a very
good list of the things we don't
understand yet that's the key part here
what are the constraints it's not like
oh this thing seems work we're done no
it's like okay it kind of works but
these are other things we know what has
to do and it's not doing those yet I
would say we're well on the way here I'm
not done yet there's a lot of trickiness
to this system but the basic principles
about how different layers in the
neocortex are doing much of this we
understand but there's some fundamental
parts that we don't understand the sums
so what would you say is one of the
harder open problems or one of them ones
that have been bothering you
Oh keeping you up at night the most oh
well right now this is a detailed thing
that wouldn't apply to most people okay
yeah please we've talked about as if to
predict what you're going to sense on
this coffee cup I need to know where my
finger is gonna be on the coffee cup
that is true but it's insufficient think
about my finger touches the edge of the
coffee cup my finger can touch it at
different orientations right I can
rotate my finger around here and that
doesn't change ice I can make that
prediction and somehow so it's not just
the location there's an orientation
component of this as well this is known
in the old parts of the brain too
there's things called head Direction
cells which which way the rat is facing
it's the same kind of base
the idea so my finger were Iraq you know
in three dimensions I have a three
dimensional orientation and I have a
three dimensional location if I was a
rat I would have it you might think it
was a 2-dimensional location a two
dimensional orientation or one
dimensional orientation like just which
way is it facing so how the the two
components work together how it is that
I I combine orientation right the
orientation my sensor as well as the the
location is a tricky problem and I think
I've made progress on it though at a
bigger version of that so prospective
super interesting but super specific
yeah it's really good there's a more
general version of that do you think
context matters the fact that we are in
a building in North America that that we
in the day and age where we have mugs I
mean there's all this extra information
that you bring to the table about
everything else in the room that's
outside of just the coffee cup how does
it get yeah so Kanab you think yeah and
that is a another really interesting
question I'm gonna throw that under the
the rubric or the name of attentional
problems first of all we have this model
I have many many models so there's a and
also the question doesn't matter because
well it matters for certain things of
course it does maybe what we think of
that as a coffee cup in another part of
the world this commute is something
totally different
or maybe the our logo which is very
benign in this part of the world it
means something very different than
another part of the world so those
things do matter I think the thing the
way to think about is the following one
way to think about it is we have all
these models of the world ok and we have
modeled we model everything and as I
said earlier it comes snuck it in there
our models are actually we we build
composite structure so every object is
composed of other objects which are
composed of other objects and they
become members of other objects so this
room is chairs and a table and a room
and the walls and so on now we can just
arrange them in these things a certain
way you go that's the new meta
conference room so
so and what we do is when we go around
the world and we experience the world
we've I walk into a room for example the
first thing I'd like say oh I'm in this
room do I recognize the room then I
could say oh look there's a there's a
table here and I by attending to the
table I'm then assigning this table in a
context of the room that's on the table
there's a coffee cup oh and on the table
there's a logo and in the logo this is
the word dementia I look in the logo
there's a letter e on look it has an
unusual Seraph and it doesn't actually
but my pretend so the point is you your
attention is kind of drilling deep in
and out of these nested structures and I
can pop back up and I can pop back down
I can pop back up and I can pop back
down so I when I attend to the coffee
cup I haven't lost the context of
everything else but but it's sort of
nested structure so the attention
filters the reference frame information
for that particular period of time yes
it basically a moment-to-moment
you attend the subcomponents and then
you can tend to sub components to sub
component so you can move up and down
you can move up and down then we do that
all the time you're not even now that
I'm aware of it I'm very conscious of it
but scintilla but most people don't
don't you think about this you know you
don't you just walk in the room and you
don't say oh I looked at the chair and I
looked at the board and looked at that
word on the board and I looked over here
what's going on right so what percent of
your day are you deeply aware of this
and what part can you actually relax and
just be Jeff me personally like my
personal day yeah unfortunately I'm
afflicted with too much of the former I
[Laughter]
fortunately or unfortunately yeah I
don't think it's useful oh I did useful
totally useful I think about this stuff
almost all the time and I meant one of
my primary ways of thinking is when I'm
in sleep at night I always wake up in
the middle of the night and then I stay
awake for at least an hour with my eyes
shut in a sort of a half sleep state
thinking about these things I come up
with answers to problems very often in
that sort of half sleeping State I think
about on my bike ride I think about on
walks I'm just constantly thing about
this I have to almost a scheduled time
to not think about this stuff because
it's very it's mentally taxing
are you when you think about the
stuffy's are you thinking
introspectively like almost
gonna step outside yourself and trying
to figure out what is your mind doing
right I do that all the time but that's
not all I do I've constantly observing
myself so as soon as I started thinking
about grid cells for example and getting
into that I started saying oh well grid
cells can't mice place a sense in the
world you know that's where you know
where you are and essentially you know
we always have a sense of where we are
unless were lost and so I started at
night when I got up to go to the
bathroom I would start trying to do a
complete with my eyes closed all the
time and I would test my sense of pretty
cells I would I would walk you know five
feet and say okay I think I'm here am I
really what's my error yeah and then I
would count in my error again and see
how the errors accumulate so even
something as simple as getting up in the
middle light or the bathroom I'm testing
these theories out it's kind of fun I
mean the coffee cup is an example of
that too so I think I find that these
sort of everyday introspections are
actually quite helpful it doesn't mean
you can ignore the science I mean I
spend hours every day reading
ridiculously complex papers that's not
nearly as much fun but you have to sort
of build up those constraints and the
knowledge about the field and who's
doing what and what exactly they think
is cooperating here and then you can sit
back and say okay let's try to piece
this all together let's come up with
some you know I I'm right in this group
here people they know they just I do
this all this time I come in with these
introspective ideas and say well do you
ever thought about this now watch well
this all do this together and it's
helpful it's not if as long as you don't
be all you did was that then you're just
making up stuff right but if you're
constraining it by the reality of the
neuroscience then it's really helpful so
let's talk a little bit about deep
learning and the successes in the apply
space of neural networks the ideas of
training model and data and these simple
computational units you're on artificial
neurons that with backpropagation as the
statistical ways of being able to
generalize from the training set onto
data that similar to that training set
so where do you think are the
limitations of those approaches what do
you think our strengths relative to your
major efforts of constructing a theory
of human intelligence yeah
well I'm not an expert in this field I'm
somewhat knowledgeable so odd but I love
it is in just your intuition what are
you well I have I have a little bit more
than intuition but you're going to say
like you know one of the things that you
asked me do I spend all my time thing
about neurons I do that's to the
exclusion of thinking about things like
convolutional neural networks in you but
I try to stay current so look I think
it's great the progress they've made
it's fantastic and as I mentioned
earlier it's very highly useful for many
things the models that we have today are
actually derived from a lot of
neuroscience principles there are
distributed processing systems and
distributed memory systems and that's
how the brain works they use things that
we we might call them neurons but
they're really not neurons at all so we
