ABOUT THE SPEAKERS
Sebastian Thrun - Educator, entrepreneur
Sebastian Thrun is a passionate technologist who is constantly looking for new opportunities to make the world better for all of us.

Why you should listen

Sebastian Thrun is an educator, entrepreneur and troublemaker. After a long life as a professor at Stanford University, Thrun resigned from tenure to join Google. At Google, he founded Google X, home to self-driving cars and many other moonshot technologies. Thrun also founded Udacity, an online university with worldwide reach, and Kitty Hawk, a "flying car" company. He has authored 11 books, 400 papers, holds 3 doctorates and has won numerous awards.

More profile about the speaker
Sebastian Thrun | Speaker | TED.com
Chris Anderson - TED Curator
After a long career in journalism and publishing, Chris Anderson became the curator of the TED Conference in 2002 and has developed it as a platform for identifying and disseminating ideas worth spreading.

Why you should listen

Chris Anderson is the Curator of TED, a nonprofit devoted to sharing valuable ideas, primarily through the medium of 'TED Talks' -- short talks that are offered free online to a global audience.

Chris was born in a remote village in Pakistan in 1957. He spent his early years in India, Pakistan and Afghanistan, where his parents worked as medical missionaries, and he attended an American school in the Himalayas for his early education. After boarding school in Bath, England, he went on to Oxford University, graduating in 1978 with a degree in philosophy, politics and economics.

Chris then trained as a journalist, working in newspapers and radio, including two years producing a world news service in the Seychelles Islands.

Back in the UK in 1984, Chris was captivated by the personal computer revolution and became an editor at one of the UK's early computer magazines. A year later he founded Future Publishing with a $25,000 bank loan. The new company initially focused on specialist computer publications but eventually expanded into other areas such as cycling, music, video games, technology and design, doubling in size every year for seven years. In 1994, Chris moved to the United States where he built Imagine Media, publisher of Business 2.0 magazine and creator of the popular video game users website IGN. Chris eventually merged Imagine and Future, taking the combined entity public in London in 1999, under the Future name. At its peak, it published 150 magazines and websites and employed 2,000 people.

This success allowed Chris to create a private nonprofit organization, the Sapling Foundation, with the hope of finding new ways to tackle tough global issues through media, technology, entrepreneurship and, most of all, ideas. In 2001, the foundation acquired the TED Conference, then an annual meeting of luminaries in the fields of Technology, Entertainment and Design held in Monterey, California, and Chris left Future to work full time on TED.

He expanded the conference's remit to cover all topics, including science, business and key global issues, while adding a Fellows program, which now has some 300 alumni, and the TED Prize, which grants its recipients "one wish to change the world." The TED stage has become a place for thinkers and doers from all fields to share their ideas and their work, capturing imaginations, sparking conversation and encouraging discovery along the way.

In 2006, TED experimented with posting some of its talks on the Internet. Their viral success encouraged Chris to begin positioning the organization as a global media initiative devoted to 'ideas worth spreading,' part of a new era of information dissemination using the power of online video. In June 2015, the organization posted its 2,000th talk online. The talks are free to view, and they have been translated into more than 100 languages with the help of volunteers from around the world. Viewership has grown to approximately one billion views per year.

Continuing a strategy of 'radical openness,' in 2009 Chris introduced the TEDx initiative, allowing free licenses to local organizers who wished to organize their own TED-like events. More than 8,000 such events have been held, generating an archive of 60,000 TEDx talks. And three years later, the TED-Ed program was launched, offering free educational videos and tools to students and teachers.

More profile about the speaker
Chris Anderson | Speaker | TED.com
TED2017

Sebastian Thrun and Chris Anderson: What AI is -- and isn't

塞巴斯蒂安·斯伦 和 克里斯·安德森: 新一代计算机会自己编程

Filmed:
1,575,780 views

教育家和企业家塞巴斯蒂安·斯伦希望我们利用人工智能使人类从重复性工作中解放出来,去发挥我们的创新力。 在与TED策展人克里斯·安德森进行的一次激发灵感、信息庞大的谈话中,斯伦讨论了深度学习的进展,为什么我们不应该害怕人工智能失控,以及如果机器帮助我们完成枯燥冗长的工作,那么社会将变得更美好。“有趣的东西才只有1%被发明了出来,”斯伦说,“我相信每个人都有不可估量的创新力......(人工智能)会赋予我们把创意转化为行动的力量。”
- Educator, entrepreneur
Sebastian Thrun is a passionate technologist who is constantly looking for new opportunities to make the world better for all of us. Full bio - TED Curator
After a long career in journalism and publishing, Chris Anderson became the curator of the TED Conference in 2002 and has developed it as a platform for identifying and disseminating ideas worth spreading. Full bio

