sponsored links
TED1994

Danny Hillis: Back to the future (of 1994)

February 20, 1994

From deep in the TED archive, Danny Hillis outlines an intriguing theory of how and why technological change seems to be accelerating, by linking it to the very evolution of life itself. The presentation techniques he uses may look dated, but the ideas are as relevant as ever.

Danny Hillis - Computer theorist
Inventor, scientist, author, engineer -- over his broad career, Danny Hillis has turned his ever-searching brain on an array of subjects, with surprising results. Full bio

sponsored links
Double-click the English subtitles below to play the video.
Because I usually take the role
00:15
of trying to explain to people
00:18
how wonderful the new technologies
00:20
that are coming along are going to be,
00:23
and I thought that, since I was among friends here,
00:25
I would tell you what I really think
00:28
and try to look back and try to understand
00:32
what is really going on here
00:34
with these amazing jumps in technology
00:37
that seem so fast that we can barely keep on top of it.
00:42
So I'm going to start out
00:45
by showing just one very boring technology slide.
00:47
And then, so if you can just turn on the slide that's on.
00:50
This is just a random slide
00:56
that I picked out of my file.
00:58
What I want to show you is not so much the details of the slide,
01:00
but the general form of it.
01:03
This happens to be a slide of some analysis that we were doing
01:05
about the power of RISC microprocessors
01:08
versus the power of local area networks.
01:11
And the interesting thing about it
01:14
is that this slide,
01:16
like so many technology slides that we're used to,
01:18
is a sort of a straight line
01:21
on a semi-log curve.
01:23
In other words, every step here
01:25
represents an order of magnitude
01:27
in performance scale.
01:29
And this is a new thing
01:31
that we talk about technology
01:33
on semi-log curves.
01:35
Something really weird is going on here.
01:37
And that's basically what I'm going to be talking about.
01:39
So, if you could bring up the lights.
01:42
If you could bring up the lights higher,
01:47
because I'm just going to use a piece of paper here.
01:49
Now why do we draw technology curves
01:52
in semi-log curves?
01:54
Well the answer is, if I drew it on a normal curve
01:56
where, let's say, this is years,
01:59
this is time of some sort,
02:01
and this is whatever measure of the technology
02:03
that I'm trying to graph,
02:06
the graphs look sort of silly.
02:09
They sort of go like this.
02:12
And they don't tell us much.
02:15
Now if I graph, for instance,
02:18
some other technology, say transportation technology,
02:21
on a semi-log curve,
02:23
it would look very stupid, it would look like a flat line.
02:25
But when something like this happens,
02:28
things are qualitatively changing.
02:30
So if transportation technology
02:32
was moving along as fast as microprocessor technology,
02:34
then the day after tomorrow,
02:37
I would be able to get in a taxi cab
02:39
and be in Tokyo in 30 seconds.
02:41
It's not moving like that.
02:43
And there's nothing precedented
02:45
in the history of technology development
02:47
of this kind of self-feeding growth
02:49
where you go by orders of magnitude every few years.
02:51
Now the question that I'd like to ask is,
02:54
if you look at these exponential curves,
02:57
they don't go on forever.
03:00
Things just can't possibly keep changing
03:03
as fast as they are.
03:06
One of two things is going to happen.
03:08
Either it's going to turn into a sort of classical S-curve like this,
03:11
until something totally different comes along,
03:15
or maybe it's going to do this.
03:19
That's about all it can do.
03:21
Now I'm an optimist,
03:23
so I sort of think it's probably going to do something like that.
03:25
If so, that means that what we're in the middle of right now
03:28
is a transition.
03:31
We're sort of on this line
03:33
in a transition from the way the world used to be
03:35
to some new way that the world is.
03:37
And so what I'm trying to ask, what I've been asking myself,
03:40
is what's this new way that the world is?
03:43
What's that new state that the world is heading toward?
03:46
Because the transition seems very, very confusing
03:49
when we're right in the middle of it.
03:52
Now when I was a kid growing up,
03:54
the future was kind of the year 2000,
03:57