can just they're not really in terrassa
distributed processing systems and and
that nature of hierarchy that came also
from neuroscience and so there's a lot
of things that the learning rules
basically not backprop but other you
know so have you int I don't know I'd be
curious to say they're not in your ons
at all he described in which way I mean
it's some of it is obvious but I'd be
curious if if you have specific ways
yeah which you think are the biggest
difference yeah we had a paper in 2016
called why neurons of thousands of
synapses and it and if you read that
paper you don't know what I'm talking
about here a real neuron in the brain is
a complex thing it let's just start with
the synapses on it which is a connection
between neurons real neurons can
everywhere from five to thirty thousand
synapses on the ones near the cell body
the ones are too close to the the soma
of the cell body those are like the ones
who people model in artificial neurons
there is a few hundred of those maybe
they can affect the cell they can make
the cell become active ninety-five
percent of the synapses can't do that
they're too far away so if you're
actually at one of those synapses it
just doesn't affect the cell body enough
to make any difference any one of them
individually anyone emanuelly or even if
you do what mass of them what what we
but what real neurons do is the
following if you activate or they you
get 10 to 20 of them active at the same
time meaning they're all receiving an
input at the same time and those 10 to
20 synapses are forty sensors within a
very short distance on the dendrite like
40 microns a very small area so if you
activate a bunch of these right next to
each other at some distant place what
happens is it creates what's called the
dendritic spike and then juridic spike
travels through the dendrites and can
reach the soma or the cell body now when
it gets there it changes the voltage
which is sort of like gonna make the
cell fire but never enough to make the
cell fire it's sort of what we call it
says we depolarize the cell you raise
the voltage a little bit but not enough
to do anything it's like well good is
that and then it goes back down again so
we proposed a theory which I'm very
confident in basics are is that what's
happening there is those ninety-five
percent of those synapses are
recognizing dozens to hundreds of unique
patterns they can write you know about
the 1020 nerve synapses at a time and
they're acting like predictions so the
neuron actually is a predictive engine
on its own it it can fire when it gets
enough what they call approximately
input from those ones near the cell fire
but it can get ready to fire from dozens
to hundreds of patterns that it
recognizes from the other guys and the
advantage of this to the neuron is that
when it actually does produce a spike in
action potential it does so slightly
sooner than it would have otherwise and
so what could is slightly sooner well
the slightly sooner part is it there's
it all the neurons in the the excitatory
neurons in the brain are surrounded by
these inhibitory neurons and they're
very fast the inhibitory neurons
it's basket cells and if I get my spike
out a little bit sooner than someone
else I inhibit all my neighbors around
me mm-hmm right and what you end up with
is a different representation you end up
with a reputation that matches your
prediction it's a it's a sparsa
representation meaning as fewest known
or interactive but it's much more
specific and so we showed how networks
of these neurons can do very
sophisticated temporal prediction
basically so so this summarize this real
neurons in the brain are time-based
prediction engines and and they and
there's no concept of this at all in
artificial what we call point neurons I
don't think you can mail the brain
without them I don't even build
intelligent
it's its theme it's where large part of
the time comes from it's it's these are
predictive models and the time is in is
there's a prior and I'm in a you know a
prediction and an action and it's
inherent to every neuron the neocortex
so so I would say that point neurons
sort of model a piece of that and not
very well with that either but you know
like for example synapses are very
unreliable and you cannot assign any
precision to them so even one digital
position is not possible so the way real
neurons work is they don't add these
they don't change these weights
accurately like artificial neural
networks do they basically form new
synapses and so what you're trying to
always do is is detect the presence of
some 10 to 20 active synapses at the
same time as opposed and they're almost
binary it's like because you can't
really represent anything much finer
than that so these are the kind of
dishes and I think that's actually
another essential component because the
brain works on sparse patterns and all
about all that mechanism is based on
sparse patterns and I don't actually
think you could build our real brains or
machine and tell us about incorporating
some of those ideas it's hard to even
think about the complex that emerges
from the fact that the timing of the
firing matters in the brain the fact
that you form new new synapses and and
the I mean everything you just mentioned
in the past okay trust me if you spend
time on it you can get your mind around
it it's not like it's no longer a
mystery to me
no but but sorry as a function in a
mathematical way it's can you get it
they're getting an intuition about what
gets it excited what not as easy as
there are many other types of neural
networks are that are more amenable to
pure analysis you know especially very
simple networks you know oh I have four
neurons and they're doing this can we
you know the scribes are mathematically
what they're doing type of thing even
the complexity of convolutional neural
networks today it's sort of a mystery
they can't really describe the whole
system and so it's different my
colleague sue Burton I am on he did a
nice paper on this you can get all the
stuff on our website if you're
interested talking about a little math
properties of sparse representations and
so we can't what we can do is we can
tell mathematically for example why 10
to 20 synapses to recognize a pattern is
the correct number it's the right number
you'd want to use and by the way that
matches biology we can show
mathematically some of these concepts
about the show why the brain is so
robust to noise and error and fallout
and so on we can show that
mathematically as well as empirically in
simulations but the system can't be
analyzed completely any complex system
can and so that's out of the realm but
there is there are mathematical benefits
and intuitions that can be derived from
mathematics and we try to do that as
well most most of our papers have the
section about that so I think it's
refreshing and useful for me to be
talking to you about deep neural
networks because your intuition
basically says that we can't achieve
anything like intelligence with
artificial neural networks well not in
their current form 9/2 can do it in the
ultimate form sure so let me dig into it
and see what your thoughts are they're a
little bit so I'm not sure if you read
this little blog post called bitter
lesson by Richard Sutton recently
recently he's a reinforcement learning
pioneer I'm not sure if you familiar
with him his basic idea is that all the
stuff we've done in AI in the past 70
years he's one of the old school guys
the the biggest lesson learned is that
all the tricky things we've done don't
you know they benefit in the short term
but in the long term what wins out is a
simple general method that just relies
on Moore's Law on on computation getting
faster and faster so this is what he's
saying this is what has worked up to now
this what has worked up to now they fear
trying to build the system if we're
talking about he's not concerned about
intelligence concern about system that
works in terms of making predictions
that applied narrow AI problems right
that's what there's the discussion is
about
that you just tried to go as general as
possible and wait years or decades for
the computation to make it actually do
you think that is a criticism or is he
saying this is the prescription of what
we ought to be doing well it's very
difficult he's saying this is what has
worked and yes a prescription with the
difficult prescription because it says
all the fun things you guys are trying
to do we are trying to do he's part of
the community they're saying it's it's
only going to be short-term gains so
this all leads up to a question I guess
on artificial neural networks and maybe
our own biological neural networks is
you think if we just scale things up
significantly so take these dumb
artificial neurons the point here as I
like that term if we just have a lot
more of them do you think some of the
elements that we see in the brain may
start emerging no I don't think so we
can do bigger problems and of the same
type I mean it's been pointed out by
many people that today's convolutional
no and that works aren't really much
different than the ones we had quite a
while ago we just they're bigger and
train more and we have more label data