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

克里斯·安德森(CA):
给我们讲讲机器学习是什么,
00:12
Chris克里斯 Anderson安德森: Help us understand理解
what machine learning学习 is,
0
904
2886
它似乎是一个关键动力,
00:15
because that seems似乎 to be the key driver司机
1
3814
2054
驱动着很多让人兴奋的事,
还有围绕着人工智能的
00:17
of so much of the excitement激动
and also of the concern关心
2
5892
2737
那么多关注。
00:20
around artificial人造 intelligence情报.
3
8653
1494
机器学习到底是怎么工作的?
00:22
How does machine learning学习 work?
4
10171
1643
塞巴斯蒂安·斯伦(ST):
人工智能和机器学习
00:23
Sebastian塞巴斯蒂安 Thrun史朗: So, artificial人造
intelligence情报 and machine learning学习
5
11838
3896
已经有60年的历史了,
00:27
is about 60 years年份 old
6
15758
2002
直到最近才初露锋芒。
00:29
and has not had a great day
in its past过去 until直到 recently最近.
7
17784
4269
原因在于,当今,
00:34
And the reason原因 is that today今天,
8
22077
2924
我们的计算和数据集的规模
00:37
we have reached到达 a scale规模
of computing计算 and datasets数据集
9
25025
3973
已经达到了机器智能化
所必需的水平。
00:41
that was necessary必要 to make machines smart聪明.
10
29022
2637
它的工作原理是这样的。
00:43
So here's这里的 how it works作品.
11
31683
1751
假设今天你想编写一个计算机程序,
给自己打造一部智能手机,
00:45
If you program程序 a computer电脑 today今天,
say, your phone电话,
12
33458
3497
那么你会聘请软件工程师
00:48
then you hire聘请 software软件 engineers工程师
13
36979
2335
编写很长很长的(类似)烹饪食谱,
00:51
that write a very,
very long kitchen厨房 recipe食谱,
14
39338
3854
比如“如果水太热,请调低温度,
00:55
like, "If the water is too hot,
turn down the temperature温度.
15
43216
3132
如果太凉,调高温度。”
00:58
If it's too cold, turn up
the temperature温度."
16
46372
2279
但我们的食谱不只是10行。
01:00
The recipes食谱 are not just 10 lines线 long.
17
48675
2849
它们长达数百万行。
01:03
They are millions百万 of lines线 long.
18
51548
2603
01:06
A modern现代 cell细胞 phone电话
has 12 million百万 lines线 of code.
19
54175
4084
现代手机拥有1200万行代码。
一个浏览器有500万行代码。
01:10
A browser浏览器 has five million百万 lines线 of code.
20
58283
2646
而这个食谱中的每个错误
都可能导致你的电脑崩溃。
01:12
And each bug窃听器 in this recipe食谱
can cause原因 your computer电脑 to crash紧急.
21
60953
4969
这就是为什么软件工程师
赚那么多钱。
01:17
That's why a software软件 engineer工程师
makes品牌 so much money.
22
65946
3075
现在的新现象是
电脑可以找到自己的规则。
01:21
The new thing now is that computers电脑
can find their own拥有 rules规则.
23
69953
3660
所以不再是专家一步一步地
01:25
So instead代替 of an expert专家
deciphering破译, step by step,
24
73637
3606
为每一个偶然事件破译出规则,
01:29
a rule规则 for every一切 contingency偶然性,
25
77267
2148
而现在的做法是
给计算机提供实例,
01:31
what you do now is you give
the computer电脑 examples例子
26
79439
3074
让计算机推断出自己的规则。
01:34
and have it infer推断 its own拥有 rules规则.
27
82537
1581
一个很好的例子是谷歌
刚获胜的阿尔法围棋。
01:36
A really good example is AlphaGoAlphaGo,
which哪一个 recently最近 was won韩元 by Google谷歌.
28
84142
4306
通常,在比赛中,
你会真的写下全部规则,
01:40
Normally一般, in game游戏 playing播放,
you would really write down all the rules规则,
29
88472
3687
而对于阿尔法围棋,
01:44
but in AlphaGo'sAlphaGo的 case案件,
30
92183
1785
它的系统观摩了
超过一百万次的比赛,
01:45
the system系统 looked看着 over a million百万 games游戏
31
93992
2066
并且能够推断出自己的规则,
01:48
and was able能够 to infer推断 its own拥有 rules规则
32
96082
2192
然后击败了当下的世界围棋冠军。
01:50
and then beat击败 the world's世界
residing居住 Go champion冠军.
33
98298
2738
这很令人兴奋,因为它不再需要
01:53
That is exciting扣人心弦, because it relieves可以减压
the software软件 engineer工程师
34
101853
3509
软件工程师必须超级聪明,
01:57
of the need of being存在 super smart聪明,
35
105386
1819
而是把负担推到数据上。
01:59
and pushes the burden负担 towards the data数据.
36
107229
2325
如我所说,这个转折点
已经真正成为可能——
02:01
As I said, the inflection拐点 point
where this has become成为 really possible可能 --
37
109578
4534
非常尴尬,我的论文
是关于机器学习的。
02:06
very embarrassing尴尬, my thesis论文
was about machine learning学习.
38
114136
2746
它的内容完全不重要,千万别读,
02:08
It was completely全然
insignificant微不足道, don't read it,
39
116906
2205
因为那是20年前的,
02:11
because it was 20 years年份 ago
40
119135
1350
那时候,计算机只有蟑螂脑袋的容量。
02:12
and back then, the computers电脑
were as big as a cockroach蟑螂 brain.
41
120509
2907
现在,计算机足够强大,
02:15
Now they are powerful强大 enough足够
to really emulate仿真
42
123440
2331
可以真正模仿特定的人类思维。
02:17
kind of specialized专门 human人的 thinking思维.
43
125795
2076
然后计算机借助一个事实,
02:19
And then the computers电脑
take advantage优点 of the fact事实
44
127895
2313
它们能读取的数据比人类多得多。
02:22
that they can look at
much more data数据 than people can.
45
130232
2500
所以我会说阿尔法围棋
看了上百万次比赛。
02:24
So I'd say AlphaGoAlphaGo looked看着 at
more than a million百万 games游戏.
46
132756
3080
没有人类专家可以
研究一百万次比赛。
02:27
No human人的 expert专家 can ever
study研究 a million百万 games游戏.
47
135860
2839
谷歌已经浏览了超过千亿的网页。
02:30
Google谷歌 has looked看着 at over
a hundred billion十亿 web卷筒纸 pages网页.
48
138723
3182
也没有人类可以做到这一点。
02:33
No person can ever study研究
a hundred billion十亿 web卷筒纸 pages网页.
49
141929
2650
因此,电脑能找到
02:36
So as a result结果,
the computer电脑 can find rules规则
50
144603
2714
连人类也找不到的规则。
02:39
that even people can't find.
51
147341
1755
CA:所以它思考的不是
“如果他那么走,我要那么走,”
02:41
CACA: So instead代替 of looking ahead
to, "If he does that, I will do that,"
52
149120
4312
而更像是“这应该是获胜模式,
02:45
it's more saying, "Here is what
looks容貌 like a winning胜利 pattern模式,
53
153456
3072
这也应该是获胜模式。”
02:48
here is what looks容貌 like
a winning胜利 pattern模式."
54
156552
2079
ST:对,想想如何抚养孩子。
02:50
STST: Yeah. I mean, think about
how you raise提高 children孩子.
55
158655
2517
你不会把前18年用来
给孩子创建每个细节的规则,
02:53
You don't spend the first 18 years年份
giving kids孩子 a rule规则 for every一切 contingency偶然性
56
161196
3644
再放他们出去,
那他们就麻烦大了。
02:56
and set them free自由
and they have this big program程序.
57
164864
2347
孩子会摸爬滚打,跌倒再站起来,
他们失败、受挫,
02:59
They stumble绊倒, fall秋季, get up,
they get slapped耳光 or spanked打屁股,
58
167235
2719
他们获得正面的经验,
在学校里取得好成绩,
03:01
and they have a positive experience经验,
a good grade年级 in school学校,
59
169978
2884
然后他们自己摸索出人生。
03:04
and they figure数字 it out on their own拥有.
60
172886
1834
现在这也发生在计算机上,
03:06
That's happening事件 with computers电脑 now,
61
174744
1737
所以突然间计算机编程容易多了。
03:08
which哪一个 makes品牌 computer电脑 programming程序设计
so much easier更轻松 all of a sudden突然.
62
176505
3029
现在不用我们思考了。
我们只要给计算机大量的数据。
03:11
Now we don't have to think anymore.
We just give them lots of data数据.
63
179558
3175
CA:所以,这才是
自动驾驶汽车的影响
03:14
CACA: And so, this has been key
to the spectacular壮观 improvement起色
64
182757
3422
大幅提升的关键。
03:18
in power功率 of self-driving自驾车 cars汽车.
65
186203
3064
我记得你给我举例了。
03:21
I think you gave me an example.
66
189291
1739
能解释下这是个什么场景吗?
03:23
Can you explain说明 what's happening事件 here?
67
191054
2685
ST:这是一个
自动驾驶汽车的行驶过程,
03:25
STST: This is a drive驾驶 of a self-driving自驾车 car汽车
68
193763
3564
这刚好是我们优达学城的车,
03:29
that we happened发生 to have at UdacityUdacity
69
197351
1957
最近做成名叫Voyage的改装车。
03:31
and recently最近 made制作
into a spin-off分拆 called Voyage航程.
70
199332
2398
我们一直用“深度学习”
03:33
We have used this thing
called deep learning学习
71
201754
2574
来训练一辆汽车自行驾驶,
03:36
to train培养 a car汽车 to drive驾驶 itself本身,
72
204352
1623
这是在一个雨天从加州的
03:37
and this is driving主动
from Mountain View视图, California加州,
73
205999
2387
山景城出发开到旧金山,
03:40
to San Francisco弗朗西斯科
74
208410
1168
行驶在El Camino Real路上,
03:41
on El萨尔瓦多 Camino卡米诺 Real真实 on a rainy多雨的 day,
75
209602
2259
路上有人骑车,有人步行,
有133个交通信号灯。
03:43
with bicyclists骑自行车 and pedestrians行人
and 133 traffic交通 lights灯火.
76
211885
3524
这里的创新点是,
03:47
And the novel小说 thing here is,
77
215433
2636
03:50
many许多, many许多 moons月亮 ago, I started开始
the Google谷歌 self-driving自驾车 car汽车 team球队.
78
218093
3120
很久以前我组建了
谷歌自动驾驶团队。
那时,我聘请了
世界上最好的软件工程师
03:53
And back in the day, I hired雇用
the world's世界 best最好 software软件 engineers工程师
79
221237
3181
来寻找世界上最好的规则。
03:56
to find the world's世界 best最好 rules规则.
80
224442
1607
而这只是训练出来的。
03:58
This is just trained熟练.
81
226073
1754
我们在这条路上跑上个20次,
03:59
We drive驾驶 this road 20 times,
82
227851
3336
把所有数据放到电脑里,
04:03
we put all this data数据
into the computer电脑 brain,
83
231211
2447
经过几个小时的处理,
04:05
and after a few少数 hours小时 of processing处理,
84
233682
2082
它创造出的行为
常常超越人类的操作能力。
04:07
it comes up with behavior行为
that often经常 surpasses超过 human人的 agility敏捷.
85
235788
3926
所以对它进行编程变得非常简单。
04:11
So it's become成为 really easy简单 to program程序 it.
86
239738
2017
这是百分之百自主操作,
大约33英里,一个半小时。
