Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:03
Could you learn to fly a helicopter
0:05
not by practicing, but
0:08
instead by uploading the information
0:10
directly into your brain? What
0:13
would society do if kids
0:15
no longer had to go to school? And
0:17
what does any of this have to do with suntan
0:20
booths or nano robots or
0:22
torking over a presidential address
0:25
or what a cowboy on a hill
0:28
is simply not able to see.
0:34
Welcome to Inner Cosmos with me David
0:36
Eagleman. I'm a neuroscientist and
0:38
author at Stanford and in these
0:40
episodes we sail deeply into
0:42
our three pound universe to
0:45
understand why and how our
0:47
lives look the way they do. Today's
0:58
episode is about the potability
1:00
of really coming to understand
1:03
the tangled forest of eighty
1:05
six billion neurons in your head
1:08
and the trillions of connections between
1:10
them. And if we could do that, could
1:12
we upload information directly
1:15
into your brain? Could we speed
1:17
up education this way? Now?
1:20
At the moment, this is all pure
1:23
fantasy because we simply don't
1:25
have the technology to allow
1:27
us to do that. But the question we're going to
1:29
ask today is whether this is theoretically
1:33
possible and something we can look
1:35
forward to around the corner of the
1:37
next century, and what are the
1:39
caveats, the things to watch out for, and
1:42
the unexpected complexities
1:44
here. So let's get started some
1:47
hundreds of years ago and still
1:49
in many impoverished places in the world,
1:52
children of the species Homo sapiens
1:55
reproduce by the time they are young
1:57
teens. But this situation
1:59
is it's totally different in modern times
2:02
and modern societies. Now, young
2:04
people go to school for
2:06
their first eighteen years or twenty
2:09
one years, and increasingly
2:11
twenty five or twenty six years
2:13
for an advanced degree, and in fields
2:16
like medicine, they take another several
2:18
years of internship and residency.
2:21
And in a field like neuroscience research
2:23
people do a postdoctoral fellowship
2:26
and then they hope to become an assistant
2:28
professor, and then an associate professor
2:31
and finally a full professor. And most
2:33
people are in their forties
2:35
by the time they get there. So
2:37
what accounts for this recent historical
2:40
change. Why do we do so much schooling
2:42
for so much of our lives now?
2:45
Well, it's because we are a runaway
2:47
species. We've gone off in a totally different
2:50
direction than all our animal
2:52
cousins, and we have made thousands
2:55
of important discoveries about
2:57
our world and produced so much art
2:59
invarious forms, And as a result,
3:02
there's so much to learn,
3:04
and so we need to spend decades
3:07
in institutions of learning,
3:09
not to mention, reading books and listening
3:12
to podcasts to understand
3:15
what millions of humans have
3:17
devoted their lives to figuring
3:19
out. But what
3:22
if there were a way that we didn't have to do
3:24
that? What if there were a way to simply
3:26
upload the information, in other words, to put
3:29
the information directly into
3:31
your brain. So let's harken
3:33
back to this great scene
3:36
in The Matrix where
3:38
Neo and Trinity are being hotly
3:40
pursued by the antagonist, agent Smith,
3:43
and our two heroes end up on top
3:45
of a building, and there they spy a
3:48
helicopter parked on the roof, and
3:50
Neo asks Trinity do
3:52
you know how to fly that? And she replies
3:55
not yet, And she flips open
3:57
her phone and she calls Tank, the operator,
4:00
and she says, I need a pilot
4:02
program for a B two twelve helicopter,
4:05
And we see the operator rotate his
4:07
chair in front of his bank of computers
4:09
and he quickly types out a bunch of commands,
4:12
and she closes her eyes and
4:14
one second later she turns confidently
4:17
to Neo and says, let's
4:19
go. So what happened
4:21
is that Tank the operator had
4:23
taken the expertise,
4:25
the complicated know how of flying
4:28
a B two twelve helicopter and just
4:30
uploaded it to her brain. So
4:33
the question we're going to ask today is
4:35
is that theoretically possible from a
4:38
neuroscience perspective, and what
4:40
will make that straightforward? And what
4:42
will make that not straightforward to
4:44
accomplish someday. Now,
4:46
in some ways, the whole idea sounds crazy
4:49
because it seems like we always have
4:51
to earn things if we want changes
4:54
to our brains or body. You can't just get something
4:56
for free. But of course, people for
4:58
decades have been climbing in into suntan
5:00
booths instead of spending days outside,
5:03
and people get botox, which binds
5:05
to receptors the ends of peripheral nerves
5:08
and changes the wrinkliness of your face.
