TED2014
Ray Kurzweil: Get ready for hybrid thinking
Filmed:
Readability: 4.4
3,548,296 views
Two hundred million years ago, our mammal ancestors developed a new brain feature: the neocortex. This stamp-sized piece of tissue (wrapped around a brain the size of a walnut) is the key to what humanity has become. Now, futurist Ray Kurzweil suggests, we should get ready for the next big leap in brain power, as we tap into the computing power in the cloud.
Ray Kurzweil - Inventor, futurist
Ray Kurzweil is an engineer who has radically advanced the fields of speech, text and audio technology. He's revered for his dizzying -- yet convincing -- writing on the advance of technology, the limits of biology and the future of the human species. Full bio
Ray Kurzweil is an engineer who has radically advanced the fields of speech, text and audio technology. He's revered for his dizzying -- yet convincing -- writing on the advance of technology, the limits of biology and the future of the human species. Full bio
Double-click the English transcript below to play the video.
00:12
Let me tell you a story.
0
988
2316
00:15
It goes back 200 million years.
1
3304
1799
00:17
It's a story of the neocortex,
2
5103
1984
00:19
which means "new rind."
3
7087
1974
00:21
So in these early mammals,
4
9061
2431
00:23
because only mammals have a neocortex,
5
11492
2055
00:25
rodent-like creatures.
6
13547
1664
00:27
It was the size of a postage stamp and just as thin,
7
15211
3579
00:30
and was a thin covering around
8
18790
1439
00:32
their walnut-sized brain,
9
20229
2264
00:34
but it was capable of a new type of thinking.
10
22493
3701
00:38
Rather than the fixed behaviors
11
26194
1567
00:39
that non-mammalian animals have,
12
27761
1992
00:41
it could invent new behaviors.
13
29753
2692
00:44
So a mouse is escaping a predator,
14
32445
2553
00:46
its path is blocked,
15
34998
1540
00:48
it'll try to invent a new solution.
16
36538
2129
00:50
That may work, it may not,
17
38667
1266
00:51
but if it does, it will remember that
18
39933
1910
00:53
and have a new behavior,
19
41843
1292
00:55
and that can actually spread virally
20
43135
1457
00:56
through the rest of the community.
21
44592
2195
00:58
Another mouse watching this could say,
22
46787
1609
01:00
"Hey, that was pretty clever, going around that rock,"
23
48396
2704
01:03
and it could adopt a new behavior as well.
24
51100
3725
01:06
Non-mammalian animals
25
54825
1717
01:08
couldn't do any of those things.
26
56542
1713
01:10
They had fixed behaviors.
27
58255
1215
01:11
Now they could learn a new behavior
28
59470
1331
01:12
but not in the course of one lifetime.
29
60801
2576
01:15
In the course of maybe a thousand lifetimes,
30
63377
1767
01:17
it could evolve a new fixed behavior.
31
65144
3330
01:20
That was perfectly okay 200 million years ago.
32
68474
3377
01:23
The environment changed very slowly.
33
71851
1981
01:25
It could take 10,000 years for there to be
34
73832
1554
01:27
a significant environmental change,
35
75386
2092
01:29
and during that period of time
36
77478
1382
01:30
it would evolve a new behavior.
37
78860
2929
01:33
Now that went along fine,
38
81789
1521
01:35
but then something happened.
39
83310
1704
01:37
Sixty-five million years ago,
40
85014
2246
01:39
there was a sudden, violent
change to the environment.
change to the environment.
41
87260
2615
01:41
We call it the Cretaceous extinction event.
42
89875
3505
01:45
That's when the dinosaurs went extinct,
43
93380
2293
01:47
that's when 75 percent of the
44
95673
3449
01:51
animal and plant species went extinct,
45
99122
2746
01:53
and that's when mammals
46
101868
1745
01:55
overtook their ecological niche,
47
103613
2152
01:57
and to anthropomorphize, biological evolution said,
48
105765
3654
02:01
"Hmm, this neocortex is pretty good stuff,"
49
109419
2025
02:03
and it began to grow it.
50
111444
1793
02:05
And mammals got bigger,
51
113237
1342
02:06
their brains got bigger at an even faster pace,
52
114579
2915
02:09
and the neocortex got bigger even faster than that
53
117494
3807
02:13
and developed these distinctive ridges and folds
54
121301
2929
02:16
basically to increase its surface area.
55
124230
2881
02:19
If you took the human neocortex
56
127111
1819
02:20
and stretched it out,
57
128930
1301
02:22
it's about the size of a table napkin,
58
130231
1713
02:23
and it's still a thin structure.