04:13
This is 100 percent百分 autonomous自主性,
about 33 miles英里, an hour小时 and a half.
87
241779
3803
CA:那么,详细说说——
这个程序的左边这一大块,
04:17
CACA: So, explain说明 it -- on the big part部分
of this program程序 on the left,
88
245606
3630
我们看到的基本上就是
电脑看到的卡车和轿车,
04:21
you're seeing眼看 basically基本上 what
the computer电脑 sees看到 as trucks卡车 and cars汽车
89
249260
3257
各种超车的亮点,等等。
04:24
and those dots overtaking超车 it and so forth向前.
90
252541
2886
ST:右侧是摄像机图像,
在这里是主要输入,
04:27
STST: On the right side, you see the camera相机
image图片, which哪一个 is the main主要 input输入 here,
91
255451
3762
用来找车道、其他车辆,
交通信号灯。
04:31
and it's used to find lanes车道,
other cars汽车, traffic交通 lights灯火.
92
259237
2676
这辆车有雷达来做测距。
04:33
The vehicle车辆 has a radar雷达
to do distance距离 estimation估计.
93
261937
2489
这在类似系统里很常见。
04:36
This is very commonly常用 used
in these kind of systems系统.
94
264450
2621
左侧的是激光图,
04:39
On the left side you see a laser激光 diagram,
95
267095
1992
可以看到激光绘制的
树木等障碍物。
04:41
where you see obstacles障碍 like trees树木
and so on depicted描绘 by the laser激光.
96
269111
3200
但是现在几乎所有有趣的工作
都集中在相机图像上。
04:44
But almost几乎 all the interesting有趣 work
is centering定心 on the camera相机 image图片 now.
97
272335
3436
我们正在从雷达和激光等
精密传感器
04:47
We're really shifting over from precision精确
sensors传感器 like radars雷达 and lasers激光器
98
275795
3476
转向非常便宜、商品化的传感器。
04:51
into very cheap低廉, commoditized商品化 sensors传感器.
99
279295
1842
成本低于8美元的相机。
04:53
A camera相机 costs成本 less than eight dollars美元.
100
281161
1987
CA:左边那个绿色的圆点是什么?
04:55
CACA: And that green绿色 dot
on the left thing, what is that?
101
283172
2793
它有什么意义吗?
04:57
Is that anything meaningful富有意义的?
102
285989
1371
ST:这是自适应巡航控制的先行点,
04:59
STST: This is a look-ahead展望 point
for your adaptive自适应 cruise巡航 control控制,
103
287384
3668
它可以帮助我们了解
05:03
so it helps帮助 us understand理解
how to regulate调节 velocity速度
104
291076
2477
如何根据车前方的距离来调节速度。
05:05
based基于 on how far
the cars汽车 in front面前 of you are.
105
293577
2634
CA:那么,我想你还有一个例子
05:08
CACA: And so, you've also
got an example, I think,
106
296235
2716
是说明实际的学习部分
是如何发生的。
05:10
of how the actual实际
learning学习 part部分 takes place地点.
107
298975
2381
也许我们可以看那个例子,
来谈谈这个话题。
05:13
Maybe we can see that. Talk about this.
108
301380
2458
ST:这个示例是我们
向优达学城的学生们发起的挑战,
05:15
STST: This is an example where we posed构成
a challenge挑战 to UdacityUdacity students学生们
109
303862
3643
用于获得我们的
自动驾驶“纳米”学位。
05:19
to take what we call
a self-driving自驾车 car汽车 NanodegreeNanodegree.
110
307529
3131
我们给他们提供这个数据集,
05:22
We gave them this dataset数据集
111
310684
1495
说:“嘿,你们能不能
找到这辆车的驾驶方法?”
05:24
and said "Hey, can you guys figure数字 out
how to steer驾驶 this car汽车?"
112
312203
3054
如果你观看图像,
05:27
And if you look at the images图片,
113
315281
1624
即使对于人类,也不大可能完美转向。
05:28
it's, even for humans人类, quite相当 impossible不可能
to get the steering操舵 right.
114
316929
4073
我们办了一场竞赛,说:
“这是深度学习竞赛,
05:33
And we ran a competition竞争 and said,
"It's a deep learning学习 competition竞争,
115
321026
3591
人工智能竞赛,”
05:36
AIAI competition竞争,"
116
324641
1173
我们给了学生48小时。
05:37
and we gave the students学生们 48 hours小时.
117
325838
1887
如果你是像谷歌或脸书
那样的软件公司,
05:39
So if you are a software软件 house
like Google谷歌 or FacebookFacebook的,
118
327749
4172
那么这样的工作至少要花费
六个月的时间。
05:43
something like this costs成本 you
at least最小 six months个月 of work.
119
331945
2717
所以我们认为48小时
就能解决问题简直太赞了。
05:46
So we figured想通 48 hours小时 is great.
120
334686
2202
在48小时内,我们收到了
大约100份学生交稿,
05:48
And within 48 hours小时, we got about
100 submissions提交 from students学生们,
121
336912
3467
前四名的答案完全正确。
05:52
and the top最佳 four got it perfectly完美 right.
122
340403
3370
它比我在这个情景中驾驶得更好,
05:55
It drives驱动器 better than I could
drive驾驶 on this imagery意象,
123
343797
2640
它用的是深度学习。
05:58
using运用 deep learning学习.
124
346461
1189
重申,方法是一样的。
05:59
And again, it's the same相同 methodology方法.
125
347674
1799
就是这个神奇的东西。
06:01
It's this magical神奇 thing.
126
349497
1164
现在如果给计算机
提供足够的数据,
06:02
When you give enough足够 data数据
to a computer电脑 now,
127
350685
2085
并且给它足够的时间来理解数据,
06:04
and give enough足够 time
to comprehend理解 the data数据,
128
352794
2140
它总会找到自己的规则。
06:06
it finds认定 its own拥有 rules规则.
129
354958
1445
06:09
CACA: And so that has led to the development发展
of powerful强大 applications应用
130
357339
4845
CA:那么它已经引起了
开发各种领域的强大应用程序。
06:14
in all sorts排序 of areas.
131
362208
1525
前几天你跟我说过癌症。
06:15
You were talking to me
the other day about cancer癌症.
132
363757
2668
我可以展示这个视频吗?
06:18
Can I show显示 this video视频?
133
366449
1189
ST:当然可以,请便。
CA:这个很酷。
06:19
STST: Yeah, absolutely绝对, please.
CACA: This is cool.
134
367662
2354
ST:这是在一个完全不同的领域
06:22
STST: This is kind of an insight眼光
into what's happening事件
135
370040
3534
洞察所发生的事。
06:25
in a completely全然 different不同 domain.
136
373598
2429
这是增强或竞争——
06:28
This is augmenting增广, or competing竞争 --
137
376051
3752
在旁观者的眼中——
06:31
it's in the eye of the beholder旁观者 --
138
379827
1749
06:33
with people who are being存在 paid支付
400,000 dollars美元 a year,
139
381600
3454
与每年拿40万美元的人、
皮肤科医生、
06:37
dermatologists皮肤科医生,
140
385078
1237
训练有素的专家的竞争。
06:38
highly高度 trained熟练 specialists专家.
141
386339
1983
需要十多年的培训才能
成为一名优秀的皮肤科医生。
06:40
It takes more than a decade of training训练
to be a good dermatologist皮肤科医生.
142
388346
3561
你在这里看到的是
它的机器学习版本。
06:43
What you see here is
the machine learning学习 version of it.
143
391931
3196
它叫做神经网络。
06:47
It's called a neural神经 network网络.
144
395151
1841
“神经网络”是这些
机器学习算法的技术术语。
06:49
"Neural神经 networks网络" is the technical技术 term术语
for these machine learning学习 algorithms算法.
145
397016
3742
20世纪80年代就有了。
06:52
They've他们已经 been around since以来 the 1980s.
146
400782
1789
而这个是1988年由脸书研究员
扬·勒丘恩发明的,
06:54
This one was invented发明 in 1988
by a FacebookFacebook的 Fellow同伴 called Yann LeCunLeCun,
147
402595
4640
它传送数据的方式
06:59
and it propagates传播 data数据 stages阶段
148
407259
3558
跟人脑分段的工作方式很相似。
07:02
through通过 what you could think of
as the human人的 brain.
149
410841
2578
不完全一样,但它模仿人脑。
07:05
It's not quite相当 the same相同 thing,
but it emulates模拟 the same相同 thing.
150
413443
2966
一个阶段一个阶段地运行。
07:08
It goes stage阶段 after stage阶段.
151
416433
1302
在第一阶段,它获取视觉输入
07:09
In the very first stage阶段, it takes
the visual视觉 input输入 and extracts提取物 edges边缘
152
417759
3637
07:13
and rods and dots.
153
421420
2612
并提取边、线、点。
下一阶段变成更复杂的边
07:16
And the next下一个 one becomes
more complicated复杂 edges边缘
154
424056
3037
以及小半月之类的形状。
07:19
and shapes形状 like little half-moons半月形.
155
427117
3191
最终,它能够构建
非常复杂的概念。
07:22
And eventually终于, it's able能够 to build建立
really complicated复杂 concepts概念.
156
430332
4443
吴恩达已经能证明
07:26
Andrew安德鲁 Ng has been able能够 to show显示
157
434799
2048
它能够在大量的图像中
07:28
that it's able能够 to find
cat faces面孔 and dog faces面孔
158
436871
3480
找出猫脸和狗脸。
07:32
in vast广大 amounts of images图片.
159
440375
1661
我在斯坦福大学的学生团队展示了
07:34
What my student学生 team球队
at Stanford斯坦福 has shown显示 is that
160
442060
2724
如果用12.9万张展示皮肤状况的
图片对它进行训练,
07:36
if you train培养 it on 129,000 images图片
of skin皮肤 conditions条件,
161
444808
6073
包括黑色素瘤和癌症,
07:42
including包含 melanoma黑色素瘤 and carcinomas,
162
450905
2565
07:45
you can do as good a job工作
163
453494
3301
那么你就可以像最好的
人类皮肤科医生一样工作。
07:48
as the best最好 human人的 dermatologists皮肤科医生.
164
456819
2197
为了证明这是真的,
07:51
And to convince说服 ourselves我们自己
that this is the case案件,
165
459040
2549
我们找到一个独立的数据组,
07:53
we captured捕获 an independent独立 dataset数据集
that we presented呈现 to our network网络
166
461613
3990
展示给我们的网络以及
25位认证的斯坦福级别皮肤医生,
07:57
and to 25 board-certified认证资格
Stanford-level斯坦福级 dermatologists皮肤科医生,
167
465627
4342
然后比较结果。
08:01
and compared相比 those.
168
469993
1672
多数情况下,
08:03
And in most cases,
169
471689
1504
它们表现出的分类准确率
08:05
they were either on par平价 or above以上
the performance性能 classification分类 accuracy准确性
170
473217
3875
等同或高于人类皮肤科医生。
08:09
of human人的 dermatologists皮肤科医生.
171
477116
1467
CA:你给我讲过一个故事。
08:10
CACA: You were telling告诉 me an anecdote轶事.
172
478607
1746
我正在这里想这个画面。
08:12
I think about this image图片 right here.
173
480377
1957
这背后的故事是什么?
08:14
What happened发生 here?
174