and so on but I don't think you can get
to the kind of things I know the brain
can do and that we think about as
intelligence by just scaling it up so I
that maybe it's a good description of
what's happened in the past what's
happened recently with the re-emergence
of artificial neural networks it may be
a good prescription for what's going to
happen in the short term but I don't
think that's the path I've said that
earlier there's an alternate path I
should mention to you by the way that
we've made sufficient progress on our
the whole cortical theory in the last
few years that last year we decided to
start actively pursuing how do we get
these ideas embedded into machine
learning well that's it again being led
by my colleague just super talked him on
and he's more of a machine learning guy
am more of a neuroscience guy so this is
now our new is I wouldn't say our focus
but it is now an equal focus
here because we we need to proselytize
what we've learned and we need to show
how it's beneficial to - to the Machine
were earlier so we're putting we have a
plan in place right now in fact we just
did our first paper on this I can tell
you about that but you know one of the
reasons I want to talk to you is because
I'm trying to get more people in the
machine learning the community say I
need to learn about this stuff and maybe
we should just think about this a bit
more about what we've learned about the
brain and what are those team Aetna
meant - what have they done is that
useful for us yeah yeah so there is
there elements of all the the cortical
Theory the things we've been talking
about that may be useful in the short
term yes in the short term yes this is
the sorry to interrupt the the open
question is it it there it certainly
feels from my perspective that in the
long term some of the ideas we've been
talking about will be extremely useful
yeah question is whether in the short
term well this is a always that what we
I would call the entrepreneurs dilemma
so you have this long term vision oh
we're gonna all be driving electric cars
or all kind of computers or or whatever
and and you're at some point in time and
you say I can see that long-term vision
I'm sure it's gonna happen how do I get
there without killing myself you know
without going out of business right
that's the challenge that's the dilemma
it's a really difficult thing to do so
we're facing that right now so ideally
what you'd want to do is find some steps
along the way you can get there
incremental you don't have to like throw
it all out and start over again the
first thing that we've done is we focus
on the sparse representations so I just
just in case you don't know what that
means or some of the listeners don't
know what that means in the brain if I
have like 10,000 neurons what you would
see is maybe 2% of them active at a time
you don't see 50 percent you know 3 30
percent you might see 2 percent and it's
always like that for any set of sensory
input it doesn't matter anything just
about any part of the brain but which
neurons differs which neurons are active
yes I take 10,000 neurons that are
representing something though it's
sitting there in a bullet block together
it's a teeny little blocking around
10,000 there right and they're
representing a location they're
representing a cop they're representing
the input for my sensors I don't know it
doesn't matter
it's representing something the way the
representations occur it's always a
sparse representation meaning it's a
population code so which 200 cells are
active tells me what's going on it's not
individual cells on it's not important
at all it's the population code that
matters and when you have sparse
population codes then all kinds of
beautiful properties come out of them so
the brain used the sparse population
codes that we've we've written and
described these benefits in some of our
papers so they give this tremendous
robustness to the system student brains
are incredibly robust neurons are dying
all the time and spasming and synapse is
falling apart and you know that all the
time and it keeps working
so what simatai and Louise one of our
other engineers here have done I've
shown they're introducing sparseness
into accomplished neural networks and
other people thinking along these lines
but we're going about it in a more
principled way I think and we're showing
that with you enforced sparseness
throughout these convolutional neural
networks in both the active the which
sort of which neurons are active and the
connections between them that you get
some very desirable properties so one of
the current hot topics in deep learning
right now are C's adversarial examples
so you know I can give me any deep
Learning Network and I can give you a
picture that looks perfect and you're
gonna call it you know you're gonna say
the monkey is you know an airplane
that's the problem and DARPA just
announced some big thing they're trying
to you know have some contest for this
but if you if you enforce sparse
representations here many of these
problems go away they're much more
robust and they're not easy to fool so
we've already shown some of those
results it was just literally in January
or February just last month we did that
and you can I think it's on bio archive
right now or on I cry for you can read
about it but so that's like a baby step
okay that's taking something from the
brain we know we know about sparseness
we know why it's important we know what
it gives the brain so let's try to
enforce that on to this what's your
intuition why sparsity leads to
robustness because it feels like it
would be less robust so why why would
you feel the Russell bust you
so it just feels like if the fewer
neurons are involved the more fragile
that represents a there was lots of food
I said it's like 200 that's a lot is
that a lot is yes so here's an intuition
for it this is a bit technical so for
you know for engineers pyram machine
land people let's be easy but all the
listeners maybe not if you're trying to
classify something you're trying to
divide some very high dimensional space
into different pieces a and B and you're
trying to create some point where you
say all these points in this high
dimensional space are a and all these
points inside dimensional space or B and
if you have points that are close to
that line it's not very robust it works
for all the points you know about but
it's it's not very robust because you
just move a little bit and you've
crossed over the line when you have
sparse representations imagine I pick I
have I'm gonna pick 200 cells active out
of out of 10,000 okay so I have to nurse
cells active now let's say I pick
randomly another a different
representation 200 the overlap between
knows is going to be very small just a
few I can pick millions of samples
randomly of 200 ons and not one of them
will overlap more than just a few so one
way to think about is if I want them
fool one of these representations to
look like one of those other
representations I can't move just one
cell or two cells or three cells or four
cells I have to move a hundred cells and
that makes them robust in terms of
further so the you mentioned sparsity
well maybe the next thing yeah okay so
what we have we picked one we don't know
if it's going to work well yet so again
we're trying to come up incremental ways
to moving from brain theory to add
pieces to machine learning current
machine learning world and one step at a
time so the next thing we're going to
try to do is sort of incorporate some of
the ideas of the thousand brains theory
that you have many many models and that
are voting now that idea is not new
there's mixture models has been around
for a long time but the way the brain
does is a little
and and the way it votes is different
and the kind of way it represents and
certain is different so we're just
starting this work but we're going to
try to see if we can sort of incorporate
some of the principles of voting or
principles of thousand brain theory like
lots of simple models that talk to each
other in this in a very certain way and
can we build more machines the systems
that learn faster and and also well
mostly are multimodal and robust to
multimodal type of issues so the one of
the challenges there is you know the
machine learning computer vision
community has certain sets of benchmarks
sets the test would based on which they
compete and I would argue especially
from your perspective that those
benchmarks not that useful for testing
the aspects that the brain is good at or
intelligent they're not only testing in
Georgia it's a very fine yeah and it's
been extremely useful for developing
specific mathematical models but it's
not useful in the long term for creating
intelligence so yeah you think you also
have a role in proposing better tests
yeah this is a very you've identified a
very serious problem first of all the
tests that they have are the tests that
they want not the tests of the other
things that we're trying to do right you
know what are the so on the second thing
is sometimes these two could be
competitive to in these tests you have
to have huge data sets and huge
computing power instead you know and we
don't have that here
