ABOUT THE SPEAKER
Kwabena Boahen - Bioengineer
Kwabena Boahen wants to understand how brains work -- and to build a computer that works like the brain by reverse-engineering the nervous system. His group at Stanford is developing Neurogrid, a hardware platform that will emulate the cortex’s inner workings.

Why you should listen

Kwabena Boahen is the principal investigator at the Brains in Silicon lab at Stanford. He writes of himself:

Being a scientist at heart, I want to understand how cognition arises from neuronal properties. Being an engineer by training, I am using silicon integrated circuits to emulate the way neurons compute, linking the seemingly disparate fields of electronics and computer science with neurobiology and medicine.

My group's contributions to the field of neuromorphic engineering include a silicon retina that could be used to give the blind sight and a self-organizing chip that emulates the way the developing brain wires itself up. Our work is widely recognized, with over sixty publications, including a cover story in the May 2005 issue of Scientific American.

My current research interest is building a simulation platform that will enable the cortex's inner workings to be modeled in detail. While progress has been made linking neuronal properties to brain rhythms, the task of scaling up these models to link neuronal properties to cognition still remains. Making the supercomputer-performance required affordable is the goal of our Neurogrid project. It is at the vanguard of a profound shift in computing, away from the sequential, step-by-step Von Neumann machine towards a parallel, interconnected architecture more like the brain.

More profile about the speaker
Kwabena Boahen | Speaker | TED.com
TEDGlobal 2007

Kwabena Boahen: A computer that works like the brain

Filmed:
718,375 views

Researcher Kwabena Boahen is looking for ways to mimic the brain's supercomputing powers in silicon -- because the messy, redundant processes inside our heads actually make for a small, light, superfast computer.
- Bioengineer
Kwabena Boahen wants to understand how brains work -- and to build a computer that works like the brain by reverse-engineering the nervous system. His group at Stanford is developing Neurogrid, a hardware platform that will emulate the cortex’s inner workings. Full bio

Double-click the English transcript below to play the video.