5:10
And people are increasingly doing things
5:13
to not have to go to the gym but instead to
5:15
have your abdominal muscles
5:17
built for you with electrical
5:20
stimulation. You just lie on the table
5:22
and your muscles contract over and over and
5:24
The idea is that your muscles can grow stronger
5:27
and look better without you having
5:30
to do a single sit up. You just lie there.
5:32
So what would be the equivalent in the realm
5:35
of education? Can we imagine a
5:37
time when you don't have to bury yourself
5:40
in a book to master some domain,
5:42
where you don't have to spend hundreds of hours
5:44
sitting in a flight simulator, but
5:47
instead you hook something up to your
5:49
brain and then it is as though
5:51
you already knew quantum
5:53
mechanics or electrical engineering, or Persian
5:55
history, or how to serve for hang
5:58
glide or repair that model of
6:00
dishwasher or whatever. Now,
6:03
how would you push information to the
6:05
brain? We currently do this by
6:08
sitting down dozens of children in
6:11
front of someone who already has the information
6:13
in their brain, and that person uses
6:16
words or pictures, and the students
6:18
attend to those stimuli and
6:21
try to translate those words
6:23
or pictures into changes
6:25
in their own private jungle of
6:27
billions of neurons. They try to convert
6:30
what they're hearing or seeing into
6:32
storage in their own internal model
6:35
in a way that makes sense to them. What
6:37
learning means is that you very
6:40
finely change the networks
6:42
in your head. That's
6:44
it. That's what we pay lots of tuition
6:46
for and go off to college for to
6:48
get someone who already has information
6:50
in their network to translate
6:53
it through the low bandwidth channel of language
6:56
over to your network. So, just
6:58
to be clear on this, before you
7:01
know some factor concept, your network
7:03
is configured in some way, and
7:05
then I tell you, oh, that dog's
7:08
name is Nebula, and then you encode
7:10
that information. This connection in
7:12
your brain gets strengthened and this one gets weakened,
7:14
and this synapse unplugs, it replugs
7:17
over there, and this happens over millions
7:19
of synapses, and then you know something
7:21
that you did not know before. And
7:24
for deeper knowledge, like flying
7:26
a B two twelve helicopter, this
7:28
requires not just the memory of
7:31
a fact, but of a procedure.
7:33
And so those changes happen in
7:35
different brain areas and they're more widespread.
7:38
But what is required in all these forms
7:41
of learning are simply changes
7:43
in the patterns of your network, presumably
7:46
just the synaptic connections, but maybe
7:49
other details as well, like which
7:51
neurotransmitter receptors are being expressed
7:54
on the membranes and whatever. But that's it,
7:56
that's what it means. To learn
7:58
something. So is
8:01
there any way to implement those
8:03
changes besides the old
8:05
fashioned way of sitting for a semester
8:08
in a classroom or spending hours
8:11
in the helicopter flight simulator. Well,
8:13
there's been a lot of excitement about brain
8:16
machine interfaces, such as the
8:18
brain electrodes that are implanted robotically
8:21
by the company Neurallink. So I'll
8:23
just take a quick moment to clarify
8:25
the landscape of electrodes
8:28
in the brain. Even though neuralink
8:30
hit the news recently. The first thing to
8:32
note is that these brain machine interfaces
8:35
have been around for many decades
8:37
since people started inserting
8:39
electrodes. These are just thin metal
8:41
wires into the brain. The idea
8:44
is that you just insert this electrode
8:46
into the neural tissue and you listen
8:49
to the electrical activity of the
8:51
cells. And researchers pretty quickly
8:53
figured out that if you send a little
8:55
bit of electricity down the wire down
8:57
this electrode, you can stimulate
9:00
the cell to make it active where
9:02
it pops off its own little electrical
9:05
spikes that travel around. So you
9:07
put in some electricity and it goes And
9:10
this is the technology behind, for example,
9:13
deep brain stimulation you might have heard
9:15
of this. Take Parkinson's disease.