we don't have it as well as other big
teams and big companies do so there's
numerous issues there you know we come
at it you know where our approach to
this is all based on in some sense you
might argue elegance we're coming at it
from like a theoretical base that we
think oh my god this so this is a
clearly elegant this how brains work
this one told uses but the machine
learning world has gotten in this phase
where they think it doesn't matter
doesn't matter what do you think as long
as you do you know point one percent
better on this benchmark that's what
that's all that matters and
and that's a problem you know we have to
figure out how to get around that that's
that's a challenge for us that's it's
one of the challenges we have to deal
with so I agree you've identified a big
issue it's difficult for those reasons
but you know what you know part of the
reasons I'm talking to here today is I
hope I'm gonna get some machine learning
people to say read those papers those
might be some interesting ideas I'll
show you I'm trying to doing this point
one percent improvement stuff you know
well that's that's why I'm here as well
because I think machine learning now as
a community is it a place where the next
step is uh needs to be orthogonal to
what has received success in the past oh
you see other leaders saying this
machine learning and leaders you know
Geoff Hinton with his capsules idea many
people have gotten up say you know we're
gonna hit road but maybe we should look
at the brain you know things like that
so hopefully that thinking walk occur
organically and then then we're in a
nice position for people to come and
look at our work and say well welcome
you learn from these guys yeah MIT is
launching a billion-dollar computing
College the center on this idea so it's
on this idea of what uh well the idea
that you know the humanities psychology
neuroscience have to work all together
to get to ability s yeah Stanford just
did this human-centered a I said yeah
I'm a little disappointed in these
initiatives because yeah you know
they're they're fuckin is sort of a
human side of it and it could very
easily slip into how humans interact
with intelligent machine interest which
is nothing wrong with that but that's
not that is orthogonal to what we're
trying to do we're trying to say like
what is the essence of intelligence I
don't care I think I want to build
intelligent machines that aren't
emotional that don't smile at you that
you know that aren't trying to tuck you
in at night yeah there is that pattern
that you when you talk about
understanding humans is important for
understanding intelligence you start
slipping into topics of ethics or yeah
like you said the interactive elements
as opposed to no no no what's the zoom
in on the brain study say what the human
brain the baby the what's funny what a
brain dolls does and then we can decide
which parts of
we want to recreate in some system but
do you have that theory about what the
brain does what's the point you know
it's just you're gonna be wasting time
right
just to break you down on the artificial
network side maybe you could speak to
this on and that biologic and you know
aside the process of learning versus the
process of inference maybe you can
explain to me what is there a difference
between you know an artificial neural
networks there's a difference between
the learning stage and the inference
stage do you see the brain is something
different one of the one of the big
distinctions that people often say I
don't know how correct it is is
artificial neural networks need a lot of
data they're very inefficient learning
do you see that as a correct distinction
from the biology of the human brain that
the human brain is very efficient or is
that just something we deceive ourselves
no it is efficient obviously we can
learn new things almost instantly and so
what elements do you think yeah I can
talk about that you brought up two
issues there so remember I talked early
about the constraints we always feel
well one of those constraints is the
fact that brains are continually
learning that's not something we said oh
we can add that later that's something
that was upfront had to be there from
the start made our problems harder but
we showed going back to the 2016 paper
on sequence memory we showed how that
happens how the brains infer and learn
at the same time and our models do that
and they're not two separate phases or
two separate sets at the time I think
that's a big big problem in AI at least
for many applications not for all so I
can talk about that there are some that
gets detailed there are some parts of
the neocortex in the brain where
actually what's going on
there's these those ease with these
cycles uh they're like cycles of
activity in the brain and there's very
strong evidence that you're doing more
of inference on one part of the phase
and more of learning on the other part
of the phase so the brain can actually
sort of separate different populations
of cells or going back and forth like
this but in general I would say that's
an important problem we have a you know
all of our networks that we've come up
with do both and it's it they're
learning continuous learning networks
and you mentioned benchmarks earlier
well there are no benchmarks about that
exactly so so we you know we have to
like you know begin our little soapbox
say hey by the way we yeah this is
important you know and here's a
mechanism for doing that but and you
know but until you can prove it to
someone in some you know commercial
system or something's a little harder so
yeah one of the things I had to linger
on that is in some ways to learn the
concept of a coffee cup you only need
this one coffee cup and maybe some time
alone in a room with it the first things
is I when I was imagine I reach my hand
into a black box and I'm reaching I'm
trying to touch something yeah I don't
know upfront if it's something I already
know or if it's a new thing right and I
have to I'm doing both at the same time
I don't say oh let's see if it's a new
thing oh let's see if it's an old thing
I don't do that I as I go my brain says
oh it's new or it's not new and if it's
new I start learning what it is so and
it by the way it starts learning from
the get-go even if we couldn't recognize
it so they're they're not separate
problems they're in so that's the
flinger the other thing you mentioned
was the fast learning um so I was
distorting my continuous learning but
there's also fast I mean literally I can
show you this coffee cup and I say
here's a new coffee cup it's got the
logo on it take a look at it done you
done you can predict what it's going to
look like you know in different
positions so I can talk about that too
yes in the brain the way learning occurs
I mentioned this earlier but I mentioned
again the way learning occurs I'm
imagining a mass section of a dendrite
of a neuron and I want to learn I'm
gonna learn something new I'm just
doesn't matter what it is I'm just gonna
learn something new I I need to
recognize a new pattern so what I'm
gonna do I'm gonna form new synapses new
synapses we're gonna rewire the brain on
to that section of the dendrite
once I've done that everything else that
neuron has learned is not affected by it
that's because it's isolated to that
small section of the dendrite they're
not all being added together like a
point neuron so if I learn something new
on this segment here it doesn't change
anything occur anywhere else in that
neuron so I can add something without
affecting previous learning and I can do
it quickly now let's talk
we can talk about the quickness how it's
done in real neurons you might say well
doesn't it take time to form synapses
yes it can take maybe an hour to form a
new synapse we can form memories quicker
than that and I can explain that albums
too if you want but it's getting a bit
neuroscience II oh that's great but is
there an understanding of these every
level yes so from the short-term
memories in the forming uh well so this
idea synaptogenesis the growth of new
synapses that's well described as well
understood and that's an essential part
of learning that is learning that is
learning okay you know back you know the
going back many many years people you
know as what's-his-name the psychologist
who proposed heavy hem Donald Hebb he
proposed that learning was the
modification of the strength of a
connection between two neurons people
interpreted that as the modification of
the strength of a synapse he didn't say
that he just said there's a modification
between the effect of one neuron another
so synaptogenesis is totally consistent
with Donald Hebb said but anyway there's
these mechanisms that growth a new sense
you can go online you can watch a video
of a synapse growing in real time it's
literally you can see this little finger
it's pretty impressive yeah so that's
those mechanisms are known now there's
another thing that we've speculated and
we've written about which is consistent
with no neuroscience but it's less
proven and this is the idea how do i
form a memory really really quickly like
instantaneously if it takes an hour to
grow synapse like that's not