482358
1484
ST:这是上个星期四的事儿。
挺让人激动的。
08:15
STST: This was last Thursday星期四.
That's a moving移动 piece.
175
483866
4008
我们之前展示过,并且今年早些时候
在“自然”杂志上发表了的
08:19
What we've我们已经 shown显示 before and we published发表
in "Nature性质" earlier this year
176
487898
3600
想法是,我们同时给皮肤科医生
08:23
was this idea理念 that we show显示
dermatologists皮肤科医生 images图片
177
491522
2484
和计算机程序看图片,
08:26
and our computer电脑 program程序 images图片,
178
494030
1539
然后统计正确率。
08:27
and count计数 how often经常 they're right.
179
495593
1627
但所有图片都是用过的。
08:29
But all these images图片 are past过去 images图片.
180
497244
1778
那些图片都做过活检,
以确保我们分类正确。
08:31
They've他们已经 all been biopsied活检 to make sure
we had the correct正确 classification分类.
181
499046
3460
但这一个不是。
08:34
This one wasn't.
182
502530
1172
这个实际上是我们在斯坦福
一个合作人得到的照片。
08:35
This one was actually其实 doneDONE at Stanford斯坦福
by one of our collaborators合作者.
183
503726
3179
故事是,我们的这位合作人
08:38
The story故事 goes that our collaborator合作者,
184
506929
2314
是世界著名的皮肤科医生,
显然是最好的三位之一,
08:41
who is a world-famous世界知名 dermatologist皮肤科医生,
one of the three best最好, apparently显然地,
185
509267
3391
他看着这个痣,说:
“这不是皮肤癌。”
08:44
looked看着 at this mole and said,
"This is not skin皮肤 cancer癌症."
186
512682
2935
然后他犹豫了一下,他说:
08:47
And then he had
a second第二 moment时刻, where he said,
187
515641
2476
“等等,让我用应用程序查查。”
08:50
"Well, let me just check with the app应用."
188
518141
1866
于是,他拿出自己的iPhone,
打开我们的软件,
08:52
So he took out his iPhone苹果手机
and ran our piece of software软件,
189
520031
2699
就是我们的“口袋皮肤医生”,
08:54
our "pocket口袋 dermatologist皮肤科医生," so to speak说话,
190
522754
2121
iPhone说:癌症。
08:56
and the iPhone苹果手机 said: cancer癌症.
191
524899
2994
它说是黑色素瘤。
08:59
It said melanoma黑色素瘤.
192
527917
1306
然后他就纠结了。
09:01
And then he was confused困惑.
193
529849
1233
他决定,“好吧,也许我
信iPhone比信自己多一点,”
09:03
And he decided决定, "OK, maybe I trust相信
the iPhone苹果手机 a little bit more than myself,"
194
531106
4551
于是把样品送到实验室进行活检。
09:07
and he sent发送 it out to the lab实验室
to get it biopsied活检.
195
535681
2735
结果它是侵略性的黑色素瘤。
09:10
And it came来了 up as an aggressive侵略性 melanoma黑色素瘤.
196
538440
2469
所以我觉得这可能是
我们第一次真正发现,
09:13
So I think this might威力 be the first time
that we actually其实 found发现,
197
541545
3067
在使用深度学习的实践中,
09:16
in the practice实践 of using运用 deep learning学习,
198
544636
2487
如果没有深度学习,
真会有人长了黑色素瘤
09:19
an actual实际 person whose谁的 melanoma黑色素瘤
would have gone走了 unclassified,
199
547147
3372
09:22
had it not been for deep learning学习.
200
550543
2115
却识别不出。
CA:真不可思议。
09:24
CACA: I mean, that's incredible难以置信.
201
552682
1560
09:26
(Applause掌声)
202
554266
1769
(掌声)
感觉现在就对这样的应用程序
有即时需求了,
09:28
It feels感觉 like there'd这红色 be an instant瞬间 demand需求
for an app应用 like this right now,
203
556059
3600
但你可能吓到了很多人。
09:31
that you might威力 freak怪物 out a lot of people.
204
559683
1966
你在考虑做这种
能自我检查的应用程序吗?
09:33
Are you thinking思维 of doing this,
making制造 an app应用 that allows允许 self-checking?
205
561673
3527
ST:我的收件箱充斥着
关于癌症应用程序的邮件,
09:37
STST: So my in-box内框 is flooded
about cancer癌症 apps应用,
206
565224
4973
还有人们让人心碎的故事。
09:42
with heartbreaking令人心碎 stories故事 of people.
207
570221
2303
有些人已经切除了
10、15、20个黑色素瘤,
09:44
I mean, some people have had
10, 15, 20 melanomas removed去除,
208
572548
3204
害怕可能会漏掉一个,就像这个,
09:47
and are scared害怕 that one
might威力 be overlooked忽视, like this one,
209
575776
3952
还有些是关于,
09:51
and also, about, I don't know,
210
579752
1741
我猜是飞行汽车和演讲咨询吧。
09:53
flying飞行 cars汽车 and speaker扬声器 inquiries查询
these days, I guess猜测.
211
581517
2732
我认为,我们需要更多的测试。
09:56
My take is, we need more testing测试.
212
584273
2738
我想要非常谨慎。
09:59
I want to be very careful小心.
213
587449
1778
给TED观众一个华丽的,
让人印象深刻的答案很容易。
10:01
It's very easy简单 to give a flashy华而不实 result结果
and impress a TEDTED audience听众.
214
589251
3666
但真正做出符合伦理道德的
事情就难得多。
10:04
It's much harder更难 to put
something out that's ethical合乎道德的.
215
592941
2627
如果人们要用这个应用程序,
10:07
And if people were to use the app应用
216
595592
2394
并选择不去寻求医生的帮助,
10:10
and choose选择 not to consult请教
the assistance帮助 of a doctor医生
217
598010
2797
但实际上是我们搞错了,
10:12
because we get it wrong错误,
218
600831
1583
我会感觉非常糟糕。
10:14
I would feel really bad about it.
219
602438
1653
所以我们现在正在进行临床试验,
10:16
So we're currently目前 doing clinical临床 tests测试,
220
604115
1925
如果这些临床试验开始后,
我们的数据还能保持正确,
10:18
and if these clinical临床 tests测试 commence开始
and our data数据 holds持有 up,
221
606064
2798
那么在某一时刻
我们或许可以采用这种技术,
10:20
we might威力 be able能够 at some point
to take this kind of technology技术
222
608886
2990
把它从斯坦福大学的诊所
10:23
and take it out of the Stanford斯坦福 clinic诊所
223
611900
1892
带到全世界,
10:25
and bring带来 it to the entire整个 world世界,
224
613816
1658
带到斯坦福的医生
从未踏足过的地方。
10:27
places地方 where Stanford斯坦福
doctors医生 never, ever set foot脚丫子.
225
615498
2456
CA:如果我听的没错,
10:30
CACA: And do I hear this right,
226
618617
2580
好像你说过,
10:33
that it seemed似乎 like what you were saying,
227
621221
1966
因为你跟优达学城的
学生军团打交道
10:35
because you are working加工
with this army军队 of UdacityUdacity students学生们,
228
623211
4254
你使用了与工业界不同形式的
10:39
that in a way, you're applying应用
a different不同 form形成 of machine learning学习
229
627489
3221
机器学习方式,
10:42
than might威力 take place地点 in a company公司,
230
630734
1735
也就是将机器学习与
群体智慧相结合。
10:44
which哪一个 is you're combining结合 machine learning学习
with a form形成 of crowd人群 wisdom智慧.
231
632493
3484
你是否在说,有时候你认为
10:48
Are you saying that sometimes有时 you think
that could actually其实 outperform跑赢大市
232
636001
3384
这能超越公司能做的事情,
甚至是一个巨型公司?
10:51
what a company公司 can do,
even a vast广大 company公司?
233
639409
2050
ST:我相信现在有一些事情
完全超乎我想象,
10:53
STST: I believe there's now
instances实例 that blow打击 my mind心神,
234
641483
2940
我还在试着去理解。
10:56
and I'm still trying to understand理解.
235
644447
1758
克里斯指的是
我们举办的这些比赛。
10:58
What Chris克里斯 is referring to
is these competitions比赛 that we run.
236
646229
3937
我们在48小时内完成,
11:02
We turn them around in 48 hours小时,
237
650190
2268
我们已经能够造出自动驾驶车,
11:04
and we've我们已经 been able能够 to build建立
a self-driving自驾车 car汽车
238
652482
2252
它能在大街上从山景城开到旧金山。
11:06
that can drive驾驶 from Mountain View视图
to San Francisco弗朗西斯科 on surface表面 streets街道.
239
654758
3387
这与谷歌的七年努力还不太能比,
11:10
It's not quite相当 on par平价 with Google谷歌
after seven years年份 of Google谷歌 work,
240
658169
3584
但是也快要实现了。
11:13
but it's getting得到 there.
241
661777
2528
而且我们只用了两个工程师,
三个月就完成了这个任务。
11:16
And it took us only two engineers工程师
and three months个月 to do this.
242
664329
3084
原因是,我们有一批
11:19
And the reason原因 is, we have
an army军队 of students学生们
243
667437
2856
参加比赛的学生军团。
11:22
who participate参加 in competitions比赛.
244
670317
1850
我们不是唯一使用众包的人。
11:24
We're not the only ones那些
who use crowdsourcing众包.
245
672191
2220
优步和滴滴也使用众包进行驾驶。
11:26
Uber尤伯杯 and Didi迪迪 use crowdsource众包 for driving主动.
246
674435
2223
Airbnb使用众包做酒店。
11:28
Airbnb制作的Airbnb uses使用 crowdsourcing众包 for hotels酒店.
247
676682
2759
现在有很多例子,
人们用众包找程序漏洞,
11:31
There's now many许多 examples例子
where people do bug-finding查找 bug crowdsourcing众包
248
679465
4007
或蛋白质折叠,各种众包。
11:35
or protein蛋白 folding折页, of all things,
in crowdsourcing众包.
249
683496
2804
但是我们已经做到
在三个月内造出这辆车,
11:38
But we've我们已经 been able能够 to build建立
this car汽车 in three months个月,
250
686324
2915
所以我实际上正在重新思考
11:41
so I am actually其实 rethinking重新思考
251
689263
3655
应该如何管理企业。
11:44
how we organize组织 corporations公司.
252
692942
2238
11:47
We have a staff员工 of 9,000 people
who are never hired雇用,
253
695204
4696
我们有从未雇用的9000员工,
我也从不解雇任何人。
11:51
that I never fire.
254
699924
1308
他们来上班,我甚至不知道。
11:53
They show显示 up to work
and I don't even know.
255
701256
2362
然后他们向我提交了
大概9000个答案。
11:55
Then they submit提交 to me
maybe 9,000 answers答案.
256
703642
3058
我并不必须使用任何一个答案。
11:58
I'm not obliged有义务的 to use any of those.
257
706724
2176
最后,我只付钱给赢家,
12:00
I end结束 up -- I pay工资 only the winners获奖者,
258
708924
1991
所以这方面我很吝啬,
这可能不太好。
12:02
so I'm actually其实 very cheapskate小气 here,
which哪一个 is maybe not the best最好 thing to do.
259
710939
3718
但他们也认为这是
教育的一部分,这很好。
12:06
But they consider考虑 it part部分
of their education教育, too, which哪一个 is nice不错.