9:17
There's a tiny brain region
9:20
called the subthalamic nucleus, and
9:22
it was discovered starting from work in the
9:24
nineteen seventies that you can insert
9:27
your electrode into this area and
9:29
zap it with a bit of electricity and
9:31
you get these amazing effects of the
9:34
movement problems of Parkinson's
9:37
essentially disappearing. And
9:39
by the way, the reason you can stick an electrode
9:41
into the brain is because the brain doesn't
9:43
have any pain receptors, so you
9:45
can just dunk the little metal wire right
9:47
in there after you've opened a little
9:50
portal in the skull. So what's
9:52
happening when you put these little bursts of electricity
9:54
in is that the cells
9:56
fire, which has effects on the rest of the network
9:59
that those cells are connected to, and it also changes
10:02
the electrical oscillations. And why this
10:04
works so well in Parkinson's is still a
10:06
bit of a mystery, but you get what
10:08
you want out of it, and people have been using
10:11
this sort of brain stimulation for all
10:13
kinds of purposes. For example, my
10:15
colleague Helen Mayberg puts electrodes
10:18
directly into a very specific area
10:21
near the singulate gyrus, and she stimulates
10:24
and can pull people out of deep
10:26
clinical depression this way. So
10:28
there are many labs and clinics using
10:30
the technique of stimulating individual
10:33
cells in the brain, and the direction of
10:35
the technology over the past couple of
10:37
decades has been getting
10:39
more and more electrodes implanted,
10:41
so that you're not just hitting one or a
10:43
few cells at the tip of the electrode, but
10:45
you're instead exciting tens
10:48
or hundreds or eventually thousands of cells
10:50
by using a whole specific
10:52
collection of electrodes.
10:55
And companies like Neuralink have become famous
10:57
in the public eye because of the idea
10:59
of sewing these electrodes very
11:02
finely into the brain and getting
11:04
a thousand of them and soon more than that.
11:06
And in all these cases, the electrodes can
11:09
read and write, in other words, they can
11:11
record the activity in the brain cells,
11:13
but they can also stimulate the brain cells
11:16
to put activity in there. So
11:33
once you have the electrodes in there, could
11:35
you just send in the right zaps
11:38
of electricity in just the right pattern,
11:40
spread over millions of neurons with precise
11:43
timing of your patterns in such
11:45
a way that you shape the
11:47
network so that you can
11:49
fly a helicopter. Now, all
11:52
that sounds pretty exciting as a theoretical
11:54
possibility, but I think
11:56
there are two major technical
11:59
hurdles here to be able to stimulate lots
12:01
and lots of cells in the brain in the way
12:03
that you might want to upload
12:06
helicopter instructions. The first is
12:08
simply a physical challenge.
12:11
The brain is very delicate, and
12:13
so Mother Nature has surrounded it
12:16
in the armored plating of the skull.
12:19
So it's very very hard to get at
12:21
this fragile, delicate tissue of
12:23
the brain, and so if you want to insert
12:25
an electrode, you have to actually drill a
12:27
small hole in the skull to expose
12:30
the brain and then you can put your electrode
12:32
in. The difficulty is that
12:34
there are eighty six billion neurons,
12:36
and at the moment, even with our
12:38
fanciest technology, we can only
12:41
get to say a thousand of
12:43
these at any time, and so that
12:46
is useless in terms of actually having
12:48
access to the whole system. It
12:50
would be equivalent to if you really
12:53
wanted to say something to all eight billion
12:55
people on the planet, but you
12:57
only had one hundred followers
13:00
on social media. The huge
13:02
majority of the world will have no idea
13:04
that you've ever said anything, or that you even
13:06
exist. And that's the situation.