instantaneous so there are there are
types of synapses called silent synapses
they look like a synapse but they don't
do anything they're just sitting there
it's like they do a action potential
that comes in it doesn't release any
neurotransmitter some parts of the brain
have more of these and others for
example the hippocampus has a lot of
them which is where we associate most
short to remember with so what we we
speculated again in that 2016 paper we
proposed that the way we form very quick
memories very short-term memories or
quick memories is that we convert
silence and synapses into axis enough
it's going it's like seeing a synapse
there's a zero weight in a one way but
the long-term memory has to be formed by
synaptogenesis so you can remember
something really quickly by just
flipping a bunch of these guys from
silent to active it's not like it's not
from point one to point one five it's
like doesn't do anything to it releases
transmitter and if I do that over a
bunch of these I've got a very quick
short-term memory so I guess the lesson
behind this is that most neural networks
today are fully connected every neuron
connects every other nerve from layer to
layer that's not correct in the brain we
don't want that we actually don't want
that it's bad if you want a very sparse
connectivity so that any neuron connects
just some subset of the neurons in the
other layer and it does so on a on a
dendrite by dendrite segment basis so
it's a very sparse elated out type of
thing and and that then learning is not
adjusting all these ways but learning is
just saying okay connect to these 10
cells here right now in that process you
know with artificial neural networks
it's a very simple process of back
propagation that adjusts the ways the
process of synaptogenesis
synaptogenesis it's even easier it's
even easier is even easier that
propagation requires something we it
really can't happen in brains this back
propagation of this error signal it
really can't happen people are trying to
make it happen and brain fits on a
vertebrate this is this is pure heavy
and learning well synaptogenesis pure
have been learning it's basically saying
there's a population of cells over here
that are active right now and there's a
population of cells over here active
right now how do i form connections
between those active cells and it's
literally saying this guy became active
this these 100 neurons here became
active before this neuron became active
so form connections to those ones that's
it
there's no propagation of error nothing
all the networks we do all models we
have work on almost completely on heavy
and learning but in in on dendritic
segments and multiple synapses at the
same time so nonetheless I have turned
the question that you already answered
and maybe you can answer it again if you
look at the history of artificial
intelligence where do you think we stand
how far are we from solving intelligence
you said you were very optimistic
yeah can you elaborate on that yeah you
know it's just always the the crazy
question to ask because you know no one
can predict the future
absolutely so I'll tell you a story I
used to I used to run a different
Neuroscience Institute called the red
burn neuroscience tattoo and we would we
would hold these symposiums we get like
35 scientists from around the world to
come together and I used to ask him all
the same question I would say well how
long do you think it'll be before we
understand his and your cortex works and
everyone went around the room and they
had introduced the name and they have to
answer that question so I got the the
typical answer was 50 to 100 years some
people would say 500 years some people
said never
I said well your size so you know but it
doesn't work like that as I mentioned
earlier these are not these are step
functions things happen and then bingo
they happen you can't predict that I
fill I've already passed a step function
so if I can do my job correctly over the
next five years
then meaning I can proselytize these
ideas I can convince other people
they're right we can show that other
people machine learning people should
pay attention to these ideas then we're
definitely in an under 20 year time
frame if I can do those things if I'm
not successful in that and this is the
last time anyone talks to me and no one
reads our papers and you know I'm wrong
or something like that then then I don't
know but it's it's not 50 years it's it
you know it'll it'll you know the same
thing about electric cars how quickly
are they going to populate the world
which probably takes about a 20 year
span it'll be something like that but I
think if I can do what I said we're
starting it and of course there could be
other you said step functions
it could be everybody gives up on your
ideas for 20 years
and then all of a sudden somebody picks
it up again wait that guy was on to
something
yeah so that would be a that would be a
failure on my part right you know yeah
think about Charles Babbage you know
Charles Babbage invented the computer
back in the eighteen hundreds
and everyone forgot about it until you
know but he was ahead of his time I
don't think you know like as I said I
recognize this is part of any
entrepreneurs challenge I use it
entrepreneur broadly in this case I'm
not meaning like I'm building a business
trying to sell something I mean I come
trying to sell ideas and this is a
challenge as to how you get people to
pay attention to you how do you get them
to give you a positive or negative
feedback how do you get the people act
differently based on your ideas so you
know we'll see how what we do on them so
you know that there's a lot of hype
behind artificial intelligence currently
do you uh as as you look to spread the
ideas that are of neocortical theory of
the things you're working on do you
think there's some possibility we'll hit
an AI winter once again it's certainly a
possibility no don't worry about yeah
well I guess do I worry about it I
haven't decided yet if that's good or
bad for my mission that's true yeah very
true because uh it's almost like you
need the the winter to refresh the
palate yeah it's so it's like I want
here's what you want to have it is you
want like the extent that everyone is so
thrilled about the current state of
machine learning and AI and they don't
imagine they need anything else that
makes my job harder right if if
everything crashed completely and every
student left the field and there was no
money for anybody to do anything and it
became an embarrassment to talk about
machine intelligence an AI that wouldn't
be good for us either you want you want
sort of the soft landing approach right
you want enough people the senior people
in AI and machine learning say you know
we need other approaches we really need
other approaches but damn we need two
approaches maybe we should look to the
brain okay let's look the brain who's
got some brain ideas okay let's let's
start a little project on the side here
trying to do brain idea related stuff
that's the ideal outcome we would want
so I don't want a total winter and yet I
don't want it to be sunny all the time
you know so what do you think it takes
to build a system with human level
intelligence where once demonstrated you
would be very impressed so does it have
to have a body this
have to have the the the c-word we used
before consciousness as an entirety as a
holistic sense first of all I don't
think the goal is to create a machine at
his human level intelligence I think
it's a false goal it back to Turing I
think it was a false statement we want
to understand what intelligence is and
then we can build intelligent machines
of all different scales all different
capabilities you know a dog is
intelligent I don't need you know that'd
be pretty good to have a dog yeah you
know but what about something that
doesn't look like an animal at all in
different spaces so my thinking about
this is that we want to define what
intelligence says agree upon what makes
an intelligence system we can then say
ok we're now going to build systems that
work on those principles or some subset
of them and we can apply them to all
different types of problems and the the
kind of the idea it's not computing we
don't ask if I take a little you know
little one ship computer I don't say
well that's not a computer because it's
not as powerful is this you know big
server over here you know no because we
know that what the principles are
computing are and I can apply those
principles to a small problem into a big
problem
insane intelligence needs to get there
we have to say these are the principles
I can make a small one a big one I can
make them distribute it I can put them
on different sensors they don't have to
be human like at all now you did bring
up a very interesting questions about
embodiment does that have to have my
body it has to have some concept of
movement it has to be able to move
through these reference frames I talked
about earlier I whether it's physically
moving like I need if I'm going to have
a a I that understands coffee cups it's
gonna have to pick up the coffee cup and
touch it and look at it with it with its
eyes and hands or something equivalent