00:18
I got my first computer when I was a teenager growing up in Accra,
0
0
5000
00:23
and it was a really cool device.
1
5000
3000
00:26
You could play games with it. You could program it in BASIC.
2
8000
5000
00:31
And I was fascinated.
3
13000
2000
00:33
So I went into the library to figure out how did this thing work.
4
15000
6000
00:39
I read about how the CPU is constantly shuffling data back and forth
5
21000
5000
00:44
between the memory, the RAM and the ALU,
6
26000
4000
00:48
the arithmetic and logic unit.
7
30000
2000
00:50
And I thought to myself, this CPU really has to work like crazy
8
32000
4000
00:54
just to keep all this data moving through the system.
9
36000
4000
00:58
But nobody was really worried about this.
10
40000
3000
01:01
When computers were first introduced,
11
43000
2000
01:03
they were said to be a million times faster than neurons.
12
45000
3000
01:06
People were really excited. They thought they would soon outstrip
13
48000
5000
01:11
the capacity of the brain.
14
53000
3000
01:14
This is a quote, actually, from Alan Turing:
15
56000
3000
01:17
"In 30 years, it will be as easy to ask a computer a question
16
59000
4000
01:21
as to ask a person."
17
63000
2000
01:23
This was in 1946. And now, in 2007, it's still not true.
18
65000
7000
01:30
And so, the question is, why aren't we really seeing
19
72000
4000
01:34
this kind of power in computers that we see in the brain?
20
76000
4000
01:38
What people didn't realize, and I'm just beginning to realize right now,
21
80000
4000
01:42
is that we pay a huge price for the speed
22
84000
2000
01:44
that we claim is a big advantage of these computers.
23
86000
4000
01:48
Let's take a look at some numbers.
24
90000
2000
01:50
This is Blue Gene, the fastest computer in the world.
25
92000
4000
01:54
It's got 120,000 processors; they can basically process
26
96000
5000
01:59
10 quadrillion bits of information per second.
27
101000
3000
02:02
That's 10 to the sixteenth. And they consume one and a half megawatts of power.
28
104000
7000
02:09
So that would be really great, if you could add that
29
111000
3000
02:12
to the production capacity in Tanzania.
30
114000
2000
02:14
It would really boost the economy.
31
116000
2000
02:16
Just to go back to the States,
32
118000
4000
02:20
if you translate the amount of power or electricity
33
122000
2000
02:22
this computer uses to the amount of households in the States,
34
124000
3000
02:25
you get 1,200 households in the U.S.
35
127000
4000
02:29
That's how much power this computer uses.
36
131000
2000
02:31
Now, let's compare this with the brain.
37
133000
3000
02:34
This is a picture of, actually Rory Sayres' girlfriend's brain.
38
136000
5000
02:39
Rory is a graduate student at Stanford.
39
141000
2000
02:41
He studies the brain using MRI, and he claims that
40
143000
4000
02:45
this is the most beautiful brain that he has ever scanned.
41
147000
3000
02:48
(Laughter)
42
150000
2000
02:50
So that's true love, right there.
43
152000
3000
02:53
Now, how much computation does the brain do?
44
155000
3000
02:56
I estimate 10 to the 16 bits per second,
45
158000
2000
02:58
which is actually about very similar to what Blue Gene does.
46
160000
4000
03:02
So that's the question. The question is, how much --
47
164000
2000
03:04
they are doing a similar amount of processing, similar amount of data --
48
166000
3000
03:07
the question is how much energy or electricity does the brain use?
49
169000
5000
03:12
And it's actually as much as your laptop computer:
50
174000
3000
03:15
it's just 10 watts.
51
177000
2000
03:17
So what we are doing right now with computers
52
179000
3000
03:20
with the energy consumed by 1,200 houses,
53
182000
3000
03:23
the brain is doing with the energy consumed by your laptop.
54
185000
5000
03:28
So the question is, how is the brain able to achieve this kind of efficiency?
55
190000
3000
03:31
And let me just summarize. So the bottom line:
56
193000
2000
03:33
the brain processes information using 100,000 times less energy
57
195000
4000
03:37
than we do right now with this computer technology that we have.
58
199000
4000
03:41
How is the brain able to do this?
59
203000
2000
03:43
Let's just take a look about how the brain works,
60
205000
3000
03:46
and then I'll compare that with how computers work.
61
208000
4000
03:50
So, this clip is from the PBS series, "The Secret Life of the Brain."
62
212000
4000
03:54
It shows you these cells that process information.
63
216000
3000
03:57
They are called neurons.
64
219000
1000
03:58
They send little pulses of electricity down their processes to each other,
65
220000
6000
04:04
and where they contact each other, those little pulses
66
226000
2000
04:06
of electricity can jump from one neuron to the other.
67
228000
2000
04:08
That process is called a synapse.
68
230000
3000
04:11
You've got this huge network of cells interacting with each other --
69
233000
2000
04:13
about 100 million of them,
70
235000
2000
04:15
sending about 10 quadrillion of these pulses around every second.
71
237000
4000
04:19
And that's basically what's going on in your brain right now as you're watching this.
72
241000
6000
04:25
How does that compare with the way computers work?
73
247000
2000
04:27
In the computer, you have all the data
74
249000
2000
04:29
going through the central processing unit,
75
251000
2000
04:31
and any piece of data basically has to go through that bottleneck,
76
253000