13:08
When you zap a few hundred neurons,
13:11
the other tens of billions of neurons
13:13
don't even know that you're knocking on the door.
13:16
So to actually insert information
13:18
into the brain, you'd somehow need to access
13:21
all or at least most of the neurons
13:24
to make any meaningful change.
13:26
Now, I'm not yet addressing how you would
13:29
know what you want to change, where I'll come back to
13:31
that in a moment. Let's just imagine for now
13:33
that you know exactly what you want to tweak
13:35
in the brain. Now, I do think that in
13:37
the future there may be a very
13:40
different solution besides electrodes
13:43
to this issue of manipulating
13:45
the network, because I don't think the idea
13:47
of dunking electrodes in there is ever
13:49
going to be a long term solution.
13:52
When I squint into the future, I think the solution
13:55
is something like nano robots.
13:58
So what are nano robots. The idea is
14:00
that you use atomically
14:03
precise three D printing to
14:05
make little molecular machines
14:07
out of atoms. Essentially, you make little
14:10
robots that carry out some
14:13
functions, so they're like little robots,
14:15
but they're microscopically small,
14:18
built out of individual atoms, by
14:20
the way, which is what proteins are. Anyway,
14:23
you could make these super durable,
14:25
for example by printing them out of carbon,
14:28
making them diamond robots.
14:31
The idea, and this is probably not for
14:33
several decades. The idea is that you
14:36
swallow a pill with tens
14:38
of billions of these little nano robots
14:40
in there, and they float through your
14:42
bloodstream and you give them the right
14:45
FedEx labels to pass the
14:47
blood brain barrier, and once
14:49
they're in there in the brain, they
14:52
wiggle their way into your neurons
14:54
where they can read the activity and they
14:56
can cause the cell to spike
14:59
to fire signal whenever they
15:01
need to. So, with proper
15:04
signaling between the nanobots, using
15:06
for example, mesh networking, you could
15:08
in theory generate whatever
15:11
patterns you needed to across the entire
15:13
brain, and if your science is
15:15
really advanced, then you
15:17
hit the correct brain wide
15:20
patterns that will cement in the
15:22
knowledge of how to fly a
15:24
B two twelve. Now, although
15:26
this is not happening anytime soon,
15:29
it certainly seems plausible that
15:31
this could be in our future. But
15:33
wait, there's actually a
15:36
difficult twist to this story. I
15:38
said before there are two technical
15:40
hurdles, and here comes the second. And
15:42
that hurdle is that there won't
15:45
be a single program
15:48
for flying a B two twelve helicopter. Why
15:51
not, because the brain
15:53
inside each of us is totally
15:55
unique. We each have a massive
15:58
forest of eighty six billion euro on each
16:00
with ten thousand connection
16:03
points reaching out and interacting
16:05
with other trees. And it's a living
16:08
forest such as each connection, every
16:11
twig on every branch finds its
16:13
place in life based on the exact
16:15
details of what you have seen
16:18
and heard and experienced in your
16:21
life. You born in your
16:23
hometown, with your family, your
16:26
neighborhood, your culture, your
16:29
moment in history. All those
16:31
things determine the exact wiring
16:33
of your brain. And your brain has a
16:35
network that is different from his brain over
16:37
there, and her brain over there, and everyone
16:40
else's brain on the planet. And
16:42
the exact wiring is
16:44
what makes you you. So
16:48
in the proposed future of the Matrix,
16:50
the operator Tank would have to specify
16:53
that he wants a program to pilot
16:56
a B two twelve helicopter that
16:58
is specified exact exactly for Trinity's
17:01
brain, that is bespoke for
17:04
her neural network only. And
17:06
if Tank tried to upload the
17:08
same program to Neo's brain
17:11
or Morpheus's brain, who knows
17:13
what that would result in. Because if
17:15
the program alters the way
17:18
that neuron nineteen million, three hundred
17:20
fifty six three hundred and two is talking to its
17:22
neighbors, and it does this over
17:25
a million other neurons with high specificity,
17:28
that might teach Trinity how to
17:30
fly a helicopter, but it certainly would not
17:32
work for someone else whose brain
17:35
is different. So how do we get around that
17:37
problem, the problem of everyone having
17:40
a unique neural network. Well,
17:43
the answer will have to rely on
17:45
what is called system identification.
