ABOUT THE SPEAKER
Kenneth Cukier - Data Editor of The Economist
Kenneth Cukier is the Data Editor of The Economist. From 2007 to 2012 he was the Tokyo correspondent, and before that, the paper’s technology correspondent in London, where his work focused on innovation, intellectual property and Internet governance. Kenneth is also the co-author of Big Data: A Revolution That Will Transform How We Live, Work, and Think with Viktor Mayer-Schönberger in 2013, which was a New York Times Bestseller and translated into 16 languages.

Why you should listen

As Data Editor of The Economist and co-author of Big Data: A Revolution That Will Transform How We Live, Work, and Think, Kenneth Cukier has spent years immersed in big data, machine learning -- and the impact of both. What's the future of big data-driven technology and design? To find out, watch this talk.

More profile about the speaker
Kenneth Cukier | Speaker | TED.com
TEDSalon Berlin 2014

Kenneth Cukier: Big data is better data

Filmed:
1,663,038 views

Self-driving cars were just the start. What's the future of big data-driven technology and design? In a thrilling science talk, Kenneth Cukier looks at what's next for machine learning -- and human knowledge.
- Data Editor of The Economist
Kenneth Cukier is the Data Editor of The Economist. From 2007 to 2012 he was the Tokyo correspondent, and before that, the paper’s technology correspondent in London, where his work focused on innovation, intellectual property and Internet governance. Kenneth is also the co-author of Big Data: A Revolution That Will Transform How We Live, Work, and Think with Viktor Mayer-Schönberger in 2013, which was a New York Times Bestseller and translated into 16 languages. Full bio