3000
04:34
whereas in the brain, what you have is these neurons,
77
256000
4000
04:38
and the data just really flows through a network of connections
78
260000
4000
04:42
among the neurons. There's no bottleneck here.
79
264000
2000
04:44
It's really a network in the literal sense of the word.
80
266000
4000
04:48
The net is doing the work in the brain.
81
270000
4000
04:52
If you just look at these two pictures,
82
274000
2000
04:54
these kind of words pop into your mind.
83
276000
2000
04:56
This is serial and it's rigid -- it's like cars on a freeway,
84
278000
4000
05:00
everything has to happen in lockstep --
85
282000
3000
05:03
whereas this is parallel and it's fluid.
86
285000
2000
05:05
Information processing is very dynamic and adaptive.
87
287000
3000
05:08
So I'm not the first to figure this out. This is a quote from Brian Eno:
88
290000
4000
05:12
"the problem with computers is that there is not enough Africa in them."
89
294000
4000
05:16
(Laughter)
90
298000
6000
05:22
Brian actually said this in 1995.
91
304000
3000
05:25
And nobody was listening then,
92
307000
3000
05:28
but now people are beginning to listen
93
310000
2000
05:30
because there's a pressing, technological problem that we face.
94
312000
5000
05:35
And I'll just take you through that a little bit in the next few slides.
95
317000
5000
05:40
This is -- it's actually really this remarkable convergence
96
322000
4000
05:44
between the devices that we use to compute in computers,
97
326000
5000
05:49
and the devices that our brains use to compute.
98
331000
4000
05:53
The devices that computers use are what's called a transistor.
99
335000
4000
05:57
This electrode here, called the gate, controls the flow of current
100
339000
4000
06:01
from the source to the drain -- these two electrodes.
101
343000
3000
06:04
And that current, electrical current,
102
346000
2000
06:06
is carried by electrons, just like in your house and so on.
103
348000
6000
06:12
And what you have here is, when you actually turn on the gate,
104
354000
5000
06:17
you get an increase in the amount of current, and you get a steady flow of current.
105
359000
4000
06:21
And when you turn off the gate, there's no current flowing through the device.
106
363000
4000
06:25
Your computer uses this presence of current to represent a one,
107
367000
5000
06:30
and the absence of current to represent a zero.
108
372000
4000
06:34
Now, what's happening is that as transistors are getting smaller and smaller and smaller,
109
376000
6000
06:40
they no longer behave like this.
110
382000
2000
06:42
In fact, they are starting to behave like the device that neurons use to compute,
111
384000
5000
06:47
which is called an ion channel.
112
389000
2000
06:49
And this is a little protein molecule.
113
391000
2000
06:51
I mean, neurons have thousands of these.
114
393000
4000
06:55
And it sits in the membrane of the cell and it's got a pore in it.
115
397000
4000
06:59
And these are individual potassium ions
116
401000
3000
07:02
that are flowing through that pore.
117
404000
2000
07:04
Now, this pore can open and close.
118
406000
2000
07:06
But, when it's open, because these ions have to line up
119
408000
5000
07:11
and flow through, one at a time, you get a kind of sporadic, not steady --
120
413000
5000
07:16
it's a sporadic flow of current.
121
418000
3000
07:19
And even when you close the pore -- which neurons can do,
122
421000
3000
07:22
they can open and close these pores to generate electrical activity --
123
424000
5000
07:27
even when it's closed, because these ions are so small,
124
429000
3000
07:30
they can actually sneak through, a few can sneak through at a time.
125
432000
3000
07:33
So, what you have is that when the pore is open,
126
435000
3000
07:36
you get some current sometimes.
127
438000
2000
07:38
These are your ones, but you've got a few zeros thrown in.
128
440000
3000
07:41
And when it's closed, you have a zero,
129
443000
4000
07:45
but you have a few ones thrown in.
130
447000
3000
07:48
Now, this is starting to happen in transistors.
131
450000
3000
07:51
And the reason why that's happening is that, right now, in 2007 --
132
453000
5000
07:56
the technology that we are using -- a transistor is big enough
133
458000
4000
08:00
that several electrons can flow through the channel simultaneously, side by side.
134
462000
5000
08:05
In fact, there's about 12 electrons can all be flowing this way.
135
467000
4000
08:09
And that means that a transistor corresponds
136
471000
2000
08:11
to about 12 ion channels in parallel.
137
473000
3000
08:14
Now, in a few years time, by 2015, we will shrink transistors so much.
138
476000
5000
08:19
This is what Intel does to keep adding more cores onto the chip.
139
481000
5000
08:24
Or your memory sticks that you have now can carry one gigabyte
140
486000
3000
08:27
of stuff on them -- before, it was 256.
141
489000
2000
08:29
Transistors are getting smaller to allow this to happen,
142
491000
3000
08:32
and technology has really benefitted from that.
143
494000
3000
08:35
But what's happening now is that in 2015, the transistor is going to become so small,
144
497000
5000
08:40
that it corresponds to only one electron at a time
145
502000
3000
08:43
can flow through that channel,
146
505000
2000
08:45
and that corresponds to a single ion channel.
147
507000
2000
08:47
And you start having the same kind of traffic jams that you have in the ion channel.
148
509000
