Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
Stay farmin' DJ Dramos from Life as
0:02
a Gringo. No making smarter financial moves
0:04
today. Secures your financial freedom for a
0:06
successful tomorrow. Tackle
0:09
these situations in stride and of course be
0:11
annoyed when unplanned expense comes up, but not
0:14
let it be something that slows me down,
0:16
right? As I did with repairing my credit,
0:18
you know, hiring somebody to do credit repair
0:20
for me. That was a gift that I
0:22
gave myself that allowed me to then, you
0:24
know, get my first apartment. Like
0:29
a good neighbor, Stay Farm is there. Stay
0:31
Farm. Proud sponsor of Michael Tura Podcast Network.
0:33
Plug in a Hyundai EV and the extraordinary
0:35
happens. From the charge time and range in
0:37
the IONIQ 5 and 6, to the adventurous
0:39
spirit of the Kona Electric to the 601
0:41
horsepower IONIQ 5N, Hyundai
0:45
EVs make the extraordinary electrifying.
0:47
There's joy in every journey. EPA estimated
0:49
303 mile driving range for 2024 IONIQ
0:52
5SEC SEL Limited real wheel drive and
0:55
361 mile driving range for 2024 IONIQ 6SEC Long
0:57
Range Rear Wheel Drive with fully charged battery. Actual
0:59
range may vary. Visit hundayusa.com or call 562-314-4603 for
1:01
more details. You're
1:05
ready for a comeback. You're ready for a comeback. and
1:07
with Purdue Global, you can do more than
1:09
take classes. you can take charge
1:11
of your story, of your
1:13
career, of your life. journal
1:16
degree you can be proud of and
1:18
get an an education employers respect. It's
1:20
time, your time, not
1:23
just to go back to school, but to come
1:25
back and move forward with
1:27
Purdue Global. Purdue's online university
1:29
for working adults. Start
1:31
your comeback at PurdueGlobal.edu
1:35
Are your lingering depression symptoms getting in
1:37
your way? If you're on an antidepressant
1:39
for major depressive disorder, MDD, and you've
1:41
made some progress but your unresolved symptoms
1:44
leave you feeling stuck, it
1:46
may be time to ask your
1:48
doctor about Vraylar. Cariprazine. Vraylar is
1:50
a prescription medicine approved for use
1:52
with antidepressant medicines to treat MDD
1:54
in adults. Adding Vraylar to an
1:56
antidepressant is clinically proven to help
1:58
relieve overall depression. symptoms better than
2:01
an antidepressant alone. Results may vary. Build
2:03
on the progress you've already made. Ask
2:05
your doctor if adding on Vraylar to
2:07
your current antidepressant could help give you
2:09
a lift in relief. Vraylar
2:11
is not approved in elderly patients with
2:13
dementia-related psychosis or for people under 18.
2:16
Elderly dementia patients have increased risk
2:18
of death or stroke. Report unusual
2:20
changes in behavior or suicidal thoughts.
2:23
Antidepressants can increase these in children and
2:25
young adults. Report fever, stiff
2:27
muscles, or confusion as these may be
2:29
life-threatening or uncontrolled muscle movements which may
2:31
be permanent. High blood sugar which can
2:34
lead to coma or death, weight gain,
2:36
and high cholesterol may occur. Difficulty
2:38
moving, tremors, slow or uncontrolled
2:40
body movements, restlessness, feeling like
2:42
you need to move around,
2:44
nausea, constipation, insomnia, dizziness, increased
2:46
appetite, and fatigue are common side effects.
2:49
Side effects may not appear for several
2:51
weeks. For a lift
2:53
in relief, ask about adding Vraylar. V-R-A-Y-L-A-R. Visit
2:56
braylar.com or call 1-877-6-BRAYLAR to learn more. What
3:07
is special about the wrinkly outer layer
3:09
of the brain, the cortex? And what
3:11
does this have to do with the
3:13
way that you come to explore and
3:16
understand the world? And by the way,
3:18
why do you see a whole image
3:20
when you open your eyes even though
3:22
each part of your visual cortex has
3:25
access to only a tiny bit of
3:27
the image? And for that
3:29
matter, the brain is divided into different areas
3:31
for sight and sound and touch and so
3:33
on. And so why when you
3:36
are petting a cat, why does the
3:38
cat seem unified? Why doesn't
3:40
the sight of the cat seem
3:42
separate from the purring and the
3:44
feel of the fur? Can
3:47
we build a new model of how the
3:49
brain works? And in what
3:51
ways is what the brain doing
3:53
something very different than what's happening
3:55
in current AI? of
6:01
our cortex. We humans have a ton
6:03
of this stuff. So take
6:06
four pieces of paper from your printer
6:09
and place them next to each other
6:11
to make one really large piece. That's
6:13
how much cortex a human has if
6:15
you were to spread out the wrinkles.