59
131944
1306
02:25
It's about the thickness of a table napkin.
60
133250
1980
02:27
But it has so many convolutions and ridges
61
135230
2497
02:29
it's now 80 percent of our brain,
62
137727
3075
02:32
and that's where we do our thinking,
63
140802
2461
02:35
and it's the great sublimator.
64
143263
1761
02:37
We still have that old brain
65
145024
1114
02:38
that provides our basic drives and motivations,
66
146138
2764
02:40
but I may have a drive for conquest,
67
148902
2716
02:43
and that'll be sublimated by the neocortex
68
151618
2715
02:46
into writing a poem or inventing an app
69
154333
2909
02:49
or giving a TED Talk,
70
157242
1509
02:50
and it's really the neocortex that's where
71
158751
3622
02:54
the action is.
72
162373
1968
02:56
Fifty years ago, I wrote a paper
73
164341
1717
02:58
describing how I thought the brain worked,
74
166058
1918
02:59
and I described it as a series of modules.
75
167976
3199
03:03
Each module could do things with a pattern.
76
171175
2128
03:05
It could learn a pattern. It could remember a pattern.
77
173303
2746
03:08
It could implement a pattern.
78
176049
1407
03:09
And these modules were organized in hierarchies,
79
177456
2679
03:12
and we created that hierarchy with our own thinking.
80
180135
2954
03:15
And there was actually very little to go on
81
183089
3333
03:18
50 years ago.
82
186422
1562
03:19
It led me to meet President Johnson.
83
187984
2115
03:22
I've been thinking about this for 50 years,
84
190099
2173
03:24
and a year and a half ago I came out with the book
85
192272
2828
03:27
"How To Create A Mind,"
86
195100
1265
03:28
which has the same thesis,
87
196365
1613
03:29
but now there's a plethora of evidence.
88
197978
2812
03:32
The amount of data we're getting about the brain
89
200790
1814
03:34
from neuroscience is doubling every year.
90
202604
2203
03:36
Spatial resolution of brainscanning of all types
91
204807
2654
03:39
is doubling every year.
92
207461
2285
03:41
We can now see inside a living brain
93
209746
1717
03:43
and see individual interneural connections
94
211463
2870
03:46
connecting in real time, firing in real time.
95
214333
2703
03:49
We can see your brain create your thoughts.
96
217036
2419
03:51
We can see your thoughts create your brain,
97
219455
1575
03:53
which is really key to how it works.
98
221030
1999
03:55
So let me describe briefly how it works.
99
223029
2219
03:57
I've actually counted these modules.
100
225248
2275
03:59
We have about 300 million of them,
101
227523
2046
04:01
and we create them in these hierarchies.
102
229569
2229
04:03
I'll give you a simple example.
103
231798
2082
04:05
I've got a bunch of modules
104
233880
2805
04:08
that can recognize the crossbar to a capital A,
105
236685
3403
04:12
and that's all they care about.
106
240088
1914
04:14
A beautiful song can play,
107
242002
1578
04:15
a pretty girl could walk by,
108
243580
1434
04:17
they don't care, but they see
a crossbar to a capital A,
a crossbar to a capital A,
109
245014
2846
04:19
they get very excited and they say "crossbar,"
110
247860
3021
04:22
and they put out a high probability
111
250881
2112
04:24
on their output axon.
112
252993
1634
04:26
That goes to the next level,
113
254627
1333
04:27
and these layers are organized in conceptual levels.
114
255960
2750
04:30
Each is more abstract than the next one,
115
258710
1856
04:32
so the next one might say "capital A."
116
260566
2418
04:34
That goes up to a higher
level that might say "Apple."
level that might say "Apple."
117
262984
2891
04:37
Information flows down also.
118
265875
2167
04:40
If the apple recognizer has seen A-P-P-L,
119
268042
2936
04:42
it'll think to itself, "Hmm, I
think an E is probably likely,"
think an E is probably likely,"
120
270978
3219
04:46
and it'll send a signal down to all the E recognizers
121
274197
2564
04:48
saying, "Be on the lookout for an E,
122
276761
1619
04:50
I think one might be coming."
123
278380
1556
04:51
The E recognizers will lower their threshold
124
279936
2843
04:54
and they see some sloppy
thing, could be an E.
thing, could be an E.
125
282779
1945
04:56
Ordinarily you wouldn't think so,
126
284724
1490
04:58
but we're expecting an E, it's good enough,
127
286214
2009
05:00
and yeah, I've seen an E, and then apple says,
128
288223
1817
05:02
"Yeah, I've seen an Apple."