260
714681
3185
但是这些学生已经能够做出
惊人的深度学习成果。
12:09
But these students学生们 have been able能够
to produce生产 amazing惊人 deep learning学习 results结果.
261
717890
4201
所以,优秀的人和优秀的的机器学习
结合起来简直太棒了。
12:14
So yeah, the synthesis合成 of great people
and great machine learning学习 is amazing惊人.
262
722115
3861
CA:加里·卡斯帕罗夫
在TED2017的第一天就说,
12:18
CACA: I mean, Gary加里 Kasparov卡斯帕罗夫 said on
the first day [of TEDTED2017]
263
726000
2814
国际象棋的胜利者
竟然是两个业余棋手,
12:20
that the winners获奖者 of chess, surprisingly出奇,
turned转身 out to be two amateur业余 chess players玩家
264
728848
5412
用三个很一般,或者
中等偏上的计算机程序,
12:26
with three mediocre-ish平庸,
mediocre-to-good平庸-好, computer电脑 programs程式,
265
734284
5371
赢了一个大师,一个很牛的棋手,
12:31
that could outperform跑赢大市 one grand盛大 master
with one great chess player播放机,
266
739679
3163
就像一切都是程序的一部分。
12:34
like it was all part部分 of the process处理.
267
742866
1743
看起来好像你正在
说的是同一想法的
12:36
And it almost几乎 seems似乎 like
you're talking about a much richer更丰富 version
268
744633
3335
更丰富的版本。
12:39
of that same相同 idea理念.
269
747992
1200
ST:是的,你也关注了昨天上午
那些很棒的小组讨论,
12:41
STST: Yeah, I mean, as you followed其次
the fantastic奇妙 panels面板 yesterday昨天 morning早上,
270
749216
3857
两个关于人工智能、
12:45
two sessions会议 about AIAI,
271
753097
1994
机器人霸主和人类反应的会议,
12:47
robotic机器人 overlords霸主 and the human人的 response响应,
272
755115
2167
说了很多很多很棒的东西。
12:49
many许多, many许多 great things were said.
273
757306
1982
但是其中一个问题是,
我们有时候
12:51
But one of the concerns关注 is
that we sometimes有时 confuse迷惑
274
759312
2687
会把人工智能真正做的事
与这种霸主威胁混淆,
12:54
what's actually其实 been doneDONE with AIAI
with this kind of overlord霸王 threat威胁,
275
762023
4062
威胁说人工智能
发展出意识了,对吧?
12:58
where your AIAI develops发展
consciousness意识, right?
276
766109
3424
我最不想看到的
就是我的人工智能有意识了。
13:01
The last thing I want
is for my AIAI to have consciousness意识.
277
769557
2971
我不想走进自己的厨房
13:04
I don't want to come into my kitchen厨房
278
772552
1716
突然发现冰箱爱上了洗碗机,
13:06
and have the refrigerator冰箱 fall秋季 in love
with the dishwasher洗碗机
279
774292
4193
还告诉我,因为我表现不错,
13:10
and tell me, because I wasn't nice不错 enough足够,
280
778509
2124
所以把我的饭热好了。
13:12
my food餐饮 is now warm.
281
780657
1837
我不会买这些产品的,
我也不想要。
13:14
I wouldn't不会 buy购买 these products制品,
and I don't want them.
282
782518
2891
但事实是,对于我来说,
13:17
But the truth真相 is, for me,
283
785825
1802
人工智能一直是对人的增强。
13:19
AIAI has always been
an augmentation增强 of people.
284
787651
2720
它是对我们的增强,
13:22
It's been an augmentation增强 of us,
285
790893
1676
使我们更强大。
13:24
to make us stronger.
286
792593
1457
我认为卡斯帕罗夫是完全正确的。
13:26
And I think Kasparov卡斯帕罗夫 was exactly究竟 correct正确.
287
794074
2831
是人类智慧和机器智慧的结合
13:28
It's been the combination组合
of human人的 smarts智慧 and machine smarts智慧
288
796929
3849
使我们变得更加强大。
13:32
that make us stronger.
289
800802
1464
机器使我们更强大的想法
与机器一样古老。
13:34
The theme主题 of machines making制造 us stronger
is as old as machines are.
290
802290
4587
13:39
The agricultural农业的 revolution革命 took
place地点 because it made制作 steam蒸汽 engines引擎
291
807567
3758
农业革命发生的原因是
它制造的蒸汽机和
农具不能自己种植,
13:43
and farming农业 equipment设备
that couldn't不能 farm农场 by itself本身,
292
811349
2666
机器从来没有取代我们;
只是让我们变得更强大。
13:46
that never replaced更换 us;
it made制作 us stronger.
293
814039
2122
我相信这个人工智能新浪潮
13:48
And I believe this new wave of AIAI
will make us much, much stronger
294
816185
3738
会让我们作为人类更加强大。
13:51
as a human人的 race种族.
295
819947
1183
CA:我们待会儿
再继续探讨这个问题,
13:53
CACA: We'll come on to that a bit more,
296
821765
1813
先说说对一些人来说可怕的部分,
13:55
but just to continue继续 with the scary害怕 part部分
of this for some people,
297
823602
3671
比如,有点让人担心的是
你有一台计算机,
13:59
like, what feels感觉 like it gets得到
scary害怕 for people is when you have
298
827297
3558
它能改写它自己的代码,
14:02
a computer电脑 that can, one,
rewrite改写 its own拥有 code,
299
830879
4618
所以,它能自己复制很多个自己,
14:07
so, it can create创建
multiple copies副本 of itself本身,
300
835521
3584
还试验好多不同的代码版本,
14:11
try a bunch of different不同 code versions版本,
301
839129
1897
甚至可能是随机的版本,
14:13
possibly或者 even at random随机,
302
841050
1775
然后自己检验,看看
目标有没有实现或得到改进。
14:14
and then check them out and see
if a goal目标 is achieved实现 and improved改善.
303
842849
3632
比如说,目标是
在智力测验上表现更好。
14:18
So, say the goal目标 is to do better
on an intelligence情报 test测试.
304
846505
3641
你知道,计算机很擅长这个,
14:22
You know, a computer电脑
that's moderately适当地 good at that,
305
850170
3894
可以尝试一百万个版本。
14:26
you could try a million百万 versions版本 of that.
306
854088
2509
可能会发现一个更好的,
14:28
You might威力 find one that was better,
307
856621
2090
然后,自己重复。
14:30
and then, you know, repeat重复.
308
858735
2004
所以让人担心的是,
会发生类似失控效应,
14:32
And so the concern关心 is that you get
some sort分类 of runaway逃跑 effect影响
309
860763
3040
比如周四晚上一切正常,
14:35
where everything is fine
on Thursday星期四 evening晚间,
310
863827
3008
周五早晨到实验室,
14:38
and you come back into the lab实验室
on Friday星期五 morning早上,
311
866859
2336
由于计算机的速度等等,
14:41
and because of the speed速度
of computers电脑 and so forth向前,
312
869219
2449
一切都开始失控,突然——
14:43
things have gone走了 crazy, and suddenly突然 --
313
871692
1903
ST:我只能说这是一种可能性,
14:45
STST: I would say this is a possibility可能性,
314
873619
2020
但是这个可能性非常遥远。
14:47
but it's a very remote远程 possibility可能性.
315
875663
1916
先让我翻译一下你所说的话。
14:49
So let me just translate翻译
what I heard听说 you say.
316
877603
3337
在阿尔法围棋中,
我们确实有这样的情况:
14:52
In the AlphaGoAlphaGo case案件,
we had exactly究竟 this thing:
317
880964
2704
计算机跟自己比赛,
14:55
the computer电脑 would play
the game游戏 against反对 itself本身
318
883692
2315
然后学到新规则。
14:58
and then learn学习 new rules规则.
319
886031
1250
而机器学习就是改写规则。
14:59
And what machine learning学习 is
is a rewriting重写 of the rules规则.
320
887305
3235
改写代码。
15:02
It's the rewriting重写 of code.
321
890564
1769
但我认为绝对不用担心
15:04
But I think there was
absolutely绝对 no concern关心
322
892357
2845
阿尔法围棋会占领世界。
15:07
that AlphaGoAlphaGo would take over the world世界.
323
895226
2426
它连国际象棋也不会玩。
15:09
It can't even play chess.
324
897676
1464
CA:没错没错,但现在
这些都是非常单一领域的东西。
15:11
CACA: No, no, no, but now,
these are all very single-domain单域 things.
325
899164
5147
但能够想象。
15:16
But it's possible可能 to imagine想像.
326
904335
2879
我是说,我们刚刚看到一个计算机
15:19
I mean, we just saw a computer电脑
that seemed似乎 nearly几乎 capable
327
907238
3089
好像几乎能够通过大学入学考试了,
15:22
of passing通过 a university大学 entrance入口 test测试,
328
910351
2655
不过——它不像我们一样阅读和理解,
15:25
that can kind of -- it can't read
and understand理解 in the sense that we can,
329
913030
3688
却能吸收所有文字,
15:28
but it can certainly当然 absorb吸收 all the text文本
330
916742
1987
还能看见更多的意义模式。
15:30
and maybe see increased增加
patterns模式 of meaning含义.
331
918753
2899
会不会有可能,
随着这个继续发展壮大,
15:33
Isn't there a chance机会 that,
as this broadens变宽 out,
332
921676
3694
会出现另一种失控效应?
15:37
there could be a different不同
kind of runaway逃跑 effect影响?
333
925394
2466
ST:老实说,这就是我
划分界限的地方。
15:39
STST: That's where
I draw the line线, honestly老老实实.
334
927884
2078
可能性是存在的——
我不想轻描淡写——
15:41
And the chance机会 exists存在 --
I don't want to downplay淡化 it --
335
929986
2643
但我认为它很遥远,
目前我脑子里不会想这个,
15:44
but I think it's remote远程, and it's not
the thing that's on my mind心神 these days,
336
932653
3672
因为我认为
大改革是指另一回事。
15:48
because I think the big revolution革命
is something else其他.
337
936349
2512
到今天,人工智能所有的成功
15:50
Everything successful成功 in AIAI
to the present当下 date日期
338
938885
2922
都是极度专业化的,
15:53
has been extremely非常 specialized专门,
339
941831
2214
并且它的繁荣一直
基于单一的理念,
15:56
and it's been thriving on a single idea理念,
340
944069
2489
就是大量的数据。
15:58
which哪一个 is massive大规模的 amounts of data数据.
341
946582
2739
阿尔法围棋这么成功的原因
是大量的围棋比赛数据,
16:01
The reason原因 AlphaGoAlphaGo works作品 so well
is because of massive大规模的 numbers数字 of Go plays播放,
342
949345
4147
阿尔法围棋不能开车
也不能开飞机。
16:05
and AlphaGoAlphaGo can't drive驾驶 a car汽车
or fly a plane平面.
343
953516
3255
谷歌自动驾驶车或
优达学城自动驾驶车
16:08
The Google谷歌 self-driving自驾车 car汽车
or the UdacityUdacity self-driving自驾车 car汽车
344
956795
3031
在海量数据上建成,
但做不了其他事。
16:11
thrives蓬勃发展 on massive大规模的 amounts of data数据,
and it can't do anything else其他.