17:48
This is an engineering approach
17:50
where you have some complicated dynamic
17:53
system and you measure lots
17:56
of input output pairs,
17:58
as in, when I put this in, what
18:01
happens? Okay, now it happens if I put that
18:03
in. So imagine you find a
18:05
really complicated machine and
18:07
you don't know exactly what it does. So you tap
18:09
one of the keys and you see how it
18:12
moves, and then you tap three
18:14
of the keys at the same time, and you look at
18:16
what it does as its output, and then
18:18
you hit a series of the keys and you
18:20
see what results. And you do this
18:22
over and over and over to try to figure
18:24
out what is the structure under
18:27
the hood. This system identification
18:30
approach is used in lots of fields. For example,
18:32
in economics, let's say you want to figure
18:35
out the guts of the stock market.
18:37
So you take lots of inputs like
18:39
gross domestic product and inflation and
18:41
unemployment and interest rates and blah blah blah,
18:44
and you look at all these as inputs
18:46
and you look at the reaction of the market this
18:48
way, and you develop better and better
18:50
mathematical models of what the machinery
18:53
of the stock market is doing, even
18:55
though you can't see it. Okay,
18:57
So the question is, could you do system
19:00
identification on a human
19:03
brain. No one's ever really
19:05
done this because there's no purpose for it now,
19:07
but someday it might make sense.
19:10
So the idea is you go into a super
19:13
futuristic brain scanner and
19:15
you get lots of inputs, and
19:17
this sophisticated brain imaging device
19:20
measures the outputs, in other words, which
19:22
cells in your brain are responding.
19:25
So you see a rapid series
19:27
of images and you hear words,
19:29
and you feel touches on your body, and you smell
19:32
smells, and you run through thousands
19:35
or maybe millions of little micro
19:38
experiences while your brain is getting
19:40
measured. And in theory, this is
19:42
how a scientist could say, Aha,
19:45
Trinity's brain is organized
19:47
like this, while Neo's brain is laid
19:49
out like that, and Morpheus's brain
19:52
has a slightly different pattern, And you
19:54
might find that for teaching
19:56
the operation of a B two
19:58
twelve helicopter, in his brain
20:01
thinks about it in analogy to
20:04
riding a horse and controlling it,
20:06
which let's say she grew up riding horses,
20:08
while Neo's brain would learn
20:11
the helicopter in analogy to
20:13
the way a motorcycle feels, which is, let's say
20:15
how he grew up. And for Morpheus,
20:18
the actions of piloting emerge
20:20
from his deep knowledge of surfing, which
20:22
is how he grew up and what is stored in his brain.
20:25
Now, it's not clear how many inputs
20:27
you'd have to ping in there to get high
20:30
enough resolution to make all the
20:32
little changes you need, but presumably
20:35
that would get figured out with enough experimentation.