129
290040
1728
05:03
Go up another five levels,
130
291768
1746
05:05
and you're now at a pretty high level
131
293514
1353
05:06
of this hierarchy,
132
294867
1569
05:08
and stretch down into the different senses,
133
296436
2353
05:10
and you may have a module
that sees a certain fabric,
that sees a certain fabric,
134
298789
2655
05:13
hears a certain voice quality,
smells a certain perfume,
smells a certain perfume,
135
301444
2844
05:16
and will say, "My wife has entered the room."
136
304288
2513
05:18
Go up another 10 levels, and now you're at
137
306801
1895
05:20
a very high level.
138
308696
1160
05:21
You're probably in the frontal cortex,
139
309856
1937
05:23
and you'll have modules that say, "That was ironic.
140
311793
3767
05:27
That's funny. She's pretty."
141
315560
2370
05:29
You might think that those are more sophisticated,
142
317930
2105
05:32
but actually what's more complicated
143
320035
1506
05:33
is the hierarchy beneath them.
144
321541
2669
05:36
There was a 16-year-old girl, she had brain surgery,
145
324210
2620
05:38
and she was conscious because the surgeons
146
326830
2051
05:40
wanted to talk to her.
147
328881
1537
05:42
You can do that because there's no pain receptors
148
330418
1822
05:44
in the brain.
149
332240
1038
05:45
And whenever they stimulated particular,
150
333278
1800
05:47
very small points on her neocortex,
151
335078
2463
05:49
shown here in red, she would laugh.
152
337541
2665
05:52
So at first they thought they were triggering
153
340206
1440
05:53
some kind of laugh reflex,
154
341646
1720
05:55
but no, they quickly realized they had found
155
343366
2519
05:57
the points in her neocortex that detect humor,
156
345885
3044
06:00
and she just found everything hilarious
157
348929
1969
06:02
whenever they stimulated these points.
158
350898
2437
06:05
"You guys are so funny just standing around,"
159
353335
1925
06:07
was the typical comment,
160
355260
1738
06:08
and they weren't funny,
161
356998
2302
06:11
not while doing surgery.
162
359300
3247
06:14
So how are we doing today?
163
362547
4830
06:19
Well, computers are actually beginning to master
164
367377
3054
06:22
human language with techniques
165
370431
2001
06:24
that are similar to the neocortex.
166
372432
2867
06:27
I actually described the algorithm,
167
375299
1514
06:28
which is similar to something called
168
376813
2054
06:30
a hierarchical hidden Markov model,
169
378867
2233
06:33
something I've worked on since the '90s.
170
381100
3241
06:36
"Jeopardy" is a very broad natural language game,
171
384341
3238
06:39
and Watson got a higher score
172
387579
1892
06:41
than the best two players combined.
173
389471
2000
06:43
It got this query correct:
174
391471
2499
06:45
"A long, tiresome speech
175
393970
2085
06:48
delivered by a frothy pie topping,"
176
396055
2152
06:50
and it quickly responded,
"What is a meringue harangue?"
"What is a meringue harangue?"
177
398207
2796
06:53
And Jennings and the other guy didn't get that.
178
401003
2635
06:55
It's a pretty sophisticated example of
179
403638
1926
06:57
computers actually understanding human language,
180
405564
1914
06:59
and it actually got its knowledge by reading
181
407478
1652
07:01
Wikipedia and several other encyclopedias.
182
409130
3785
07:04
Five to 10 years from now,
183
412915
2133
07:07
search engines will actually be based on
184
415048
2184
07:09
not just looking for combinations of words and links
185
417232
2794
07:12
but actually understanding,
186
420026
1914
07:13
reading for understanding the billions of pages
187
421940
2411
07:16
on the web and in books.
188
424351
2733
07:19
So you'll be walking along, and Google will pop up
189
427084
2616
07:21
and say, "You know, Mary, you expressed concern
190
429700
3081
07:24
to me a month ago that your glutathione supplement
191
432781
3019
07:27
wasn't getting past the blood-brain barrier.
192
435800
2231
07:30
Well, new research just came out 13 seconds ago
193
438031
2593
07:32
that shows a whole new approach to that
194
440624
1711
07:34
and a new way to take glutathione.
195
442335
1993
07:36
Let me summarize it for you."
196
444328
2562
07:38
Twenty years from now, we'll have nanobots,
197
446890
3684
07:42
because another exponential trend
198
450574
1627
07:44
is the shrinking of technology.