6:18
Now our nearest cousins, the great apes,
6:20
only have about one piece of paper
6:22
worth and most mammals have a lot
6:24
less than that. So something about
6:26
the story of the runaway human
6:28
success has to do with the
6:31
fact that we have way more
6:33
cortex for our body size than
6:35
any other creature. And
6:37
side note, I'm really talking about
6:40
what's called the neocortex or new
6:42
cortex because we also have a
6:44
little bit of paleocortex or old
6:46
cortex, but the thing that really
6:48
makes us outstanding is the amount
6:50
of neocortex that we have. But
6:53
what is this neocortex doing?
6:56
Well, if you look at any neuroscience
6:58
textbook, you'll see that this part of
7:01
the brain, the cortex is often drawn
7:03
with different colored regions like this red
7:05
region over here is devoted to vision
7:07
and this green one is devoted to
7:10
hearing and this yellow one to touch
7:12
and so on. But something I've been
7:14
obsessed with and write about in my
7:17
latest book, LiveWired, is that this is
7:19
the wrong way to think about it
7:21
because the neocortex is remarkably flexible. It's
7:23
not a fixed map. If
7:25
you are born blind, the part of
7:27
your cortex that we would have thought
7:30
of as visual cortex gets taken over
7:32
by hearing and touch and so on.
7:35
Now let me just be really clear what
7:37
I mean by taken over. The neurons there
7:39
are the same. The cortex
7:41
looks exactly the same from the outside,
7:43
but the function of those particular neurons
7:45
is now not visual. They
7:48
have nothing to do with visual information
7:50
anymore. Now that same neuron,
7:52
instead of firing when
7:54
it detects a moving object,
7:56
now it responds to a
7:58
touch your toe or
8:01
hearing a B flat note or
8:03
whatever. So the little
8:05
labels that we draw onto the brain,
8:08
these maps that we impose, these
8:10
are actually massively flexible. And as you
8:13
may know, I gave a talk at
8:15
Ted about this a while ago, where
8:17
I showed that you can feed in
8:19
new kinds of information, let's say through
8:21
the ears or the skin, and the
8:24
brain will figure out how to deal
8:26
with that data. It will flexibly devote
8:28
part of its cortical real estate to
8:30
that. And
8:32
this line of thinking led some scientists
8:35
like Vernon Mount Castle some decades ago
8:37
to realize that the cells of the
8:40
cortex are a one trick
8:42
pony. No neuron is
8:44
inherently a visual neuron
8:47
or a neuron devoted to hearing
8:49
or touch or smell or taste
8:51
or memory or whatever. All
8:53
parts of the cortex are perfectly
8:55
capable and willing to take on
8:57
any job. So that
9:00
suggests they're all running some sort of
9:02
basic algorithm. And it doesn't matter what
9:04
kind of data you feed in different
9:06
parts of the cortex. We'll say, cool,
9:09
I'll build a representation of that data.
9:11
I don't care if it comes from
9:13
photons or air compression waves or temperature
9:15
or whatever. I'm
9:17
on the job here to build
9:20
an understanding of whatever is coming
9:22
in locally. Now,
9:24
it's not individual neurons that are
9:26
building models, but instead groups
9:29
of many tens of thousands of
9:31
neurons arranged in a
9:33
six layered cylinder. So think about
9:35
this like you're a geologist
9:37
and you drilled out a cylinder of
9:40
rock and you saw
9:42
six layers in it, six sedimentary
9:44
layers. That's what the neocortex looks
9:46
like six layers. And it's built
9:49
out of these columns, which have
9:51
the same types of neurons with
9:53
the same connection patterns in each
9:55
column. And so think
9:57
about the cortex as being made. of
10:00
lots of these columns, like taking hundreds
10:02
of thousands of grains of rice and standing them
10:04
up on their end and packing them all next
10:07
to each other. People
10:09
have known about cortical columns for many
10:11
decades since Vernon Mount Castle first discovered
10:13
these in 1957. But
10:16
recently, someone has pulled together several
10:19
different threads to propose
10:21
how this could underlie what
10:23
the cortex is all about.