4000
08:51
The current will turn on and off at random,
149
513000
3000
08:54
even when it's supposed to be on.
150
516000
2000
08:56
And that means your computer is going to get
151
518000
2000
08:58
its ones and zeros mixed up, and that's going to crash your machine.
152
520000
4000
09:02
So, we are at the stage where we
153
524000
4000
09:06
don't really know how to compute with these kinds of devices.
154
528000
3000
09:09
And the only kind of thing -- the only thing we know right now
155
531000
3000
09:12
that can compute with these kinds of devices are the brain.
156
534000
3000
09:15
OK, so a computer picks a specific item of data from memory,
157
537000
4000
09:19
it sends it into the processor or the ALU,
158
541000
3000
09:22
and then it puts the result back into memory.
159
544000
2000
09:24
That's the red path that's highlighted.
160
546000
2000
09:26
The way brains work, I told you all, you have got all these neurons.
161
548000
4000
09:30
And the way they represent information is
162
552000
2000
09:32
they break up that data into little pieces
163
554000
2000
09:34
that are represented by pulses and different neurons.
164
556000
3000
09:37
So you have all these pieces of data
165
559000
2000
09:39
distributed throughout the network.
166
561000
2000
09:41
And then the way that you process that data to get a result
167
563000
3000
09:44
is that you translate this pattern of activity into a new pattern of activity,
168
566000
4000
09:48
just by it flowing through the network.
169
570000
3000
09:51
So you set up these connections
170
573000
2000
09:53
such that the input pattern just flows
171
575000
3000
09:56
and generates the output pattern.
172
578000
2000
09:58
What you see here is that there's these redundant connections.
173
580000
4000
10:02
So if this piece of data or this piece of the data gets clobbered,
174
584000
4000
10:06
it doesn't show up over here, these two pieces can activate the missing part
175
588000
5000
10:11
with these redundant connections.
176
593000
2000
10:13
So even when you go to these crappy devices
177
595000
2000
10:15
where sometimes you want a one and you get a zero, and it doesn't show up,
178
597000
3000
10:18
there's redundancy in the network
179
600000
2000
10:20
that can actually recover the missing information.
180
602000
3000
10:23
It makes the brain inherently robust.
181
605000
3000
10:26
What you have here is a system where you store data locally.
182
608000
3000
10:29
And it's brittle, because each of these steps has to be flawless,
183
611000
4000
10:33
otherwise you lose that data, whereas in the brain, you have a system
184
615000
3000
10:36
that stores data in a distributed way, and it's robust.
185
618000
4000
10:40
What I want to basically talk about is my dream,
186
622000
4000
10:44
which is to build a computer that works like the brain.
187
626000
3000
10:47
This is something that we've been working on for the last couple of years.
188
629000
4000
10:51
And I'm going to show you a system that we designed
189
633000
3000
10:54
to model the retina,
190
636000
3000
10:57
which is a piece of brain that lines the inside of your eyeball.
191
639000
5000
11:02
We didn't do this by actually writing code, like you do in a computer.
192
644000
6000
11:08
In fact, the processing that happens
193
650000
3000
11:11
in that little piece of brain is very similar
194
653000
2000
11:13
to the kind of processing that computers
195
655000
1000
11:14
do when they stream video over the Internet.
196
656000
4000
11:18
They want to compress the information --
197
660000
1000
11:19
they just want to send the changes, what's new in the image, and so on --
198
661000
4000
11:23
and that is how your eyeball
199
665000
3000
11:26
is able to squeeze all that information down to your optic nerve,
200
668000
3000
11:29
to send to the rest of the brain.
201
671000
2000
11:31
Instead of doing this in software, or doing those kinds of algorithms,
202
673000
3000
11:34
we went and talked to neurobiologists
203
676000
3000
11:37
who have actually reverse engineered that piece of brain that's called the retina.
204
679000
4000
11:41
And they figured out all the different cells,
205
683000
2000
11:43
and they figured out the network, and we just took that network
206
685000
3000
11:46
and we used it as the blueprint for the design of a silicon chip.
207
688000
4000
11:50
So now the neurons are represented by little nodes or circuits on the chip,
208
692000
6000
11:56
and the connections among the neurons are represented, actually modeled by transistors.
209
698000
5000
12:01
And these transistors are behaving essentially
210
703000
2000
12:03
just like ion channels behave in the brain.
211
705000
3000
12:06
It will give you the same kind of robust architecture that I described.
212
708000
5000
12:11
Here is actually what our artificial eye looks like.
213
713000
4000
12:15
The retina chip that we designed sits behind this lens here.
214
717000
5000
12:20
And the chip -- I'm going to show you a video
215
722000
2000
12:22
that the silicon retina put out of its output
216
724000
3000
12:25
when it was looking at Kareem Zaghloul,
217
727000
3000
12:28
who's the student who designed this chip.
218
730000
2000
12:30
Let me explain what you're going to see, OK,
219
732000
2000
12:32
because it's putting out different kinds of information,
220
734000
3000
12:35
it's not as straightforward as a camera.
221
737000
2000
12:37