345
959850
3240
甚至控制不了摩托车。
16:15
It can't even control控制 a motorcycle摩托车.
346
963114
1727
这是一个非常具体的、
特定领域的功能,
16:16
It's a very specific具体,
domain-specific特定领域 function功能,
347
964865
2762
我们的癌症应用程序也是如此。
16:19
and the same相同 is true真正 for our cancer癌症 app应用.
348
967651
1907
而所谓“通用人工智能”,
几乎没有进展,
16:21
There has been almost几乎 no progress进展
on this thing called "general一般 AIAI,"
349
969582
3236
“通用”就是你去对人工智能说:
“嘿,为我发明个狭义相对论
16:24
where you go to an AIAI and say,
"Hey, invent发明 for me special特别 relativity相对论
350
972842
4000
或弦理论。”
16:28
or string theory理论."
351
976866
1666
那完全是在婴儿期。
16:30
It's totally完全 in the infancy婴儿期.
352
978556
1931
我想强调这一点的原因是,
16:32
The reason原因 I want to emphasize注重 this,
353
980511
2127
我明白大家的担忧,
我想告诉大家我了解。
16:34
I see the concerns关注,
and I want to acknowledge确认 them.
354
982662
3838
但是如果我只能考虑一件事情,
16:38
But if I were to think about one thing,
355
986524
2886
我会问自己: “如果我们
把所有重复性的事情解决掉,
16:41
I would ask myself the question,
"What if we can take anything repetitive重复
356
989434
5563
让自己的效率提高100倍,会怎样?”
16:47
and make ourselves我们自己
100 times as efficient高效?"
357
995021
3473
事实证明,三百年前,我们都务农,
16:51
It so turns out, 300 years年份 ago,
we all worked工作 in agriculture农业
358
999170
4249
耕种,做重复的事。
16:55
and did farming农业 and did repetitive重复 things.
359
1003443
2051
今天,我们75%的人
在办公室里工作,
16:57
Today今天, 75 percent百分 of us work in offices办事处
360
1005518
2556
仍然做重复的事。
17:00
and do repetitive重复 things.
361
1008098
2124
我们已经变成专做表格的猴子。
17:02
We've我们已经 become成为 spreadsheet电子表格 monkeys猴子.
362
1010246
2183
不只是低端劳动力,
17:04
And not just low-end低端 labor劳动.
363
1012453
2054
我们已经变成了
皮肤科医生在做重复的工作,
17:06
We've我们已经 become成为 dermatologists皮肤科医生
doing repetitive重复 things,
364
1014531
2754
律师也在做重复的工作。
17:09
lawyers律师 doing repetitive重复 things.
365
1017309
1749
我想我们处于一个边缘,
能够利用人工智能
17:11
I think we are at the brink边缘
of being存在 able能够 to take an AIAI,
366
1019082
3823
替我们仔细查看,
17:14
look over our shoulders肩膀,
367
1022929
1718
帮我们在这些重复的事情上
把效率提高10倍或50倍。
17:16
and they make us maybe 10 or 50 times
as effective有效 in these repetitive重复 things.
368
1024671
4058
这才是我在考虑的事。
17:20
That's what is on my mind心神.
369
1028753
1275
CA:听起来很刺激。
17:22
CACA: That sounds声音 super exciting扣人心弦.
370
1030052
2450
实现这些的过程会让
一些人内心多少有些抵触,
17:24
The process处理 of getting得到 there seems似乎
a little terrifying可怕的 to some people,
371
1032526
3530
因为一旦电脑可以比皮肤科医生,
17:28
because once一旦 a computer电脑
can do this repetitive重复 thing
372
1036080
3180
尤其是比司机
17:31
much better than the dermatologist皮肤科医生
373
1039284
3434
更能胜任重复劳动,
17:34
or than the driver司机, especially特别,
is the thing that's talked about
374
1042742
3230
现在这是热门话题,
17:37
so much now,
375
1045996
1290
突然上百万工作消失了,
17:39
suddenly突然 millions百万 of jobs工作 go,
376
1047310
1958
并且,你知道,国家变得速度很快,
17:41
and, you know, the country's in revolution革命
377
1049292
2695
我们根本来不及实现更耀眼的成就。
17:44
before we ever get to the more
glorious辉煌 aspects方面 of what's possible可能.
378
1052011
4329
ST:是的,这是个问题,
是个大问题,
17:48
STST: Yeah, and that's an issue问题,
and it's a big issue问题,
379
1056364
2517
昨天上午也有几位演讲嘉宾提到了。
17:50
and it was pointed out yesterday昨天 morning早上
by several一些 guest客人 speakers音箱.
380
1058905
4196
在我上台之前,
17:55
Now, prior to me showing展示 up onstage在舞台上,
381
1063125
2754
我承认我是一个积极乐观的人,
17:57
I confessed供认不讳 I'm a positive,
optimistic乐观 person,
382
1065903
3739
所以让我给你一个乐观的意见,
18:01
so let me give you an optimistic乐观 pitch沥青,
383
1069666
2389
假想你在300年前。
18:04
which哪一个 is, think of yourself你自己
back 300 years年份 ago.
384
1072079
4795
欧洲刚刚经历了140年的连续战争,
18:08
Europe欧洲 just survived幸存 140 years年份
of continuous连续 war战争,
385
1076898
3996
没有人会读书写字,
18:12
none没有 of you could read or write,
386
1080918
1711
没有现代社会的工作,
18:14
there were no jobs工作 that you hold保持 today今天,
387
1082653
2945
比如投资银行家、
软件工程师或电视主播。
18:17
like investment投资 banker银行家
or software软件 engineer工程师 or TV电视 anchor.
388
1085622
4096
我们都要在田野里种地。
18:21
We would all be in the fields领域 and farming农业.
389
1089742
2414
现在小塞巴斯蒂安来了,
口袋里装着一个小蒸汽机,
18:24
Now here comes little Sebastian塞巴斯蒂安
with a little steam蒸汽 engine发动机 in his pocket口袋,
390
1092180
3573
他说:“嘿,伙计们,看看这个,
18:27
saying, "Hey guys, look at this.
391
1095777
1548
它会让你强壮100倍,
然后你就可以做点别的了。”
18:29
It's going to make you 100 times
as strong强大, so you can do something else其他."
392
1097349
3595
那时候,没有真正的舞台,
18:32
And then back in the day,
there was no real真实 stage阶段,
393
1100968
2470
我和克里斯在牛棚里跟牛闲晃,
18:35
but Chris克里斯 and I hang out
with the cows奶牛 in the stable稳定,
394
1103462
2526
他说,“我真的很担心,
18:38
and he says, "I'm really
concerned关心 about it,
395
1106012
2100
因为我每天挤牛奶,如果机器
也能干这活儿了,我可怎么办呐?”
18:40
because I milk牛奶 my cow every一切 day,
and what if the machine does this for me?"
396
1108136
3652
我之所以提到这个,
18:43
The reason原因 why I mention提到 this is,
397
1111812
1702
是因为我们总是擅长
承认过去的进步和好处,
18:46
we're always good in acknowledging承认
past过去 progress进展 and the benefit效益 of it,
398
1114360
3603
比如iPhone或飞机,
电力或者医疗供应。
18:49
like our iPhonesiPhone手机 or our planes飞机
or electricity电力 or medical supply供应.
399
1117987
3354
我们都喜欢活到80年,
这在300年前是不可能的。
18:53
We all love to live生活 to 80,
which哪一个 was impossible不可能 300 years年份 ago.
400
1121365
4245
但是我们对未来的态度
却并不基于相同的规则。
18:57
But we kind of don't apply应用
the same相同 rules规则 to the future未来.
401
1125634
4156
19:02
So if I look at my own拥有 job工作 as a CEOCEO,
402
1130621
3207
如果我审视自己的
首席执行官工作,
我认为我的工作中
有90%是重复性的,
19:05
I would say 90 percent百分
of my work is repetitive重复,
403
1133852
3140
我不喜欢,
19:09
I don't enjoy请享用 it,
404
1137016
1351
我每天花四个小时在
愚蠢、重复的电子邮件上。
19:10
I spend about four hours小时 per day
on stupid, repetitive重复 email电子邮件.
405
1138391
3978
我正心急如焚想要
找谁帮我摆脱这一点。
19:14
And I'm burning燃烧 to have something
that helps帮助 me get rid摆脱 of this.
406
1142393
3641
为什么?
19:18
Why?
407
1146058
1158
因为我相信每个人都有无限创造力。
19:19
Because I believe all of us
are insanely疯狂地 creative创作的;
408
1147240
3003
我认为TED社区更是如此。
19:22
I think the TEDTED community社区
more than anybody任何人 else其他.
409
1150731
3194
但即使是蓝领工人,
你可以找酒店清洁工
19:25
But even blue-collar蓝领 workers工人;
I think you can go to your hotel旅馆 maid女佣
410
1153949
3559
跟他或她喝一杯,
19:29
and have a drink with him or her,
411
1157532
2402
一小时后,你就会发现
有创意的想法。
19:31
and an hour小时 later后来,
you find a creative创作的 idea理念.
412
1159958
2717
人工智能将赋予我们的力量是
将这种创造力转化为行动。
19:34
What this will empower授权
is to turn this creativity创造力 into action行动.
413
1162699
4140
比如,如果你能
在一天内造出谷歌会怎样?
19:39
Like, what if you could
build建立 Google谷歌 in a day?
414
1167265
3442
如果你坐这儿喝着啤酒,
就发明出下一个Snapchat会怎样?
19:43
What if you could sit over beer啤酒
and invent发明 the next下一个 SnapchatSnapchat,
415
1171221
3316
不管发明的是什么吧,
19:46
whatever随你 it is,
416
1174561
1165
第二天早上它就完工、
投入运行会怎样?
19:47
and tomorrow明天 morning早上 it's up and running赛跑?
417
1175750
2187
那不是科幻小说。
19:49
And that is not science科学 fiction小说.
418
1177961
1773
可以预见的是,
19:51
What's going to happen发生 is,
419
1179758
1254
我们已经处于历史当中。
19:53
we are already已经 in history历史.
420
1181036
1867
我们已经释放出惊人的创造力,
19:54
We've我们已经 unleashed如虎添翼 this amazing惊人 creativity创造力
421
1182927
3228
先从农耕解放出来,
19:58
by de-slavingde 奴 us from farming农业
422
1186179
1611
又从工厂劳动解放出来,
19:59
and later后来, of course课程, from factory work
423
1187814
3363
我们发明了这么多东西。
20:03
and have invented发明 so many许多 things.
424
1191201
3162
我认为,将来会更好的。
20:06
It's going to be even better,
in my opinion意见.
425
1194387
2178
当然也会有更大的副作用。
20:08
And there's going to be
great side effects效果.
426
1196589
2072
其中一个副作用就是
20:10
One of the side effects效果 will be
427
1198685
1489
比如食物、医疗、教育、庇护
20:12
that things like food餐饮 and medical supply供应
and education教育 and shelter庇护
428
1200198
4795
交通等这些东西,
20:17
and transportation运输
429
1205017
1177
将会让所有人都承受得起,
20:18
will all become成为 much more
affordable实惠 to all of us,
430
1206218
2441
而不只是富人。
20:20
not just the rich丰富 people.
431
1208683
1322
CA:嗯。
20:22
CACA: Hmm.
432
1210029
1182
所以,之前马丁·福特提出的,
与这一次有所不同,
20:23
So when Martin马丁 Ford argued争论, you know,