and people used to talk about what would happen in the year 2000.
04:00
Now here's a conference
04:04
in which people talk about the future,
04:06
and you notice that the future is still at about the year 2000.
04:08
It's about as far as we go out.
04:11
So in other words, the future has kind of been shrinking
04:13
one year per year
04:16
for my whole lifetime.
04:19
Now I think that the reason
04:22
is because we all feel
04:24
that something's happening there.
04:26
That transition is happening. We can all sense it.
04:28
And we know that it just doesn't make too much sense
04:30
to think out 30, 50 years
04:32
because everything's going to be so different
04:34
that a simple extrapolation of what we're doing
04:37
just doesn't make any sense at all.
04:39
So what I would like to talk about
04:42
is what that could be,
04:44
what that transition could be that we're going through.
04:46
Now in order to do that
04:49
I'm going to have to talk about a bunch of stuff
04:52
that really has nothing to do
04:54
with technology and computers.
04:56
Because I think the only way to understand this
04:58
is to really step back
05:00
and take a long time scale look at things.
05:02
So the time scale that I would like to look at this on
05:04
is the time scale of life on Earth.
05:07
So I think this picture makes sense
05:13
if you look at it a few billion years at a time.
05:15
So if you go back
05:19
about two and a half billion years,
05:21
the Earth was this big, sterile hunk of rock
05:23
with a lot of chemicals floating around on it.
05:26
And if you look at the way
05:29
that the chemicals got organized,
05:31
we begin to get a pretty good idea of how they do it.
05:33
And I think that there's theories that are beginning to understand
05:36
about how it started with RNA,
05:39
but I'm going to tell a sort of simple story of it,
05:41
which is that, at that time,
05:44
there were little drops of oil floating around
05:46
with all kinds of different recipes of chemicals in them.
05:49
And some of those drops of oil
05:52
had a particular combination of chemicals in them
05:54
which caused them to incorporate chemicals from the outside
05:56
and grow the drops of oil.
05:59
And those that were like that
06:02
started to split and divide.
06:04
And those were the most primitive forms of cells in a sense,
06:06
those little drops of oil.
06:09
But now those drops of oil weren't really alive, as we say it now,
06:11
because every one of them
06:14
was a little random recipe of chemicals.
06:16
And every time it divided,
06:18
they got sort of unequal division
06:20
of the chemicals within them.
06:23
And so every drop was a little bit different.
06:25
In fact, the drops that were different in a way
06:28
that caused them to be better
06:30
at incorporating chemicals around them,
06:32
grew more and incorporated more chemicals and divided more.
06:34
So those tended to live longer,
06:37
get expressed more.
06:39
Now that's sort of just a very simple
06:42
chemical form of life,
06:45
but when things got interesting
06:47
was when these drops
06:50
learned a trick about abstraction.
06:52
Somehow by ways that we don't quite understand,
06:55
these little drops learned to write down information.
06:58
They learned to record the information
07:01
that was the recipe of the cell
07:03
onto a particular kind of chemical
07:05
called DNA.
07:07
So in other words, they worked out,
07:09
in this mindless sort of evolutionary way,
07:11
a form of writing that let them write down what they were,
07:14
so that that way of writing it down could get copied.
07:17
The amazing thing is that that way of writing
07:20
seems to have stayed steady
07:23
since it evolved two and a half billion years ago.
07:25
In fact the recipe for us, our genes,
07:27
is exactly that same code and that same way of writing.
07:30
In fact, every living creature is written
07:33
in exactly the same set of letters and the same code.
07:36
In fact, one of the things that I did
07:38
just for amusement purposes
07:40
is we can now write things in this code.
07:42
And I've got here a little 100 micrograms of white powder,
07:44
which I try not to let the security people see at airports.
07:50
(Laughter)
07:54
But this has in it --
07:56
what I did is I took this code --
07:58
the code has standard letters that we use for symbolizing it --
08:00
and I wrote my business card onto a piece of DNA
08:03
and amplified it 10 to the 22 times.
08:06