10:26
And that someone is Jeff Hawkins and
10:28
his team. And so I
10:30
met with Jeff in my studio. Now,
10:32
Jeff is one of my favorite people
10:34
because he does theoretical neuroscience. He really
10:36
tries to figure out the big picture
10:38
of what the brain is
10:41
doing. Now, Jeff has a very interesting
10:43
history. So I'll just mention that in
10:45
the 1980s, he was a graduate student
10:47
at Berkeley where he proposed a PhD
10:49
thesis on a new theory of the
10:52
cortex. But his proposal was rejected.
10:54
And so he ended up pursuing
10:57
his vision for mobile computing instead.
10:59
And in 1992, he launched the
11:01
company Palm, which made the Palm
11:04
pilot. If you remember that, this
11:06
was this little handheld device and
11:08
you could write on it with
11:10
a stylus and it would translate
11:12
your handwriting into text. And
11:14
you can use this for your address book and your
11:17
calendar and your contacts and note taking. This
11:19
was the first entrant into the world
11:21
of portable computing. It really changed the
11:23
world. Anyhow, a decade
11:26
later, Jeff returned to his original
11:28
love, which was theoretical neuroscience, trying
11:30
to figure out what's going
11:32
on with the brain. And
11:35
he wrote a book in 2004 called
11:37
on intelligence, which was very influential
11:39
on me and lots of other
11:41
thinkers I know. So I was
11:43
very excited when Jeff recently came
11:45
out with his next book that
11:47
represents his last decade and a
11:49
half of research. It's called a
11:51
thousand brains, a new theory of
11:53
intelligence, and it describes his framework
11:55
for thinking about the brain. So
11:58
without further ado, let's dive into.
12:00
a very cool new model of
12:02
the brain. Okay,
12:07
Jeff, so you are a theoretician. You think about
12:10
the brain from a high level. We're in
12:12
this era now of AI, where
12:14
AI is doing all kinds of things that
12:16
are amazing and no one expected, but you
12:18
see the brain as being very different from
12:20
what is going on with, let's say, large
12:22
language models. So tell us about that. That's
12:24
absolutely true. You know,
12:27
the current AI wave is really amazing, but
12:29
those models don't work at all like the
12:31
brain. And I think you
12:33
could start with one really fundamental difference.
12:36
Brains work through movement. We
12:39
move our bodies through the world. We move our
12:41
hands over objects to touch and learn what they
12:43
are. We move our eyes constantly. So
12:45
the inputs of the brain are constantly changing, mostly
12:47
because we're moving through the world. And
12:50
the term for that is a sensorimotor system. And
12:53
the brain can't understand its inputs unless it
12:55
knows how it's moving through the world. So
12:57
we learned by exploring, by moving
12:59
different places, picking things up, touching them,
13:01
and so on. And that's
13:03
all animals that move in the
13:06
world learn this way. So this idea that the
13:08
brain is a sensorimotor system has been known back
13:10
in the late 1800s, but
13:12
it's pretty much ignored by everybody. But
13:15
it leads to a very fundamental different
13:17
way of how we acquire knowledge and
13:19
how knowledge is represented in the brain.
13:22
Whereas today's AI most
13:25
of it's built on deep learning
13:27
or transformer technologies, which essentially we feed
13:29
it to it. It
13:31
doesn't explore it. And we feed to
13:34
large language models, we just feed it language.
13:36
So there's no inherent knowledge about what these
13:38
words mean, only what these words mean in
13:40
the context of other words. But
13:42
you and I can pick up a
13:44
cat and touch it and feel it and know
13:46
that it's warmth and we understand how its body
13:48
moves because no one has to tell us that.