that this time it's different不同
433
1211235
4341
说因为我们以前的
20:27
because the intelligence情报
that we've我们已经 used in the past过去
434
1215600
3453
用来寻找新方法的智慧
20:31
to find new ways方法 to be
435
1219077
2483
将被计算机接管,
20:33
will be matched匹配 at the same相同 pace步伐
436
1221584
2279
以相同的步调继续下去,
20:35
by computers电脑 taking服用 over those things,
437
1223887
2291
而我听你的意思,那不完全对,
20:38
what I hear you saying
is that, not completely全然,
438
1226202
3078
原因是人的创造力。
20:41
because of human人的 creativity创造力.
439
1229304
2951
你是否认为人的创造力
与计算机的那种创造力
20:44
Do you think that that's fundamentally从根本上
different不同 from the kind of creativity创造力
440
1232279
3785
有着根本的区别?
20:48
that computers电脑 can do?
441
1236088
2696
ST:那是我作为
一个AI人的坚定信念——
20:50
STST: So, that's my firm公司
belief信仰 as an AIAI person --
442
1238808
4434
在创造力和创新思维方面,
我并没有看到
20:55
that I haven't没有 seen看到
any real真实 progress进展 on creativity创造力
443
1243266
3803
20:59
and out-of-the-box盒子外面 thinking思维.
444
1247949
1407
任何真正的进展。
我现在所看到的——
大家也一定要意识到,
21:01
What I see right now -- and this is
really important重要 for people to realize实现,
445
1249380
3623
由于“人工智能”一词
如此有威胁性,
21:05
because the word "artificial人造
intelligence情报" is so threatening危险的,
446
1253027
2903
而且史蒂夫·斯皮尔伯格
又加进一部电影,
21:07
and then we have Steve史蒂夫 Spielberg斯皮尔伯格
tossing折腾 a movie电影 in,
447
1255954
2523
电影里突然之间
计算机变成我们的霸主——
21:10
where all of a sudden突然
the computer电脑 is our overlord霸王,
448
1258501
2413
但人工智能真的只是一种技术。
21:12
but it's really a technology技术.
449
1260938
1452
是帮我们做重复工作的技术。
21:14
It's a technology技术 that helps帮助 us
do repetitive重复 things.
450
1262414
2982
而且进展完全发生在
重复性事件上。
21:17
And the progress进展 has been
entirely完全 on the repetitive重复 end结束.
451
1265420
2913
比如法律文件探索、
21:20
It's been in legal法律 document文件 discovery发现.
452
1268357
2228
合同起草、
21:22
It's been contract合同 drafting制图.
453
1270609
1680
胸部X光片筛查,
21:24
It's been screening筛查 X-raysX射线 of your chest胸部.
454
1272313
4223
这些都是非常专业的,
21:28
And these things are so specialized专门,
455
1276560
1773
我不觉得对人类有什么大威胁。
21:30
I don't see the big threat威胁 of humanity人性.
456
1278357
2391
事实上,我们作为人类——
21:32
In fact事实, we as people --
457
1280772
1794
让我们面对事实:
我们已经变成了超人。
21:34
I mean, let's face面对 it:
we've我们已经 become成为 superhuman超人.
458
1282590
2385
我们把自己变成了超人。
21:36
We've我们已经 made制作 us superhuman超人.
459
1284999
1764
我们能用11个小时游过大西洋。
21:38
We can swim游泳 across横过
the Atlantic大西洋 in 11 hours小时.
460
1286787
2632
我们能从口袋里掏出设备
21:41
We can take a device设备 out of our pocket口袋
461
1289443
2074
喊到澳大利亚去,
21:43
and shout all the way to Australia澳大利亚,
462
1291541
2147
并且同时,那人可以喊回来。
21:45
and in real真实 time, have that person
shouting叫喊 back to us.
463
1293712
2600
这在物理学上是不可能的。
我们正在打破物理规则。
21:48
That's physically物理 not possible可能.
We're breaking破坏 the rules规则 of physics物理.
464
1296336
3624
当这样说了,这样做了,我们会记住
21:51
When this is said and doneDONE,
we're going to remember记得 everything
465
1299984
2943
我们曾说过和见过的一切,
21:54
we've我们已经 ever said and seen看到,
466
1302951
1213
你会记得每一个人,
21:56
you'll你会 remember记得 every一切 person,
467
1304188
1496
这对我的早期老年痴呆有好处。
21:57
which哪一个 is good for me
in my early stages阶段 of Alzheimer's老年痴呆症.
468
1305708
2626
对不起,我在说什么?我忘了。
22:00
Sorry, what was I saying? I forgot忘记.
469
1308358
1677
CA:(笑)
22:02
CACA: (Laughs)
470
1310059
1578
ST:我们的智商可能超过1000。
22:03
STST: We will probably大概 have
an IQ智商 of 1,000 or more.
471
1311661
3077
我们的孩子将不再有拼写课,
22:06
There will be no more
spelling拼字 classes for our kids孩子,
472
1314762
3425
因为不存在拼写问题了。
22:10
because there's no spelling拼字 issue问题 anymore.
473
1318211
2086
也不存在数学问题了。
22:12
There's no math数学 issue问题 anymore.
474
1320321
1832
我认为真正会发生的是,
我们将变得充满创意。
22:14
And I think what really will happen发生
is that we can be super creative创作的.
475
1322177
3510
是的,我们很有创意。
22:17
And we are. We are creative创作的.
476
1325711
1857
这是我们的秘密武器。
22:19
That's our secret秘密 weapon武器.
477
1327592
1552
CA:所以那些将要消失的工作,
22:21
CACA: So the jobs工作 that are getting得到 lost丢失,
478
1329168
2153
某种程度上,即使痛苦,
22:23
in a way, even though虽然
it's going to be painful痛苦,
479
1331345
2494
人类能够做的远不止那些工作。
22:25
humans人类 are capable
of more than those jobs工作.
480
1333863
2047
这才是(人工智能的最终)梦想。
22:27
This is the dream梦想.
481
1335934
1218
梦想人类可以上升到能量与探索的
22:29
The dream梦想 is that humans人类 can rise上升
to just a new level水平 of empowerment权力
482
1337176
4247
新高度。
22:33
and discovery发现.
483
1341447
1657
那才是梦想。
22:35
That's the dream梦想.
484
1343128
1452
ST:想想看:
22:36
STST: And think about this:
485
1344604
1643
如果你看一下人类的历史,
22:38
if you look at the history历史 of humanity人性,
486
1346271
2021
可能是大概6万至10万年的岁月,
22:40
that might威力 be whatever随你 --
60-100,000 years年份 old, give or take --
487
1348316
3328
几乎每一件珍贵的发明
22:43
almost几乎 everything that you cherish珍视
in terms条款 of invention发明,
488
1351668
3726
技术发明,或建造的作品,
22:47
of technology技术, of things we've我们已经 built内置,
489
1355418
2151
都是在最近150年完成的。
22:49
has been invented发明 in the last 150 years年份.
490
1357593
3099
如果算上书本和车轮,还要更久一点。
22:53
If you toss折腾 in the book and the wheel,
it's a little bit older旧的.
491
1361756
3048
或斧头。
22:56
Or the axe斧头.
492
1364828
1169
但你的手机、跑鞋,
22:58
But your phone电话, your sneakers球鞋,
493
1366021
2790
这些椅子、现代制造、青霉素——
23:00
these chairs椅子, modern现代
manufacturing制造业, penicillin青霉素 --
494
1368835
3551
这些我们珍惜的东西。
23:04
the things we cherish珍视.
495
1372410
1714
现在,对我而言它意味着,
23:06
Now, that to me means手段
496
1374148
3658
接下来的150年将会发现更多的东西。
23:09
the next下一个 150 years年份 will find more things.
497
1377830
3041
事实上,在我看来,发明的速度
已经上升了,没有下降。
23:12
In fact事实, the pace步伐 of invention发明
has gone走了 up, not gone走了 down, in my opinion意见.
498
1380895
4154
我相信有趣的东西只有
1%被发明出来了。可以理解吧?
23:17
I believe only one percent百分 of interesting有趣
things have been invented发明 yet然而. Right?
499
1385073
4905
我们还没有治愈癌症。
23:22
We haven't没有 cured治愈 cancer癌症.
500
1390002
1988
我们没有飞行汽车——目前还没有,
希望我会改变这一点。
23:24
We don't have flying飞行 cars汽车 -- yet然而.
Hopefully希望, I'll change更改 this.
501
1392014
3718
那曾经是大家的笑料。(笑)
23:27
That used to be an example
people laughed笑了 about. (Laughs)
502
1395756
3257
是不是很逗,
秘密地研究飞行汽车?
23:31
It's funny滑稽, isn't it?
Working加工 secretly偷偷 on flying飞行 cars汽车.
503
1399037
2992
我们的寿命还没有翻倍,对吧?
23:34
We don't live生活 twice两次 as long yet然而. OK?
504
1402053
2683
我们还没有神奇的脑植入物
23:36
We don't have this magic魔法
implant注入 in our brain
505
1404760
2785
来提供我们想要的信息。
23:39
that gives us the information信息 we want.
506
1407569
1832
你可能会为此感到惊恐,
23:41
And you might威力 be appalled by it,
507
1409425
1526
但我向你保证,一旦拥有了,
你一定会喜欢的。
23:42
but I promise诺言 you,
once一旦 you have it, you'll你会 love it.
508
1410975
2444
我希望你会的。
23:45
I hope希望 you will.
509
1413443
1166
有点吓人,我明白。
23:46
It's a bit scary害怕, I know.
510
1414633
1909
还有这么多没有出现的东西
23:48
There are so many许多 things
we haven't没有 invented发明 yet然而
511
1416566
2254
我想我们会发明出来的。
23:50
that I think we'll invent发明.
512
1418844
1268
我们没有引力盾。
23:52
We have no gravity重力 shields盾牌.
513
1420136
1306
我们不能把自己从一个地点
转移到另一个地点。
23:53
We can't beam光束 ourselves我们自己
from one location位置 to another另一个.
514
1421466
2553
这听起来挺荒唐,
但大约200年前,
23:56
That sounds声音 ridiculous荒谬,
515
1424043
1151
23:57
but about 200 years年份 ago,
516
1425218
1288
专家们还认为飞机不会存在,
23:58
experts专家 were of the opinion意见
that flight飞行 wouldn't不会 exist存在,
517
1426530
2667
即使120年前,
24:01
even 120 years年份 ago,
518
1429221
1324
如果你的移动速度
比你跑步还快,
24:02
and if you moved移动 faster更快
than you could run,
519
1430569
2582
你会立即死掉。
24:05
you would instantly即刻 die.
520
1433175
1520
那么今天有谁敢说我们肯定不能把人
24:06
So who says we are correct正确 today今天
that you can't beam光束 a person
521
1434719
3569
从这儿送到火星呢?
24:10
from here to Mars火星?
522
1438312
2249
CA:塞巴斯蒂安,非常感谢你今天来
24:12
CACA: Sebastian塞巴斯蒂安, thank you so much
523
1440585
1569
分享你无比激励的展望和你的才华。
24:14
for your incredibly令人难以置信 inspiring鼓舞人心 vision视力
and your brilliance.
524
1442178
2682
谢谢塞巴斯蒂安·斯伦。
24:16
Thank you, Sebastian塞巴斯蒂安 Thrun史朗.
525
1444884
1323
ST:真棒。 (掌声)
24:18
STST: That was fantastic奇妙. (Applause掌声)
526
1446231
1895
Translated by Yan Gao