So if anyone would like a hundred million copies of my business card,
08:09
I have plenty for everyone in the room,
08:12
and, in fact, everyone in the world,
08:14
and it's right here.
08:16
(Laughter)
08:19
If I had really been a egotist,
08:26
I would have put it into a virus and released it in the room.
08:28
(Laughter)
08:31
So what was the next step?
08:39
Writing down the DNA was an interesting step.
08:41
And that caused these cells --
08:43
that kept them happy for another billion years.
08:45
But then there was another really interesting step
08:47
where things became completely different,
08:49
which is these cells started exchanging and communicating information,
08:52
so that they began to get communities of cells.
08:55
I don't know if you know this,
08:57
but bacteria can actually exchange DNA.
08:59
Now that's why, for instance,
09:01
antibiotic resistance has evolved.
09:03
Some bacteria figured out how to stay away from penicillin,
09:05
and it went around sort of creating its little DNA information
09:08
with other bacteria,
09:11
and now we have a lot of bacteria that are resistant to penicillin,
09:13
because bacteria communicate.
09:16
Now what this communication allowed
09:18
was communities to form
09:20
that, in some sense, were in the same boat together;
09:22
they were synergistic.
09:24
So they survived
09:26
or they failed together,
09:28
which means that if a community was very successful,
09:30
all the individuals in that community
09:32
were repeated more
09:34
and they were favored by evolution.
09:36
Now the transition point happened
09:39
when these communities got so close
09:41
that, in fact, they got together
09:43
and decided to write down the whole recipe for the community
09:45
together on one string of DNA.
09:48
And so the next stage that's interesting in life
09:51
took about another billion years.
09:53
And at that stage,
09:55
we have multi-cellular communities,
09:57
communities of lots of different types of cells,
09:59
working together as a single organism.
10:01
And in fact, we're such a multi-cellular community.
10:03
We have lots of cells
10:06
that are not out for themselves anymore.
10:08
Your skin cell is really useless
10:10
without a heart cell, muscle cell,
10:13
a brain cell and so on.
10:15
So these communities began to evolve
10:17
so that the interesting level on which evolution was taking place
10:19
was no longer a cell,
10:22
but a community which we call an organism.
10:24
Now the next step that happened
10:28
is within these communities.
10:30
These communities of cells,
10:32
again, began to abstract information.
10:34
And they began building very special structures
10:36
that did nothing but process information within the community.
10:39
And those are the neural structures.
10:42
So neurons are the information processing apparatus
10:44
that those communities of cells built up.
10:47
And in fact, they began to get specialists in the community
10:50
and special structures
10:52
that were responsible for recording,
10:54
understanding, learning information.
10:56
And that was the brains and the nervous system
10:59
of those communities.
11:01
And that gave them an evolutionary advantage.
11:03
Because at that point,
11:05
an individual --
11:08
learning could happen
11:11
within the time span of a single organism,
11:13
instead of over this evolutionary time span.
11:15
So an organism could, for instance,
11:18
learn not to eat a certain kind of fruit
11:20
because it tasted bad and it got sick last time it ate it.
11:22
That could happen within the lifetime of a single organism,
11:26
whereas before they'd built these special information processing structures,
11:29
that would have had to be learned evolutionarily
11:33
over hundreds of thousands of years
11:35
by the individuals dying off that ate that kind of fruit.
11:38
So that nervous system,
11:41
the fact that they built these special information structures,
11:43
tremendously sped up the whole process of evolution.
11:46
Because evolution could now happen within an individual.
11:49
It could happen in learning time scales.
11:52
But then what happened
11:55
was the individuals worked out,
11:57
of course, tricks of communicating.
11:59
And for example,
12:01
the most sophisticated version that we're aware of is human language.
12:03
It's really a pretty amazing invention if you think about it.
12:06
Here I have a very complicated, messy,
12:09