to that if I have a mathematical AI
maybe it needs to move through
mathematical spaces I could have a
virtual AI that lives in the internet
and it's true its movements are
traversing links and digging into files
but it's got a location that it span is
traveling through some space you can't
have an AI that just takes some flash
thing input and we call it flash
different system here's a pattern Thun
know its movement moving pattern moving
pad and moving pad attention digging
building building structure just so I
figure out the model the world so some
sort of embodiment
whether it's physical or not has to be
part of it so self-awareness in the way
to be able to answer where my bring up
self I was two different topics
self-awareness or no the very narrow
definition of self meaning knowing a
sense of self enough to know where am I
yeah the space was yeah yeah basically
the system the system needs to know its
location where each component of the
system needs to know where it is in the
world at that point in time so self
awareness and consciousness do you think
one from the perspective neuroscience
and your cortex these are interesting
topics solvable topics give any ideas of
what why the heck it is that we have a
subjective experience at all yeah I
belong is it useful or is it just a side
effect it's interesting to think about I
don't think it's useful as a means to
figure out how to build intelligent
machines it's it's something that
systems do and we can talk about what it
is that are like well I build the system
like this then it would be self-aware or
and if I build it like this it wouldn't
be self-aware so that's a choice I can
have it's not like oh my god itself away
I can't turn oh I I heard interview
recently with this philosopher from Yale
I can't remember his name apologize for
that but he was talking about well if
these computers are self-aware then it
would be a crime to unplug them I'm like
oh come on you know I employed myself
every night go to sleep what is that a
crime you know I plugged myself in again
in the morning I am so people get kind
of bent out of shape about this I have
very different very detailed
understanding or opinions about what it
means to be conscious and what it means
to be self-aware I don't think it's that
interesting a problem you've talked
about Christoph caulk you know he thinks
that's the only problem I didn't
actually listen to your interview with
him but I know him and I know that's the
thing he also thinks intelligence the
cautions are disjoint so I mean it's not
I don't have to have one or the other so
he is I just agree with that I just
totally agree with that
so where's hear your thoughts the
cautions were doesn't emerge from
because it is so we then we have to
break it down to the two parts okay
because consciousness isn't one thing
that's part of the problem that term is
it means different things to different
people
and there's different components of it
there is a concept of self-awareness
okay that it can be very easily
explained you have a model of your own
body
the your cortex models the things in the
world and it also models your own body
and and then it has a memory it can
remember what you've done okay so it can
remember what you did this morning can
remember what you had for breakfast and
so on and so I can say to you okay Lex
were you conscious this morning when you
know I had your you know bagel and you'd
say yes I was conscious now what if I
could take your brain and revert all the
synapses back to the state they were
this morning and then I said to you Lex
were you conscious when you ate the
bagel you should know and I wasn't hot
just actually here's a video of eating
the bagel he's saying I wasn't there I
have no I that's not possible because I
was I must have been unconscious at that
time so we can just make this one-to-one
correlation between memory of your
body's trajectories through the world
over some period of time a memory that
and the ability to recall that memory is
what you would call conscious I was
conscious of that it's a self awareness
um and and in any system that can recall
memorize what it's done recently and
bring that back and invoke it again
would say yeah I'm aware I remember what
I did yeah all right I got it
that's an easy one although some people
think that's a hard one the more
challenging part of consciousness is
this one that sometimes you just go by
the word of quality um which is you know
why does an object seem red or what is
pain and why just pain feel like
something why do I feel redness so what
do I feel a little pain is in no way and
then I could say well why does sight
seems different than just hearing you
know it's the same problem it's really
yeah these are all dis neurons and so
how is it that why does looking at you
feel different than you know I'm hearing
you it feels different but this is noise
in my head they're all doing the same
thing so that's the interesting question
the best treatise I've read about this
is by guy named Oh Reagan
or Regan he wrote a book called why red
doesn't sound like a bill it's a little
it's not it's not a trade book easy read
but it and and it's an interesting
question take something like color
color really doesn't exist in the world
it's not a property of the world
property the world that exists is light
frequency and that gets turned into we
have certain cells in the retina that
respond to different frequencies
different than others and so when they
enter the brain you have a bunch of
axons that are firing at different rates
and from that we perceive color but
there is no color in the brain I mean
there's there's no color coming in on
those synapses it's just a correlation
between some some some axons and some
property of frequency and that isn't
even color itself frequency doesn't have
a color it's just a it's just what it is
so then the question is well why does it
even appear to have a color at all just
as you're describing it there seems to
be a connection of these those ideas of
reference frames I mean it just feels
like consciousness having the subject
assigning the feeling of red to the
actual color or to the wavelength it's
useful for intelligent that's a good way
putting it it's useful as a predictive
mechanism or useful there's a
generalization I did it's a way of
grouping things together to say it's
useful to have a model like this yeah
think about the the the there's a
well-known syndrome that people who've
lost a limb experience called phantom
limbs and what they claim is they can
have their arm is removed but they feel
their arm that not only feel it they
know it's there they it's there I can I
know it's there they'll swear to you
that it's there and then they can feel
pain in the arm and feeling their finger
in it they move their they move their
non-existent arm behind your back then
they feel the pain behind their back so
this whole idea that your arm exists is
a model of your brain it may or may not
really exist and just like but it's
useful to have a model of something that
sort of correlates to things in the
world so you can make predictions about
what would happen when those things
occur it's a little bit of a fuzzy but I
think you're getting quite towards the
answer there it's it's useful for the
model of to express things certain ways
that we can then map them into these
reference frames and make predictions
about them I need to spend more time on
this topic it doesn't bother me do you
really need to spend more time yeah
yeah it does feel special that we have
subjective experience but I'm yet to
know why I'm just I'm just personally
curious it's not it's not necessary for
the work we're doing here I don't think
I need to solve that problem to build
intelligent machines at all not at all
but there is so the the silly notion
that you described briefly that doesn't
seem so silly does humans is you know if
you're successful building intelligent
machines it feels wrong to then turn
them off
because if you're able to build a lot of
them it feels wrong to then be able to
you know to turn off the Y but just be
let's let's break it down a bit as
humans why do we fear death there's
there's two reasons we fear death well
first of all stay when you're dead
doesn't matter oh okay
you're doing it so why do we fear death
we fear death for two reasons one is
because we are programmed genetically to
fear death that's a that's a survival
and propagating the genes thing and we
also a program to feel sad when people
we know die
we don't feel sad for someone we don't
know dies it's people dying right now
they're always come saying I'm so bad
about because I don't know them but I
knew them I'd feel really bad so again
this these are old brain genetically
embedded things that we fear death
there's outside of those those
uncomfortable feelings there's nothing
else to worry about wait a second do you
know the denial of death by becquer I
don't know you know there's a thought
that death is you know our whole
conception of our world model kind of
assumes immortality and then death is
this terror that underlies it all so
like well some people's world not mine
but okay so what what Becker would say