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

00:12
America's favorite pie is?
0
787
3845
00:16
Audience: Apple.
Kenneth Cukier: Apple. Of course it is.
1
4632
3506
00:20
How do we know it?
2
8138
1231
00:21
Because of data.
3
9369
2753
00:24
You look at supermarket sales.
4
12122
2066
00:26
You look at supermarket
sales of 30-centimeter pies
5
14188
2866
00:29
that are frozen, and apple wins, no contest.
6
17054
4075
00:33
The majority of the sales are apple.
7
21129
5180
00:38
But then supermarkets started selling
8
26309
2964
00:41
smaller, 11-centimeter pies,
9
29273
2583
00:43
and suddenly, apple fell to fourth or fifth place.
10
31856
4174
00:48
Why? What happened?
11
36030
2875
00:50
Okay, think about it.
12
38905
2818
00:53
When you buy a 30-centimeter pie,
13
41723
3848
00:57
the whole family has to agree,
14
45571
2261
00:59
and apple is everyone's second favorite.
15
47832
3791
01:03
(Laughter)
16
51623
1935
01:05
But when you buy an individual 11-centimeter pie,
17
53558
3615
01:09
you can buy the one that you want.
18
57173
3745
01:12
You can get your first choice.
19
60918
4015
01:16
You have more data.
20
64933
1641
01:18
You can see something
21
66574
1554
01:20
that you couldn't see
22
68128
1132
01:21
when you only had smaller amounts of it.
23
69260
3953
01:25
Now, the point here is that more data
24
73213
2475
01:27
doesn't just let us see more,
25
75688
2283
01:29
more of the same thing we were looking at.
26
77971
1854
01:31
More data allows us to see new.
27
79825
3613
01:35
It allows us to see better.
28
83438
3094
01:38
It allows us to see different.
29
86532
3656
01:42
In this case, it allows us to see
30
90188
3173
01:45
what America's favorite pie is:
31
93361
2913
01:48
not apple.
32
96274
2542
01:50
Now, you probably all have heard the term big data.
33
98816
3614
01:54
In fact, you're probably sick of hearing the term
34
102430
2057
01:56
big data.
35
104487
1630
01:58
It is true that there is a lot of hype around the term,
36
106117
3330
02:01
and that is very unfortunate,
37
109447
2332
02:03
because big data is an extremely important tool
38
111779
3046
02:06
by which society is going to advance.
39
114825
3734
02:10
In the past, we used to look at small data
40
118559
3561
02:14
and think about what it would mean
41
122120
1704
02:15
to try to understand the world,
42
123824
1496
02:17
and now we have a lot more of it,
43
125320
1991
02:19
more than we ever could before.
44
127311
2722
02:22
What we find is that when we have
45
130033
1877
02:23
a large body of data, we can fundamentally do things
46
131910
2724
02:26
that we couldn't do when we
only had smaller amounts.
47
134634
3276
02:29
Big data is important, and big data is new,
48
137910
2641
02:32
and when you think about it,
49
140551
1777
02:34
the only way this planet is going to deal
50
142328
2216
02:36
with its global challenges —
51
144544
1789
02:38
to feed people, supply them with medical care,
52
146333
3537
02:41
supply them with energy, electricity,
53
149870
2810
02:44
and to make sure they're not burnt to a crisp
54
152680
1789
02:46
because of global warming —
55
154469
1238
02:47
is because of the effective use of data.
56
155707
4195
02:51
So what is new about big
data? What is the big deal?
57
159902
3870
02:55
Well, to answer that question, let's think about
58
163772
2517
02:58
what information looked like,
59
166289
1896
03:00
physically looked like in the past.
60
168185
3034
03:03
In 1908, on the island of Crete,
61
171219
3611
03:06
archaeologists discovered a clay disc.
62
174830
4735
03:11
They dated it from 2000 B.C., so it's 4,000 years old.
63
179565
4059
03:15
Now, there's inscriptions on this disc,
64
183624
2004
03:17
but we actually don't know what it means.
65
185628
1327
03:18
It's a complete mystery, but the point is that
66
186955
2098
03:21
this is what information used to look like
67
189053
1928
03:22
4,000 years ago.
68
190981
2089
03:25
This is how society stored
69
193070
2548
03:27
and transmitted information.
70
195618
3524
03:31
Now, society hasn't advanced all that much.
71
199142
4160
03:35
We still store information on discs,
72
203302
3474
03:38
but now we can store a lot more information,
73
206776
3184
03:41
more than ever before.
74
209960
1260
03:43
Searching it is easier. Copying it easier.
75
211220
3093
03:46
Sharing it is easier. Processing it is easier.
76
214313
3500
03:49
And what we can do is we can reuse this information
77
217813
2766
03:52
for uses that we never even imagined
78
220579
1834
03:54
when we first collected the data.
79
222413
3195
03:57
In this respect, the data has gone
80
225608
2252
03:59
from a stock to a flow,
81
227860
3532
04:03
from something that is stationary and static
82
231392
3938
04:07
to something that is fluid and dynamic.
83
235330
3609
04:10
There is, if you will, a liquidity to information.
84
238939
4023
04:14
The disc that was discovered off of Crete
85
242962
3474
04:18
that's 4,000 years old, is heavy,
86
246436
3764
04:22
it doesn't store a lot of information,
87
250200
1962
04:24
and that information is unchangeable.
88
252162
3116
04:27
By contrast, all of the files
89
255278
4011
04:31
that Edward Snowden took
90
259289
1861
04:33
from the National Security
Agency in the United States