199
452201
1615
07:45
They'll go into our brain
200
453816
2370
07:48
through the capillaries
201
456186
1703
07:49
and basically connect our neocortex
202
457889
2477
07:52
to a synthetic neocortex in the cloud
203
460366
3185
07:55
providing an extension of our neocortex.
204
463551
3591
07:59
Now today, I mean,
205
467142
1578
08:00
you have a computer in your phone,
206
468720
1530
08:02
but if you need 10,000 computers for a few seconds
207
470250
2754
08:05
to do a complex search,
208
473004
1495
08:06
you can access that for a second or two in the cloud.
209
474499
3396
08:09
In the 2030s, if you need some extra neocortex,
210
477895
3095
08:12
you'll be able to connect to that in the cloud
211
480990
2273
08:15
directly from your brain.
212
483263
1648
08:16
So I'm walking along and I say,
213
484911
1543
08:18
"Oh, there's Chris Anderson.
214
486454
1363
08:19
He's coming my way.
215
487817
1525
08:21
I'd better think of something clever to say.
216
489342
2335
08:23
I've got three seconds.
217
491677
1524
08:25
My 300 million modules in my neocortex
218
493201
3097
08:28
isn't going to cut it.
219
496298
1240
08:29
I need a billion more."
220
497538
1246
08:30
I'll be able to access that in the cloud.
221
498784
3323
08:34
And our thinking, then, will be a hybrid
222
502107
2812
08:36
of biological and non-biological thinking,
223
504919
3522
08:40
but the non-biological portion
224
508441
1898
08:42
is subject to my law of accelerating returns.
225
510339
2682
08:45
It will grow exponentially.
226
513021
2239
08:47
And remember what happens
227
515260
2016
08:49
the last time we expanded our neocortex?
228
517276
2645
08:51
That was two million years ago
229
519921
1426
08:53
when we became humanoids
230
521347
1236
08:54
and developed these large foreheads.
231
522583
1594
08:56
Other primates have a slanted brow.
232
524177
2583
08:58
They don't have the frontal cortex.
233
526760
1745
09:00
But the frontal cortex is not
really qualitatively different.
really qualitatively different.
234
528505
3685
09:04
It's a quantitative expansion of neocortex,
235
532190
2743
09:06
but that additional quantity of thinking
236
534933
2703
09:09
was the enabling factor for us to take
237
537636
1779
09:11
a qualitative leap and invent language
238
539415
3346
09:14
and art and science and technology
239
542761
1967
09:16
and TED conferences.
240
544728
1454
09:18
No other species has done that.
241
546182
2131
09:20
And so, over the next few decades,
242
548313
2075
09:22
we're going to do it again.
243
550388
1760
09:24
We're going to again expand our neocortex,
244
552148
2274
09:26
only this time we won't be limited
245
554422
1756
09:28
by a fixed architecture of enclosure.
246
556178
4280
09:32
It'll be expanded without limit.
247
560458
3304
09:35
That additional quantity will again
248
563762
2243
09:38
be the enabling factor for another qualitative leap
249
566005
3005
09:41
in culture and technology.
250
569010
1635
09:42
Thank you very much.
251
570645
2054
09:44
(Applause)
252
572699
3086
ABOUT THE SPEAKER
Ray Kurzweil - Inventor, futuristRay Kurzweil is an engineer who has radically advanced the fields of speech, text and audio technology. He's revered for his dizzying -- yet convincing -- writing on the advance of technology, the limits of biology and the future of the human species.
Why you should listen
Inventor, entrepreneur, visionary, Ray Kurzweil's accomplishments read as a startling series of firsts -- a litany of technological breakthroughs we've come to take for granted. Kurzweil invented the first optical character recognition (OCR) software for transforming the written word into data, the first print-to-speech software for the blind, the first text-to-speech synthesizer, and the first music synthesizer capable of recreating the grand piano and other orchestral instruments, and the first commercially marketed large-vocabulary speech recognition.
Yet his impact as a futurist and philosopher is no less significant. In his best-selling books, which include How to Create a Mind, The Age of Spiritual Machines, The Singularity Is Near: When Humans Transcend Biology, Kurzweil depicts in detail a portrait of the human condition over the next few decades, as accelerating technologies forever blur the line between human and machine.
In 2009, he unveiled Singularity University, an institution that aims to "assemble, educate and inspire leaders who strive to understand and facilitate the development of exponentially advancing technologies." He is a Director of Engineering at Google, where he heads up a team developing machine intelligence and natural language comprehension.
Ray Kurzweil | Speaker | TED.com