The retina chip extracts four different kinds of information.
222
739000
3000
12:40
It extracts regions with dark contrast,
223
742000
3000
12:43
which will show up on the video as red.
224
745000
3000
12:46
And it extracts regions with white or light contrast,
225
748000
4000
12:50
which will show up on the video as green.
226
752000
2000
12:52
This is Kareem's dark eyes
227
754000
2000
12:54
and that's the white background that you see here.
228
756000
3000
12:57
And then it also extracts movement.
229
759000
2000
12:59
When Kareem moves his head to the right,
230
761000
2000
13:01
you will see this blue activity there;
231
763000
2000
13:03
it represents regions where the contrast is increasing in the image,
232
765000
3000
13:06
that's where it's going from dark to light.
233
768000
3000
13:09
And you also see this yellow activity,
234
771000
2000
13:11
which represents regions where contrast is decreasing;
235
773000
4000
13:15
it's going from light to dark.
236
777000
2000
13:17
And these four types of information --
237
779000
3000
13:20
your optic nerve has about a million fibers in it,
238
782000
4000
13:24
and 900,000 of those fibers
239
786000
3000
13:27
send these four types of information.
240
789000
2000
13:29
So we are really duplicating the kind of signals that you have on the optic nerve.
241
791000
4000
13:33
What you notice here is that these snapshots
242
795000
3000
13:36
taken from the output of the retina chip are very sparse, right?
243
798000
4000
13:40
It doesn't light up green everywhere in the background,
244
802000
2000
13:42
only on the edges, and then in the hair, and so on.
245
804000
3000
13:45
And this is the same thing you see
246
807000
1000
13:46
when people compress video to send: they want to make it very sparse,
247
808000
4000
13:50
because that file is smaller. And this is what the retina is doing,
248
812000
3000
13:53
and it's doing it just with the circuitry, and how this network of neurons
249
815000
4000
13:57
that are interacting in there, which we've captured on the chip.
250
819000
3000
14:00
But the point that I want to make -- I'll show you up here.
251
822000
3000
14:03
So this image here is going to look like these ones,
252
825000
3000
14:06
but here I'll show you that we can reconstruct the image,
253
828000
2000
14:08
so, you know, you can almost recognize Kareem in that top part there.
254
830000
5000
14:13
And so, here you go.
255
835000
2000
14:24
Yes, so that's the idea.
256
846000
3000
14:27
When you stand still, you just see the light and dark contrasts.
257
849000
2000
14:29
But when it's moving back and forth,
258
851000
2000
14:31
the retina picks up these changes.
259
853000
3000
14:34
And that's why, you know, when you're sitting here
260
856000
1000
14:35
and something happens in your background,
261
857000
2000
14:37
you merely move your eyes to it.
262
859000
2000
14:39
There are these cells that detect change
263
861000
2000
14:41
and you move your attention to it.
264
863000
2000
14:43
So those are very important for catching somebody
265
865000
2000
14:45
who's trying to sneak up on you.
266
867000
2000
14:47
Let me just end by saying that this is what happens
267
869000
3000
14:50
when you put Africa in a piano, OK.
268
872000
3000
14:53
This is a steel drum here that has been modified,
269
875000
3000
14:56
and that's what happens when you put Africa in a piano.
270
878000
3000
14:59
And what I would like us to do is put Africa in the computer,
271
881000
4000
15:03
and come up with a new kind of computer
272
885000
2000
15:05
that will generate thought, imagination, be creative and things like that.
273
887000
3000
15:08
Thank you.
274
890000
2000
15:10
(Applause)
275
892000
2000
15:12
Chris Anderson: Question for you, Kwabena.
276
894000
2000
15:14
Do you put together in your mind the work you're doing,
277
896000
4000
15:18
the future of Africa, this conference --
278
900000
3000
15:21
what connections can we make, if any, between them?
279
903000
3000
15:24
Kwabena Boahen: Yes, like I said at the beginning,
280
906000
2000
15:26
I got my first computer when I was a teenager, growing up in Accra.
281
908000
4000
15:30
And I had this gut reaction that this was the wrong way to do it.
282
912000
4000
15:34
It was very brute force; it was very inelegant.
283
916000
3000
15:37
I don't think that I would've had that reaction,
284
919000
2000
15:39
if I'd grown up reading all this science fiction,
285
921000
3000
15:42
hearing about RD2D2, whatever it was called, and just -- you know,
286
924000
4000
15:46
buying into this hype about computers.
287
928000
1000
15:47
I was coming at it from a different perspective,
288
929000
2000
15:49
where I was bringing that different perspective
289
931000
2000
15:51
to bear on the problem.
290
933000
2000
15:53
And I think a lot of people in Africa have this different perspective,
291
935000
3000
15:56
and I think that's going to impact technology.
292
938000
2000
15:58
And that's going to impact how it's going to evolve.
293
940000
2000
16:00
And I think you're going to be able to see, use that infusion,
294
942000
2000
16:02
to come up with new things,
295
944000
2000
16:04
because you're coming from a different perspective.
296
946000
3000
16:07
I think we can contribute. We can dream like everybody else.
297
949000
4000
16:11
CA: Thanks Kwabena, that was really interesting.
298
953000
2000
16:13
Thank you.
299
955000
1000
16:14
(Applause)
300
956000
2000