13:51
We just experience it directly. So
13:54
this is a huge gap between
13:57
brains, pretty much all brains work by
13:59
sensorimotor learning. long
20:00
and around. And as you do, you build a
20:02
three-dimensional model of the cup, even though you're only
20:04
getting input from one fingertip. The
20:06
eyes are doing the same thing. It's
20:08
surprising. You don't realize this. So every
20:10
cortical column we understand now is doing
20:13
this sort of processing movement information and
20:15
sensor information, building what we call structure
20:17
or 3D models of things in the
20:19
world. So it's quite different than even
20:21
neuroscientists think about it. And
20:24
there's a lot of reasons we can talk about
20:26
how it was missed for all these years. So
20:28
in the cortex, you have essentially six layers of
20:30
cells. And a column is...
20:33
All six layers. Is all six layers. It's going
20:35
up and down. It's
20:38
like, think of it like layers of a cake. Right. Column
20:40
is you're taking a straw and
20:42
shoving it through the top. And so you've got this Good
20:44
analysis. This is the call. Okay. You
20:46
got a straw of cake. Okay,
20:48
great. And so the idea is
20:50
if you're looking at some column
20:52
in primary visual cortex,
20:55
yeah, your point, Jeff, was that it's
20:58
like looking at the world
21:00
through a straw. It only sees a little
21:02
tiny piece of the world, but because the
21:05
eyes are moving out, because you're exploring the
21:07
world, this is actually getting lots
21:09
of parts of information. It's exploring the world in
21:11
the same way that your fingertip explores the world.
21:13
Right. And it has to integrate information over time.
21:15
That's the key, right? And you can literally do
21:17
this. You can look at the world through a
21:19
straw. And you
21:21
can say, oh, what am I looking at? Well, you can't
21:23
tell until you start moving the straw. And then
21:26
you can start and you can also learn objects
21:28
that way. So literally, you can learn by looking
21:30
through a straw, which is what sort of one
21:32
column is doing. And
21:34
in your model, there are
21:37
thousands of such columns. And
21:39
each one of these is
21:41
learning a
21:43
model of the world as it's going. So tell
21:45
us about it. Right, right. So figure out what
21:47
it is and what it's doing. So
21:49
the trick of this thing is, it's a little tricky
21:51
here. You know, when you look out at the world,
21:53
you have a sense, anybody, you have
21:55
a sense where things are. I have a sense where you
21:57
are relative to me. I have a sense for this microphone.
22:00
to me, I know where my hand is, relative to this
22:02
cop. Now, it turns out
22:04
that you have any kind of sense of
22:06
location and space, you have to have neurons
22:08
representing it. There's nothing goes on
22:10
in the brain if there aren't neurons firing doing
22:12
it. It turns out most of
22:14
the machinery in the neocortex is keeping track
22:16
of where things are relative to other things.
22:19
So those six layers, all those cells, at
22:21
least half of that circuitry is
22:24
tracking where the sensory input is
22:26
coming from in the world. So if I move
22:28
my finger over this coffee cup, the
22:30
part that's getting information from my sensory, like
22:33
I'm sensing an edge, for example, as
22:35
I move my finger, it has to keep
22:37
track of where my finger is, a location
22:39
of it and its orientation relative to this
22:41
cup. It's quite complicated. But
22:44
that's what it has to do to build those models. And
22:46
now we know how it does it. There's all this evidence
22:48
for it. So the brain
22:50
is just trying to keep track of where all of
22:53
its inputs are in the world, all relative to other
22:55
things. Then it builds up these three dimensional models of
22:57
the world. So tell us about how it does that
22:59
then. Right. So you can think
23:01
about when you're in high school, you learned about Cartesian
23:03
coordinates, x, y, z coordinates, right? And
23:05
so if I wanted to say, where is something, where are
23:08
you relative to me? I might say, okay, your nose is
23:10
the origin. I could say it's some distance from here. And
23:12
so, you know, x, y, and z. Well,
23:14
you have to have something like that. But
23:17
brains don't do it that way. They do it another way. And
23:20
this was some very clever research in
23:22
the last 20 years that people discovered in
23:25
the antirhinal cortex and hippocampus, these cells
23:27
called grid cells and play cells, which
23:29
actually operate as reference frames. They are
23:32
a way of neurons to
23:34
represent locations. And
23:36
they work differently than x, y, and z. So
23:38
there's no origin. It's kind of really clever how
23:41
they work. Nature
23:43
has discovered a different way of doing this. So make
23:45
sure you tell us a little bit about that. Well,
23:47
okay. But these are well-known things.