20:38
Okay, so let's say we
20:41
as a society grow to
20:43
a point where we can do system
20:46
identification on an individual's
20:48
brain and then use nanobots
20:51
to upload knowledge of
20:53
helicopter piloting. I need to emphasize
20:56
that this is not right around the corner,
20:58
but it certainly seems the theoretically
21:00
plausible. Another century
21:03
of advancement, and suddenly the
21:05
network that makes you can
21:08
get directed and shaped
21:10
in a bespoke manner. And if
21:12
we come to a point where we can do it,
21:15
that's possibly the biggest societal
21:18
change. I can imagine you say
21:20
to your three year old kid, Okay, we're gonna upload
21:22
first grade now. Great, Now, go play
21:24
outside for an hour, and then we're gonna upload
21:26
second grade after lunch. Imagine
21:29
that by the end of the week, your three
21:31
year old knows as much as
21:33
a full professor does. Now, so
21:35
what becomes of society and
21:38
the way we run it now? You may
21:40
think the analogy here is to look at super
21:42
smart, genius kids in our current world,
21:45
But these kids often go
21:47
off to attend college at twelve years old,
21:49
and they very often end up lonely
21:51
and socially misplaced, because really what they
21:53
want is to play with their colleagues other kids
21:56
their age, But they get stuck with a bunch
21:58
of older kids who have gone through puberty
22:01
and are running deeply carved evolutionary
22:04
programs that cause their brains
22:06
to be taken over by sexuality, and
22:08
that software hasn't yet turned on in the
22:10
heads of these young genuses, and as
22:12
a result, they can't mesh with what is
22:14
happening around them, and they can feel very lonely
22:16
in these contexts. But the
22:19
future scenario of uploading knowledge
22:22
is totally different because now every
22:25
single kid can stay among colleagues.
22:28
But the question is if
22:30
education is uploaded, what do the kids
22:33
do all day? Do they launch
22:36
startups at the age of six, do
22:38
they write epic novels by
22:40
the time they're eight years old? Do
22:42
they return to reproducing
22:44
as teenagers like their distant ancestors
22:47
did? And is it dangerous that
22:49
they have all the knowledge
22:51
of decades of schooling but
22:53
without the maturity. The most
22:56
slowly developing part of the brain is the prefrontal
22:58
cortex, and this underlies our ability
23:01
to simulate possible futures
23:03
and think about consequences. So imagine
23:06
a kid with an undeveloped prefederal cortex
23:08
who has all the knowledge that Albert
23:10
Einstein commanded at midlife.
23:13
But this child lacks the ability to simulate
23:16
consequences, so they think something like,
23:18
wouldn't it be hilarious to build a
23:21
small nuclear bomb and blow up my neighbor's
23:23
porch, Or wouldn't it be
23:25
a crackup to disrupt
23:27
the presidential broadcast by
23:30
hijacking the frequency and imposing
23:32
a video of me twerking or
23:34
whatever? Because children
23:36
don't yet have a fully developed
23:39
profederal cortex that can't simulate
23:41
consequences the way an adult can, and
23:44
this is why it could be dangerous
23:46
to inject the knowledge of an adult
23:48
into a child's brain. Now,
24:04
perhaps I'm being shortsighted here, and we
24:06
could somehow upload maturity as
24:08
well. We could figure out
24:11
the learning that translates
24:13
to morally complex situations
24:16
and simulate those over and over do
24:18
the synaptic equivalent of working
24:21
through the possibilities and
24:23
feeling the consequences. Maybe
24:25
you could massively speed up emotional
24:28
learning that way. After all, as my
24:30
father would always tell me, the wise
24:32
person learns from experience,
24:35
but the wiser person learns
24:37
from the experience of others.
24:40
So maybe there could be enough uploaded
24:42
knowledge where a kid understands
24:45
various possible scenarios and outcomes,
24:48
and the good decision making simply results
24:50
from a deep knowledge of previous
24:53
examples, things that have happened to other people,
24:56
all of which have been uploaded.
24:58
So maybe the maturity problem could be taken
25:01
care of, but still we're looking
25:03
at massive societal shifts
25:05
that would render our current civilization
25:09
totally unrecognizable. Now,
25:11
we all like to be very thoughtful about the
25:13
future, but it doesn't matter what
25:16
we speculate about it, because we are guaranteed
25:19
to be wrong. We can only envision
25:21
what we're capable of, in this case, a
25:23
cartoonish version of a bunch
25:26
of super intelligent kids running around while
25:28
their parents go off to their jobs. But
25:30
the world is likely to be very
25:32
different by then. Presuming that
25:34
everything is massively sped up
25:37
by artificial intelligence, it seems
25:39
very possible that society
25:41
is going to evolve exponentially
25:44
faster at a pace that we really
25:46
can't conceive of here in the first
25:49
third of the twenty first century. I mean, just
25:51
imagine that AI knocks
25:54
down scientific problems rapidly,
25:56
such that we move from our current
25:58
state of pretty wide spread ignorance
26:01
to perfect, wonderful
26:03
models of everything in the cosmos.