91
261150
2621
04:35
fits on a memory stick
92
263771
2419
04:38
the size of a fingernail,
93
266190
3010
04:41
and it can be shared at the speed of light.
94
269200
4745
04:45
More data. More.
95
273945
5255
04:51
Now, one reason why we have
so much data in the world today
96
279200
1974
04:53
is we are collecting things
97
281174
1432
04:54
that we've always collected information on,
98
282606
3280
04:57
but another reason why is we're taking things
99
285886
2656
05:00
that have always been informational
100
288542
2812
05:03
but have never been rendered into a data format
101
291354
2486
05:05
and we are putting it into data.
102
293840
2419
05:08
Think, for example, the question of location.
103
296259
3308
05:11
Take, for example, Martin Luther.
104
299567
2249
05:13
If we wanted to know in the 1500s
105
301816
1597
05:15
where Martin Luther was,
106
303413
2667
05:18
we would have to follow him at all times,
107
306080
2092
05:20
maybe with a feathery quill and an inkwell,
108
308172
2137
05:22
and record it,
109
310309
1676
05:23
but now think about what it looks like today.
110
311985
2183
05:26
You know that somewhere,
111
314168
2122
05:28
probably in a telecommunications carrier's database,
112
316290
2446
05:30
there is a spreadsheet or at least a database entry
113
318736
3036
05:33
that records your information
114
321772
2088
05:35
of where you've been at all times.
115
323860
2063
05:37
If you have a cell phone,
116
325923
1360
05:39
and that cell phone has GPS,
but even if it doesn't have GPS,
117
327283
2847
05:42
it can record your information.
118
330130
2385
05:44
In this respect, location has been datafied.
119
332515
4084
05:48
Now think, for example, of the issue of posture,
120
336599
4601
05:53
the way that you are all sitting right now,
121
341200
1285
05:54
the way that you sit,
122
342485
2030
05:56
the way that you sit, the way that you sit.
123
344515
2771
05:59
It's all different, and it's a function of your leg length
124
347286
2077
06:01
and your back and the contours of your back,
125
349363
2093
06:03
and if I were to put sensors,
maybe 100 sensors
126
351456
2531
06:05
into all of your chairs right now,
127
353987
1766
06:07
I could create an index that's fairly unique to you,
128
355753
3600
06:11
sort of like a fingerprint, but it's not your finger.
129
359353
4409
06:15
So what could we do with this?
130
363762
2969
06:18
Researchers in Tokyo are using it
131
366731
2397
06:21
as a potential anti-theft device in cars.
132
369128
4388
06:25
The idea is that the carjacker sits behind the wheel,
133
373516
2924
06:28
tries to stream off, but the car recognizes
134
376440
2104
06:30
that a non-approved driver is behind the wheel,
135
378544
2362
06:32
and maybe the engine just stops, unless you
136
380906
2164
06:35
type in a password into the dashboard
137
383070
3177
06:38
to say, "Hey, I have authorization to drive." Great.
138
386247
4658
06:42
What if every single car in Europe
139
390905
2553
06:45
had this technology in it?
140
393458
1457
06:46
What could we do then?
141
394915
3165
06:50
Maybe, if we aggregated the data,
142
398080
2240
06:52
maybe we could identify telltale signs
143
400320
3814
06:56
that best predict that a car accident
144
404134
2709
06:58
is going to take place in the next five seconds.
145
406843
5893
07:04
And then what we will have datafied
146
412736
2557
07:07
is driver fatigue,
147
415293
1783
07:09
and the service would be when the car senses
148
417076
2334
07:11
that the person slumps into that position,
149
419410
3437
07:14
automatically knows, hey, set an internal alarm
150
422847
3994
07:18
that would vibrate the steering wheel, honk inside
151
426841
2025
07:20
to say, "Hey, wake up,
152
428866
1721
07:22
pay more attention to the road."
153
430587
1904
07:24
These are the sorts of things we can do
154
432491
1853
07:26
when we datafy more aspects of our lives.
155
434344
2821
07:29
So what is the value of big data?
156
437165
3675
07:32
Well, think about it.
157
440840
2190
07:35
You have more information.
158
443030
2412
07:37
You can do things that you couldn't do before.
159
445442
3341
07:40
One of the most impressive areas
160
448783
1676
07:42
where this concept is taking place
161
450459
1729
07:44
is in the area of machine learning.
162
452188
3307
07:47
Machine learning is a branch of artificial intelligence,
163
455495
3077
07:50
which itself is a branch of computer science.
164
458572
3378
07:53
The general idea is that instead of
165
461950
1543
07:55
instructing a computer what do do,
166
463493
2117
07:57
we are going to simply throw data at the problem
167
465610
2620
08:00
and tell the computer to figure it out for itself.
168
468230
3206
08:03
And it will help you understand it
169
471436
1777
08:05
by seeing its origins.
170
473213
3552
08:08
In the 1950s, a computer scientist
171
476765
2388
08:11
at IBM named Arthur Samuel liked to play checkers,
172
479153
3592
08:14
so he wrote a computer program
173
482745
1402
08:16
so he could play against the computer.
174
484147
2813
08:18
He played. He won.
175
486960
2711
08:21
He played. He won.
176
489671
2103
08:23
He played. He won,
177
491774
3015
08:26
because the computer only knew
178
494789
1778
08:28