23:49
Like grid cells, which are antirhinal cortex. What
23:51
they do is they, these cells, if you
23:53
take a set of them, individual cells are
23:56
not unique. And these real cells,
23:58
I fire at different locations in space. But if
24:00
you take a set of them, they're unique. And
24:02
so you can encode a unique location in space.
24:05
And the key thing about them is
24:07
these cells automatically update as
24:10
you move. So the original grid cells are where your
24:12
body is in a room. And
24:14
as you move, it's called path integration. It says,
24:17
OK, you're moving at this direction,
24:19
at this speed. So we'll just automatically update these
24:21
neurons as if we know where you are, right?
24:24
And so it's what sales used to do,
24:26
dead reckoning. You just say, oh, I'm heading
24:29
north for an hour or three knots there
24:31
for all these three miles in
24:33
this direction. So we know that these
24:35
cells exist. They've been well studied. People
24:37
wouldn't know about prize for these things. So
24:40
we speculated that the same
24:42
neural mechanisms, these grid cells
24:45
and equivalents, would be in the cortex in every cortical
24:47
column. And sure enough, they're
24:49
finding that now. So there's all
24:51
kinds of research now. They're finding in humans and
24:53
other animals that there are grid cell-like structures in
24:55
cortical columns. And so what does that tell you?
24:58
It tells me that that's the mechanism by which
25:00
the brain uses for reference frames. And so literally,
25:02
when you build a model of something in the
25:04
world, like a model of a cup or a
25:06
model of anything, essentially what
25:08
you're doing is you're saying, here's the sensation,
25:10
and here it's location. Here's another sensation at
25:12
a different location. Here's another sensation at a
25:14
different location. You add all these together, and
25:16
you get a three-dimensional model. You
25:18
can say, this thing consists of these features
25:20
in these locations relative to each other. And
25:24
so literally in our head, we build
25:26
models of the world that are three-dimensional
25:28
analogues of the physical things we
25:31
interact with. And that's
25:33
why you appear three-dimensional to me. You're
25:35
not an image. You're a three-dimensional structure, because I have
25:37
a three-dimensional model of humans and I have a special
25:39
model for you, David. OK,
25:43
great. India.
25:53
Let's post in the outfit of the day. We
25:55
look so good. Yeah, we do, but is that
25:57
a new phone I see? You were still
25:59
paying off your old phone- Plug
27:51
in a Hyundai EV and the extraordinary
27:54
happens. It's not just the ultra-fast charging
27:56
capability and long range in the IONIQ
27:58
5 and IONIQ 6, or
28:00
the adventure-seeking spirit of the Kona Electric,
28:02
or the groundbreaking 601 horsepower IONIQ 5N.
28:07
And it's not just the comfort and knowing that every
28:09
Hyundai EV is backed by a
28:11
10-year, 100,000-mile limited electric battery warranty.
28:13
Hyundai's EVs transform a low hum
28:15
into a loud adventure. They bring
28:17
color to your journey and turn
28:19
energy into main character energy. So
28:21
forget everything you thought you knew
28:23
about EVs and turn the
28:25
extraordinary into something truly electrifying.
28:28
There's joy in every journey. EPA estimated 3 to
28:30
3-mile driving range for 2024 IONIQ 5SCL limited
28:33
rear-wheel drive and 361-mile driving range for
28:36
2024 IONIQ 6SCL long-range rear-wheel drive with fully charged
28:38
battery. Estimated range varies for other models and trims
28:40
and depends on battery capacity, temperature options, and other
28:42
factors. Limited battery warranty covers defects in materials for
28:44
10 years or 100,000 miles, whichever occurs first. See
28:46
owner's manual for complete warranty details and limitations. Visit
28:48
usa.com or call 562-314-4603 for more details.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More