26:05
Just think about the incredibly
26:08
slow pace between the Stone
26:10
Age and the Bronze Age, and then
26:12
the Bronze Age to the Silver Age.
26:15
Now imagine this pace goes up by a
26:17
thousandfold or a millionfold. So we
26:19
find ourselves a few decades from now
26:22
in the Diamond Age, where we
26:24
can manipulate carbon atoms however we like.
26:27
And then a few years later we're
26:29
past that and into a new era where we
26:31
can entangle photons and
26:33
find ourselves in the quantum age
26:36
and so on. Like everyone, I
26:38
love to speculate about the future, but the truth
26:41
is that it is impossible to
26:43
picture what things will become and how
26:45
quickly. And I want to share an example.
26:48
Last month here in Silicon Valley,
26:50
I saw a black and white photograph
26:53
from nineteen forty. It was a man
26:55
on horseback ambling up
26:58
a dirt road on a hills and
27:00
there was nothing particularly special about
27:02
this sandy hill with its
27:04
scrubbrush. So I was intrigued
27:07
to read the caption and find
27:10
out that this little dirt road was
27:13
sand Hill Road. Now
27:15
you may know that sand Hill Road is nowadays
27:18
a road almost as famous
27:20
as Wall Street in New York. Sand
27:22
Hill Road is where many
27:24
of the world's most elite venture
27:26
capitalists do their business. They
27:29
invest hundreds of billions. This road
27:32
is the mecca for startups
27:34
who are seeking investment. Now,
27:36
the thing that was so striking to me is
27:39
that for the horseman sauntering
27:41
up this sandy hillside in nineteen forty
27:43
in the hot sun, there's no way
27:46
he could have imagined that the
27:49
lonely hoof prints he was leaving
27:51
would in just sixty years mark
27:53
this spot of one of the world's economic
27:56
engines. And there's no way
27:58
he could have envisioned what advances
28:01
would get funded on that spot. The
28:03
worldwide light speed network
28:06
that allows anyone on the planet to
28:08
effortlessly communicate to anyone else,
28:10
or rectangles that everyone
28:13
would carry in their pocket like a
28:15
handkerchief or a tobacco tin. But
28:17
these rectangles would contain the
28:20
accumulated knowledge of all
28:22
humankind. Or satellites
28:25
or quantum computers or blockchain
28:27
cryptocurrencies, or large
28:29
language models that could read every
28:31
book ever written. None of these
28:34
would be even vaguely imaginable
28:37
to the cowboy in nineteen
28:39
forty, moving slowly up
28:41
the hill. We are blind
28:44
to the future. I often wish I could
28:46
talk to whoever is listening to
28:48
this historical podcast
28:50
in the year twenty eighty four, because
28:53
the world will be so different by then,
28:55
and I am incapable of
28:57
imagining it. And it's not just that
29:00
we are not being creative about extrapolating
29:03
technology curves into the future. It's that
29:05
there will be new technologies
29:07
and novel sciences and new
29:10
convergences that will make it
29:12
intrinsically unpredictable.