what a legal move was.
179
496567
2227
08:30
Arthur Samuel knew something else.
180
498794
2087
08:32
Arthur Samuel knew strategy.
181
500881
4629
08:37
So he wrote a small sub-program alongside it
182
505510
2396
08:39
operating in the background, and all it did
183
507906
1974
08:41
was score the probability
184
509880
1817
08:43
that a given board configuration would likely lead
185
511697
2563
08:46
to a winning board versus a losing board
186
514260
2910
08:49
after every move.
187
517170
2508
08:51
He plays the computer. He wins.
188
519678
3150
08:54
He plays the computer. He wins.
189
522828
2508
08:57
He plays the computer. He wins.
190
525336
3731
09:01
And then Arthur Samuel leaves the computer
191
529067
2277
09:03
to play itself.
192
531344
2227
09:05
It plays itself. It collects more data.
193
533571
3509
09:09
It collects more data. It increases
the accuracy of its prediction.
194
537080
4309
09:13
And then Arthur Samuel goes back to the computer
195
541389
2104
09:15
and he plays it, and he loses,
196
543493
2318
09:17
and he plays it, and he loses,
197
545811
2069
09:19
and he plays it, and he loses,
198
547880
2047
09:21
and Arthur Samuel has created a machine
199
549927
2599
09:24
that surpasses his ability in a task that he taught it.
200
552526
6288
09:30
And this idea of machine learning
201
558814
2498
09:33
is going everywhere.
202
561312
3927
09:37
How do you think we have self-driving cars?
203
565239
3149
09:40
Are we any better off as a society
204
568388
2137
09:42
enshrining all the rules of the road into software?
205
570525
3285
09:45
No. Memory is cheaper. No.
206
573810
2598
09:48
Algorithms are faster. No. Processors are better. No.
207
576408
3994
09:52
All of those things matter, but that's not why.
208
580402
2772
09:55
It's because we changed the nature of the problem.
209
583174
3141
09:58
We changed the nature of the problem from one
210
586315
1530
09:59
in which we tried to overtly and explicitly
211
587845
2245
10:02
explain to the computer how to drive
212
590090
2581
10:04
to one in which we say,
213
592671
1316
10:05
"Here's a lot of data around the vehicle.
214
593987
1876
10:07
You figure it out.
215
595863
1533
10:09
You figure it out that that is a traffic light,
216
597396
1867
10:11
that that traffic light is red and not green,
217
599263
2081
10:13
that that means that you need to stop
218
601344
2014
10:15
and not go forward."
219
603358
3083
10:18
Machine learning is at the basis
220
606441
1518
10:19
of many of the things that we do online:
221
607959
1991
10:21
search engines,
222
609950
1857
10:23
Amazon's personalization algorithm,
223
611807
3801
10:27
computer translation,
224
615608
2212
10:29
voice recognition systems.
225
617820
4290
10:34
Researchers recently have looked at
226
622110
2835
10:36
the question of biopsies,
227
624945
3195
10:40
cancerous biopsies,
228
628140
2767
10:42
and they've asked the computer to identify
229
630907
2315
10:45
by looking at the data and survival rates
230
633222
2471
10:47
to determine whether cells are actually
231
635693
4667
10:52
cancerous or not,
232
640360
2544
10:54
and sure enough, when you throw the data at it,
233
642904
1778
10:56
through a machine-learning algorithm,
234
644682
2047
10:58
the machine was able to identify
235
646729
1877
11:00
the 12 telltale signs that best predict
236
648606
2262
11:02
that this biopsy of the breast cancer cells
237
650868
3299
11:06
are indeed cancerous.
238
654167
3218
11:09
The problem: The medical literature
239
657385
2498
11:11
only knew nine of them.
240
659883
2789
11:14
Three of the traits were ones
241
662672
1800
11:16
that people didn't need to look for,
242
664472
2975
11:19
but that the machine spotted.
243
667447
5531
11:24
Now, there are dark sides to big data as well.
244
672978
5925
11:30
It will improve our lives, but there are problems
245
678903
2074
11:32
that we need to be conscious of,
246
680977
2640
11:35
and the first one is the idea
247
683617
2623
11:38
that we may be punished for predictions,
248
686240
2686
11:40
that the police may use big data for their purposes,
249
688926
3870
11:44
a little bit like "Minority Report."
250
692796
2351
11:47
Now, it's a term called predictive policing,
251
695147
2441
11:49
or algorithmic criminology,
252
697588
2363
11:51
and the idea is that if we take a lot of data,
253
699951
2036
11:53
for example where past crimes have been,
254
701987
2159
11:56
we know where to send the patrols.
255
704146
2543
11:58
That makes sense, but the problem, of course,
256
706689
2115
12:00
is that it's not simply going to stop on location data,
257
708804
4544
12:05
it's going to go down to the level of the individual.
258
713348
2959
12:08
Why don't we use data about the person's
259
716307
2250
12:10
high school transcript?
260
718557
2228
12:12
Maybe we should use the fact that
261
720785
1561
12:14
they're unemployed or not, their credit score,
262
722346
2028
12:16
their web-surfing behavior,
263
724374
1552
12:17
whether they're up late at night.
264
725926
1878
12:19
Their Fitbit, when it's able
to identify biochemistries,
265
727804
3161
12:22
will show that they have aggressive thoughts.
266
730965
4236
12:27
We may have algorithms that are likely to predict
267
735201
2221
12:29