confused idea in my head.
12:11
I'm sitting here making grunting sounds basically,
12:14
and hopefully constructing a similar messy, confused idea in your head
12:17
that bears some analogy to it.
12:20
But we're taking something very complicated,
12:22
turning it into sound, sequences of sounds,
12:24
and producing something very complicated in your brain.
12:27
So this allows us now
12:31
to begin to start functioning
12:33
as a single organism.
12:35
And so, in fact, what we've done
12:38
is we, humanity,
12:41
have started abstracting out.
12:43
We're going through the same levels
12:45
that multi-cellular organisms have gone through --
12:47
abstracting out our methods of recording,
12:49
presenting, processing information.
12:52
So for example, the invention of language
12:54
was a tiny step in that direction.
12:56
Telephony, computers,
12:59
videotapes, CD-ROMs and so on
13:01
are all our specialized mechanisms
13:04
that we've now built within our society
13:06
for handling that information.
13:08
And it all connects us together
13:10
into something
13:13
that is much bigger
13:15
and much faster
13:17
and able to evolve
13:19
than what we were before.
13:21
So now, evolution can take place
13:23
on a scale of microseconds.
13:25
And you saw Ty's little evolutionary example
13:27
where he sort of did a little bit of evolution
13:29
on the Convolution program right before your eyes.
13:31
So now we've speeded up the time scales once again.
13:34
So the first steps of the story that I told you about
13:37
took a billion years a piece.
13:39
And the next steps,
13:41
like nervous systems and brains,
13:43
took a few hundred million years.
13:45
Then the next steps, like language and so on,
13:47
took less than a million years.
13:50
And these next steps, like electronics,
13:52
seem to be taking only a few decades.
13:54
The process is feeding on itself
13:56
and becoming, I guess, autocatalytic is the word for it --
13:58
when something reinforces its rate of change.
14:01
The more it changes, the faster it changes.
14:04
And I think that that's what we're seeing here in this explosion of curve.
14:07
We're seeing this process feeding back on itself.
14:10
Now I design computers for a living,
14:13
and I know that the mechanisms
14:16
that I use to design computers
14:18
would be impossible
14:21
without recent advances in computers.
14:23
So right now, what I do
14:25
is I design objects at such complexity
14:27
that it's really impossible for me to design them in the traditional sense.
14:30
I don't know what every transistor in the connection machine does.
14:33
There are billions of them.
14:37
Instead, what I do
14:39
and what the designers at Thinking Machines do
14:41
is we think at some level of abstraction
14:44
and then we hand it to the machine
14:46
and the machine takes it beyond what we could ever do,
14:48
much farther and faster than we could ever do.
14:51
And in fact, sometimes it takes it by methods
14:54
that we don't quite even understand.
14:56
One method that's particularly interesting
14:59
that I've been using a lot lately
15:01
is evolution itself.
15:04
So what we do
15:06
is we put inside the machine
15:08
a process of evolution
15:10
that takes place on the microsecond time scale.
15:12
So for example,
15:14
in the most extreme cases,
15:16
we can actually evolve a program
15:18
by starting out with random sequences of instructions.
15:20
Say, "Computer, would you please make
15:24
a hundred million random sequences of instructions.
15:26
Now would you please run all of those random sequences of instructions,
15:29
run all of those programs,
15:32
and pick out the ones that came closest to doing what I wanted."
15:34
So in other words, I define what I wanted.
15:37
Let's say I want to sort numbers,
15:39
as a simple example I've done it with.
15:41
So find the programs that come closest to sorting numbers.
15:43
So of course, random sequences of instructions
15:46
are very unlikely to sort numbers,
15:49
so none of them will really do it.
15:51
But one of them, by luck,
15:53
may put two numbers in the right order.
15:55
And I say, "Computer,
15:57
would you please now take the 10 percent
15:59
of those random sequences that did the best job.
16:02
Save those. Kill off the rest.
16:04
And now let's reproduce
16:06
the ones that sorted numbers the best.
16:08
And let's reproduce them by a process of recombination
16:10
analogous to sex."
16:13