is that you're just living an illusion
you've constructed an illusion for
yourself because it's such a terrible
terror the fact that what is the
illusion that deathless about you still
not coming to grips with the delusion of
what that death is are going to happen
it's not going to happen
you're a mess you're actually operating
you haven't even though you said you've
accepted it you haven't really
except in Russia guys what do you say so
it sounds like it sounds like you
disagree with that notion every night I
go to bed it's like dying a little
deaths and if I didn't wake up it
wouldn't matter to me only if I knew
that was gonna happen would it be
bothers him if I didn't know was gonna
happen how would I know know it then I
would worry about my wife
so imagine imagine I was a loner and I
lived in Alaska and and I lived them out
there and there's no animals nobody knew
I existed I was just eating these roots
all the time and nobody knew was there
and one day I didn't wake up where what
what pain in the world would there exist
well so most people that think about
this problem would say that you're just
deeply enlightened or are completely
delusional but I would say I would say
that's a very light enlightened way to
see the world is that that's the
rational rational that's right but the
fact is we don't I mean we really don't
have an understanding of why the heck it
is were born and why we die and what
happens after well maybe there isn't a
reason maybe there is so I mentioned
those big problems too right you know
you you interviewed max tegmark you know
and there's people like that right I'm
missing those big problems as well and
in fact when I was young I made a list
of the biggest problems I could think of
first why is anything exists second why
did we have the laws of physics that we
have third is life inevitable and why is
it here fourth is intelligence
inevitable and why is it here I stopped
there because I figured if you can make
a truly intelligent system will be that
will be the quickest way to answer the
first three questions I'm serious yeah
and and so I said my mission I mean I
you asked me earlier my first missions
understand the brain but I felt that is
the shortest way to get to true machine
intelligence and I want to get the true
machine tells us because even if it
doesn't occur in my lifetime other
people will benefit from it because I
think it'll occur in my lifetime but you
know 20 years it's you never know and
but that would be the quickest way for
us to you know we can make super
mathematicians we can make soup
space explorers we can make super
physicists brains that do these things
and that can run experiments that we
can't run we don't have the abilities to
manipulate things and so on but we can
build intelligent machines that do all
those things and with the ultimate goal
of finding out the answers to the other
questions let me ask you know the
depressing and difficult question which
is once we achieved that goal do you of
creating it over know of understanding
intelligence do you think we would be
happier more fulfilled as a species
the understand intelligent understanding
the answers to the big questions
understanding intelligence Oh totally
totally for more fun place to live you
think so oh yeah I mean beside this you
know terminator nonsense and and and and
just think about you can think about we
can talk about the risk of AI if you
want I'd love to so let's uh I think
world's before better knowing things
we're always better than no things do
you think it's better better place to
work the living that I know that our
planet is one of many in the solar
system and the soleus is one of many of
the calluses I think it's a more I I
dread I used to I sometimes think like
God what would be like the list three
hundred years ago I'd be looking at the
sky god I can't understand anything oh
my god I'd be like throwing a bit of
light going what's going on here well I
mean in some sense I agree with you but
I'm not exactly sure that I'm also a
scientist so I have I share you've used
but I'm not we're like rolling down the
hill together oh oh what's down the hill
I feel for climbing a hill whatever
anything cooler getting closer to
enlightenment
we're climbing we're getting pulled up a
hill the way you're putting our polio
studies put we're pulling ourselves up
the hill by our curiosity
yeah Sisyphus is doing the same thing
with the rock yeah yeah
but okay our happiness decide do you
have concerns about you know you talk
about sam harris you know a musk of
existential threats of
intelligence no I'm not worried about
exercise there are there are some things
we really do need to worry about even
today's things we have to worry about we
have to worry about privacy and about
how impacts false beliefs in the world
and and we have real problems that and
things to worry about with today's AI
and that will continue as we create more
intelligent systems there's no question
you know the whole issue about you know
making intelligent armament and weapons
it's something that really we have to
think about carefully I don't think of
those as existential threats I think
those are the kind of threats we always
face and we all have to face them here
and hope to deal with them
the ie we can we could talk about what
people think are the existential threats
but when I hear people talking about
them they all sound hollow to me they're
based on ideas they're based on people
who really have no idea what
intelligence is and and if they knew
what intelligence was they wouldn't say
those things so those are not experts in
the field in at home so yeah so there's
two right there's so one is like super
intelligence so a system that becomes
far far superior in reasoning ability
than us humans how is that an
existential threat then so there's a lot
of ways in which it could be one way as
us humans are actually irrational
inefficient and get in the way of of not
happiness but whatever the objective
function is of maximizing that objective
function yeah super intelligent
paperclip problem things like but so the
paperclip problem but with a super
intelligent yeah so we already face this
threat in some sense
they're called bacteria these are
organisms in the world that would like
to turn everything into bacteria and
they're constantly morphing they're
constantly changing to evade our
protections and in the past they have
killed huge swathes of populations of
humans on this planet so if you want to
worry about something that's going to
multiply endlessly we have it and I'm
far more worried in that regard I'm far
more worried that some scientists in a
laboratory will create a super virus or
a super bacteria that we cannot control
that is a more existential strep putting
putting in its halogen thing on top of
it actually seems to make it less
existential to me it's like it's it
limits its power is limits where it can
go and limits the number of things that
can do in many ways a bacteria is
something you can't you can't even see
so that's the only one of those problems
yes exactly so the the other one just in
your intuition about intelligent you
think about intelligence as humans do
you think of that as something if you
look at intelligence on a spectrum from
zero to us humans do you think you can
scale that to something far superior
yeah all the mechanisms with me I want
to make another point here that Lex
before I get there sure intelligence is
the neocortex it is not the entire brain
if I the goal is not to be make a human
the goal is not to make an emotional
system the goal is not to make a system
that wants to have sex and reproduce
why would I build that if I want to have
a system that wants to reproduce enough
sex make bacteria make computer viruses
those are bad things don't do that just
those are really bad don't do those
things regulate those but if I just say
I want to intelligent system why does it
have to have any human like emotions why
couldn't I does he even care if it lives
why does it even care if it has food it
doesn't care about those things it's
just you know it's just in a trance
thinking about mathematics or it's out
there just trying to build the space
plant you know for it on Mars it's C we
don't that's a choice we make don't make
human-like things don't make replicating
things don't make things which have
emotions just stick to the neocortex so
that's that's a view actually that I
shared but not everybody shares in the
sense that you have faith and optimism
about us
years of systems humans as builders of
systems got to to do not put in stupid
not so this is why I mentioned the
bacteria one yeah because you might say
well some person's gonna do that well
some person today could create a
bacteria that's resistant to all the non
antibacterial agents so we already have
that threat we already knows this is
going on it's not a new threat so just
accept that and then we have to deal
with it right yeah so my point has
nothing to do with intelligence it
intelligence is the separate component
that you might apply to a system that
wants to reproduce and do stupid things
let's not do that and in fact it is a
mystery why people haven't done that