what we are about to do,
268
737422
1633
12:31
and we may be held accountable
269
739055
1244
12:32
before we've actually acted.
270
740299
2590
12:34
Privacy was the central challenge
271
742889
1732
12:36
in a small data era.
272
744621
2880
12:39
In the big data age,
273
747501
2149
12:41
the challenge will be safeguarding free will,
274
749650
4523
12:46
moral choice, human volition,
275
754173
3779
12:49
human agency.
276
757952
3068
12:54
There is another problem:
277
762540
2225
12:56
Big data is going to steal our jobs.
278
764765
3556
13:00
Big data and algorithms are going to challenge
279
768321
3512
13:03
white collar, professional knowledge work
280
771833
3061
13:06
in the 21st century
281
774894
1653
13:08
in the same way that factory automation
282
776547
2434
13:10
and the assembly line
283
778981
2189
13:13
challenged blue collar labor in the 20th century.
284
781170
3026
13:16
Think about a lab technician
285
784196
2092
13:18
who is looking through a microscope
286
786288
1409
13:19
at a cancer biopsy
287
787697
1624
13:21
and determining whether it's cancerous or not.
288
789321
2637
13:23
The person went to university.
289
791958
1972
13:25
The person buys property.
290
793930
1430
13:27
He or she votes.
291
795360
1741
13:29
He or she is a stakeholder in society.
292
797101
3666
13:32
And that person's job,
293
800767
1394
13:34
as well as an entire fleet
294
802161
1609
13:35
of professionals like that person,
295
803770
1969
13:37
is going to find that their jobs are radically changed
296
805739
3150
13:40
or actually completely eliminated.
297
808889
2357
13:43
Now, we like to think
298
811246
1284
13:44
that technology creates jobs over a period of time
299
812530
3187
13:47
after a short, temporary period of dislocation,
300
815717
3465
13:51
and that is true for the frame of reference
301
819182
1941
13:53
with which we all live, the Industrial Revolution,
302
821123
2142
13:55
because that's precisely what happened.
303
823265
2328
13:57
But we forget something in that analysis:
304
825593
2333
13:59
There are some categories of jobs
305
827926
1830
14:01
that simply get eliminated and never come back.
306
829756
3420
14:05
The Industrial Revolution wasn't very good
307
833176
2004
14:07
if you were a horse.
308
835180
4002
14:11
So we're going to need to be careful
309
839182
2055
14:13
and take big data and adjust it for our needs,
310
841237
3514
14:16
our very human needs.
311
844751
3185
14:19
We have to be the master of this technology,
312
847936
1954
14:21
not its servant.
313
849890
1656
14:23
We are just at the outset of the big data era,
314
851546
2958
14:26
and honestly, we are not very good
315
854504
3150
14:29
at handling all the data that we can now collect.
316
857654
4207
14:33
It's not just a problem for
the National Security Agency.
317
861861
3330
14:37
Businesses collect lots of
data, and they misuse it too,
318
865191
3038
14:40
and we need to get better at
this, and this will take time.
319
868229
3667
14:43
It's a little bit like the challenge that was faced
320
871896
1822
14:45
by primitive man and fire.
321
873718
2407
14:48
This is a tool, but this is a tool that,
322
876125
1885
14:50
unless we're careful, will burn us.
323
878010
3559
14:56
Big data is going to transform how we live,
324
884008
3120
14:59
how we work and how we think.
325
887128
2801
15:01
It is going to help us manage our careers
326
889929
1889
15:03
and lead lives of satisfaction and hope
327
891818
3634
15:07
and happiness and health,
328
895452
2992
15:10
but in the past, we've often
looked at information technology
329
898444
3306
15:13
and our eyes have only seen the T,
330
901750
2208
15:15
the technology, the hardware,
331
903958
1686
15:17
because that's what was physical.
332
905644
2262
15:19
We now need to recast our gaze at the I,
333
907906
2924
15:22
the information,
334
910830
1380
15:24
which is less apparent,
335
912210
1373
15:25
but in some ways a lot more important.
336
913583
4109
15:29
Humanity can finally learn from the information
337
917692
3465
15:33
that it can collect,
338
921157
2418
15:35
as part of our timeless quest
339
923575
2115
15:37
to understand the world and our place in it,
340
925690
3159
15:40
and that's why big data is a big deal.
341
928849
5631
15:46
(Applause)
342
934480
3568

▲Back to top

ABOUT THE SPEAKER
Kenneth Cukier - Data Editor of The Economist
Kenneth Cukier is the Data Editor of The Economist. From 2007 to 2012 he was the Tokyo correspondent, and before that, the paper’s technology correspondent in London, where his work focused on innovation, intellectual property and Internet governance. Kenneth is also the co-author of Big Data: A Revolution That Will Transform How We Live, Work, and Think with Viktor Mayer-Schönberger in 2013, which was a New York Times Bestseller and translated into 16 languages.

Why you should listen

As Data Editor of The Economist and co-author of Big Data: A Revolution That Will Transform How We Live, Work, and Think, Kenneth Cukier has spent years immersed in big data, machine learning -- and the impact of both. What's the future of big data-driven technology and design? To find out, watch this talk.

More profile about the speaker
Kenneth Cukier | Speaker | TED.com