Take two programs and they produce children
16:15
by exchanging their subroutines,
16:18
and the children inherit the traits of the subroutines of the two programs.
16:20
So I've got now a new generation of programs
16:23
that are produced by combinations
16:26
of the programs that did a little bit better job.
16:28
Say, "Please repeat that process."
16:30
Score them again.
16:32
Introduce some mutations perhaps.
16:34
And try that again and do that for another generation.
16:36
Well every one of those generations just takes a few milliseconds.
16:39
So I can do the equivalent
16:42
of millions of years of evolution on that
16:44
within the computer in a few minutes,
16:46
or in the complicated cases, in a few hours.
16:49
At the end of that, I end up with programs
16:51
that are absolutely perfect at sorting numbers.
16:54
In fact, they are programs that are much more efficient
16:56
than programs I could have ever written by hand.
16:59
Now if I look at those programs,
17:01
I can't tell you how they work.
17:03
I've tried looking at them and telling you how they work.
17:05
They're obscure, weird programs.
17:07
But they do the job.
17:09
And in fact, I know, I'm very confident that they do the job
17:11
because they come from a line
17:14
of hundreds of thousands of programs that did the job.
17:16
In fact, their life depended on doing the job.
17:18
(Laughter)
17:21
I was riding in a 747
17:26
with Marvin Minsky once,
17:28
and he pulls out this card and says, "Oh look. Look at this.
17:30
It says, 'This plane has hundreds of thousands of tiny parts
17:33
working together to make you a safe flight.'
17:37
Doesn't that make you feel confident?"
17:41
(Laughter)
17:43
In fact, we know that the engineering process doesn't work very well
17:45
when it gets complicated.
17:48
So we're beginning to depend on computers
17:50
to do a process that's very different than engineering.
17:52
And it lets us produce things of much more complexity
17:56
than normal engineering lets us produce.
17:59
And yet, we don't quite understand the options of it.
18:01
So in a sense, it's getting ahead of us.
18:04
We're now using those programs
18:06
to make much faster computers
18:08
so that we'll be able to run this process much faster.
18:10
So it's feeding back on itself.
18:13
The thing is becoming faster
18:16
and that's why I think it seems so confusing.
18:18
Because all of these technologies are feeding back on themselves.
18:20
We're taking off.
18:23
And what we are is we're at a point in time
18:25
which is analogous to when single-celled organisms
18:28
were turning into multi-celled organisms.
18:30
So we're the amoebas
18:33
and we can't quite figure out what the hell this thing is we're creating.
18:35
We're right at that point of transition.
18:38
But I think that there really is something coming along after us.
18:40
I think it's very haughty of us
18:43
to think that we're the end product of evolution.
18:45
And I think all of us here
18:48
are a part of producing
18:50
whatever that next thing is.
18:52
So lunch is coming along,
18:54
and I think I will stop at that point,
18:56
before I get selected out.
18:58
(Applause)
19:00

sponsored links

Danny Hillis - Computer theorist
Inventor, scientist, author, engineer -- over his broad career, Danny Hillis has turned his ever-searching brain on an array of subjects, with surprising results.

Why you should listen

Danny Hillis is an inventor, scientist, author and engineer. While completing his doctorate at MIT, he pioneered the concept of parallel computers that is now the basis for most supercomputers, as well as the RAID array. He holds over 100 US patents, covering parallel computers, disk arrays, forgery prevention methods, and various electronic and mechanical devices, and has recently been working on problems in medicine as well. He is also the designer of a 10,000-year mechanical clock, and he gave a TED Talk in 1994 that is practically prophetic. Throughout his career, Hillis has worked at places like Disney and now Applied Minds, always looking for the next fascinating problem.

The original video is available on TED.com
sponsored links

If you need translations, you can install "Google Translate" extension into your Chrome Browser.
Furthermore, you can change playback rate by installing "Video Speed Controller" extension.

Data provided by TED.

This website is owned and operated by Tokyo English Network.
The developer's blog is here.