yeah my my dad is a physicist believes
that the reason you so for some nuclear
weapons haven't proliferated amongst
evil people so one is one belief that I
share is that there's not that many evil
people in the world that would that that
would use Spectre whether it's bacteria
and you clear weapons or maybe the
future AI systems to do bad so the
fraction is small and the second is that
it's actually really hard technically
yeah so the the intersection between
evil and competent is small in terms and
otherwise it really annihilate humanity
you'd have to have you know sort of the
the nuclear winter phenomena which is
not one person shooting you know or even
ten bombs you'd have to have some
automated system that you know detonates
a million bombs or whatever many
thousands we have extreme evil combined
with extreme competence and it's just
like only some stupid system that would
automatically you know dr. Strangelove
type of thing you know I mean look we
could have some nuclear bomb go off in
some major city in the world like no I
think that's actually quite likely even
in my lifetime I don't think that's on I
like to think and it'd be a tragedy but
it won't be an existential threat and
it's the same as you know the virus of
1917 whatever it was you know the
influenza these bad things can happen
and the plague and so on we can't always
prevent them we always to always try but
we can't but they're not existential
threats until we combine all those crazy
things together one
so on the on the spectrum of
intelligence from zero to human do you
have a sense of if whether it's possible
to create several orders of magnitude or
at least double that of human
intelligence type on your cortex
I think the wrong thing to say double
the intelligence you break it down into
different components can I make
something that's a million times faster
than a human brain
yes I can do that could I make something
that is has a lot more storage than the
human brain yes I could more common more
copies of comp can I make something that
attaches the different sensors than
human brain yes I can do that could I
make something that's distributed so
these people yet we talked early about
that important in your cortex voting's
well they don't have to be co-located
why you know they can be all around the
places I could do that too those are the
levers I have but is it more intelligent
what depends what I train it on what is
it doing if it's oh here's the thing so
let's say larger neocortex and or
whatever size that allows for higher and
higher hierarchies yeah to form right
we're talking about rains in canto I
could could I have something as a super
physicist or a super mathematician yes
and the question is once you have a
super physicist will they be able to
understand something do a sense that
it'll be orders to make like us compared
to ever understand it yeah most people
cannot understand general relativity
right it's a really hard thing together
I mean paint in a fuzzy picture stretchy
space you know yeah but the the field
equations to do that in the deep
intuitions are really really hard and
I've tried I unable to do it is to get
you know it's easy to get special
relativity general that's it man that's
too much and so we already live with
this to some extent the vast majority of
people can't understand actually what
the vast majority other people actually
know we're just either we don't have the
effort to or we can't or it on time are
just not smart enough whatever so but we
have ways of communicating
Einstein has spoken in a way that I can
understand he's given me analogies that
are useful
I can use those analogies from my own
work and think about you know concepts
that are similar it's not stupid it's
not like he's existed some other plane
there's no connection to my plane in the
world here so that will occur it already
has occurred that's from my point that
this story is it already has a kirby
liveth everyday one could argue that
with me crepe machine intelligence that
think a million times faster than us
that it'll be so far we can't make the
connections but you know at the moment
everything that seems really really hard
to figure out in the world when you
actually figure it out it's not that
hard you know we can everyone most
everyone can understand the multiverses
and most everyone can understand quantum
physics we can understand these basic
things even though hardly any baby
people could figure those things out
yeah but really understand so only a few
people really understand you need to
only understand the the projections the
sprinkles of the useful my example of
Einstein right his general theory of
relativity is one thing that very very
very few people can get and what if we
just said those other few people are
also artificial intelligences how bad is
that in some sense they right yeah they
say already you mean Einstein wasn't a
very normal person he had a lot of where
the quirks and so the other people who
work with him so you know maybe they
already were sort of this astral plane
of intelligence that we live with it
already it's not a problem it's still
useful and you know so do you think we
are the only intelligent life out there
in the universe I would say that
intelligent life has and will exist
elsewhere in the universe I'll say that
there is a question about
contemporaneous intelligence life which
is hard to even answer when we think
about relativity in the the nature of
space-time you can't say what exactly is
this time someplace else in the world
but I think it's it's you know I do
worry a lot about the the filter idea
which is that perhaps intelligent
species don't last very long
and so we haven't been around very long
you know as a technological species
we've been around for almost nothing man
you know what 200 years I'm like that
and we don't have any data a good data
point on whether it's likely they will
survive or not so do I think that there
have been intelligent life elsewhere in
the universe almost certain that of
course in the past in the future yes
does it survive for a long time I don't
know
this is another reason I'm excited about
our work is our work meaning that
general Worlds of AI and I think we can
build intelligent machines that outlast
us and you know they don't have to be
tied to earth they don't have to you
know I'm not saying that recreating you
know you know aliens I'm just saying
well if I asked myself and this might be
a good point to end on here if I asked
myself you know what's special about our
species we're not particularly
interesting physically we're not we
don't fly we're not good swimmers we're
not very fast from that very strong you
know it's our brain that's the only
thing and we are the only species on
this planet it's built the model of the
world it extends beyond what we can
actually sense we're the only people who
know about the far side of the Moon and
the other universes and I mean other
other galaxies and other stars and and
but what happens in the atom there's no
what that knowledge doesn't exist
anywhere else it's only in our heads
cats don't do it dogs into a monkey's
don't do it it's just on and that is
what we've created that's unique not our
genes it's knowledge and if I asked me
what is the legacy of humanity what what
what should our legacy be it should be
knowledge we should preserve our
knowledge in a way that it can exist
beyond us and I think the best way of
doing that in fact you have to do it is
to has to go along with intelligent
machines to understand that knowledge
it's a very broad idea but we should be
thinking I call it a state planning for
Humanity we should be thinking about
what we want to leave behind when as a
species we're no longer here and that'll
happen sometime it sooner or later it's
gonna happen and understanding
intelligence and creating intelligence
gives us a better chance to prolong it
does give us a better chance prolonging
life yes it gives us a chance to live on
other planets but even beyond that I
mean our solar system will disappear
one day just give enough time so I don't
know I thought we'll ever be able to
travel to other things but we could tell
the stars but we could send Intel's
machines to do that say you have a you
have an optimistic a hopeful view of our
knowledge of the echoes of human
civilization living through the
intelligence systems we create Oh
totally well I think the telephone
systems are created in some sense the
the vessel for bring him beyond Earth or
making him last beyond humans themselves
so how do you feel about that that they
won't be human quote-unquote human what
does human our species are changing all
the time human today is not the same as
human just fifty years ago its what does
human do we care about our genetics why
is that important as I point out our
genetics are no more interesting than
about two Miam genetics there's no more
interesting them you know monkeys
genetics what we have what what's unique
and what's family better I start is our
knowledge art what we've learned about
the world and that is the rare thing
that's the thing we want to preserve its
genes the knowledge the knowledge that's
a really good place to end thank you so
much for talking
you