29:15
There will be serendipitous discoveries
29:17
and socioeconomic changes and geopolitical
29:20
events. While we always make
29:23
guesses based on our current trends and
29:25
research, the future is
29:27
shaped by hundreds of things we just can't
29:29
see. Not only that, but you've heard me
29:32
speak before about our limited
29:34
perspective, our inability
29:36
to see past the fence line of
29:39
what we already know. Our current
29:41
knowledge understanding are based on the technologies
29:43
and paradigms that exist right now, so
29:46
it's really hard for us to anticipate
29:49
breakthroughs or paradigm shifts
29:51
that are going to radically alter our
29:54
society in the future. But this idea
29:56
of putting information directly
29:58
into the brain, that's it certainly
30:00
seems like that could be a big shift. So
30:03
when we think about the future, it's
30:05
more than just adults like us
30:07
riding around on a spaceship with a robot
30:10
or two. Things are guaranteed
30:12
to be weirder than we expect.
30:15
While brain uploads our science fiction
30:17
right now, assuming we don't blow
30:19
ourselves up, this inevitably
30:21
seems like it will become science fact.
30:25
So let's wrap this up. This
30:27
episode is not about what's going to happen
30:29
anytime soon, but I think it is inevitably
30:32
what will happen in the future. After
30:34
all, the brain is made of billions of cells,
30:37
each one of which is very complicated, and
30:39
each is connected in very complicated
30:41
patterns. But fundamentally, learning
30:44
and memory take place in the
30:46
changes of connectivity,
30:48
and as far as we can tell, that's all
30:51
learning is. So what we talked
30:53
about is the way that the jungle
30:55
of neurons in your head is wired up differently
30:57
than in your friend's head because you have
31:00
different genetic predispositions, and more
31:02
importantly, you have different experiences
31:04
in life. So in order to upload
31:07
any changes into your network, we'd
31:09
have to know your brain in exquisitely
31:12
fine detail, and we'd have to know those
31:14
patterns right now, because it's just a
31:16
little bit different than it was yesterday.
31:19
But in theory, if we had this information
31:22
and understood the language of the connections,
31:25
we could dial knobs here and there
31:27
in a million other spots, strengthening
31:30
or weakening synapta connections, tickling
31:32
the genome to express a little more neurotransmit
31:35
or receptor over here, a little less over there,
31:38
and after that you might be able
31:40
to suddenly possess some knowledge
31:44
you didn't have before. Now,
31:47
obviously, society will have to be very
31:49
careful about this technology when that
31:51
century comes, because in theory,
31:53
you could use it to implant
31:56
false memories, or to erase
31:58
knowledge, or to do any number of
32:01
nefarious things. So we will enter
32:03
a very strange time, and like every
32:05
technology, a whole raft
32:08
of protections and legislation
32:10
will grow up around it. Again,
32:12
this is likely impossible to achieve
32:15
in our generation because of the
32:17
size of the problem. It would take
32:19
about a zetabyte of information
32:22
to store the detailed structure
32:24
of one human brain, and that, by the way,
32:26
would only tell you the structure of the forest
32:28
of neurons, but wouldn't even tell you anything
32:31
about their individual details,
32:33
like which genes are getting expressed
32:35
and which proteins are getting put where. So
32:37
for us, the citizens of the
32:40
twenty first century, this is likely
32:42
to be an unsolvably huge
32:44
problem to capture a detailed
32:47
description of an individual
32:49
brain. But as a species,
32:52
we're in an interesting situation because
32:55
we can see that this is all coming,
32:58
and we can speculate on
33:00
the size of the changes this will
33:02
have on society writ large.
33:05
Now, what I find amazing is our guaranteed
33:08
inability to correctly picture
33:10
this future world, even though it
33:12
will be populated by our own great
33:15
grandchildren. Given all this,
33:17
I think the only specific
33:19
prediction we can make is that
33:21
we have more in common with
33:24
our ancestors two million years
33:26
ago than we do with our
33:28
descendants two hundred
33:30
years from now. In
33:36
the meantime, go to eagleman dot com
33:38
slash podcast for more information
33:40
and to find further reading. Send
33:42
me an email at podcast at eagleman
33:44
dot com with questions for discussion,
33:47
and check out and subscribe to Inner Cosmos
33:50
on YouTube for videos of each
33:52
episode and to leave comments.
33:54
Until next time. I'm David Eagleman,
33:56
and this is Inner Cosmos.
34:02
You and not you. You
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More