▲Back to top

ABOUT THE SPEAKER
Kwabena Boahen - Bioengineer
Kwabena Boahen wants to understand how brains work -- and to build a computer that works like the brain by reverse-engineering the nervous system. His group at Stanford is developing Neurogrid, a hardware platform that will emulate the cortex’s inner workings.

Why you should listen

Kwabena Boahen is the principal investigator at the Brains in Silicon lab at Stanford. He writes of himself:

Being a scientist at heart, I want to understand how cognition arises from neuronal properties. Being an engineer by training, I am using silicon integrated circuits to emulate the way neurons compute, linking the seemingly disparate fields of electronics and computer science with neurobiology and medicine.

My group's contributions to the field of neuromorphic engineering include a silicon retina that could be used to give the blind sight and a self-organizing chip that emulates the way the developing brain wires itself up. Our work is widely recognized, with over sixty publications, including a cover story in the May 2005 issue of Scientific American.

My current research interest is building a simulation platform that will enable the cortex's inner workings to be modeled in detail. While progress has been made linking neuronal properties to brain rhythms, the task of scaling up these models to link neuronal properties to cognition still remains. Making the supercomputer-performance required affordable is the goal of our Neurogrid project. It is at the vanguard of a profound shift in computing, away from the sequential, step-by-step Von Neumann machine towards a parallel, interconnected architecture more like the brain.

More profile about the speaker
Kwabena Boahen | Speaker | TED.com