▲Back to top

ABOUT THE SPEAKERS
Sebastian Thrun - Educator, entrepreneur
Sebastian Thrun is a passionate technologist who is constantly looking for new opportunities to make the world better for all of us.

Why you should listen

Sebastian Thrun is an educator, entrepreneur and troublemaker. After a long life as a professor at Stanford University, Thrun resigned from tenure to join Google. At Google, he founded Google X, home to self-driving cars and many other moonshot technologies. Thrun also founded Udacity, an online university with worldwide reach, and Kitty Hawk, a "flying car" company. He has authored 11 books, 400 papers, holds 3 doctorates and has won numerous awards.

More profile about the speaker
Sebastian Thrun | Speaker | TED.com
Chris Anderson - TED Curator
After a long career in journalism and publishing, Chris Anderson became the curator of the TED Conference in 2002 and has developed it as a platform for identifying and disseminating ideas worth spreading.

Why you should listen

Chris Anderson is the Curator of TED, a nonprofit devoted to sharing valuable ideas, primarily through the medium of 'TED Talks' -- short talks that are offered free online to a global audience.

Chris was born in a remote village in Pakistan in 1957. He spent his early years in India, Pakistan and Afghanistan, where his parents worked as medical missionaries, and he attended an American school in the Himalayas for his early education. After boarding school in Bath, England, he went on to Oxford University, graduating in 1978 with a degree in philosophy, politics and economics.

Chris then trained as a journalist, working in newspapers and radio, including two years producing a world news service in the Seychelles Islands.

Back in the UK in 1984, Chris was captivated by the personal computer revolution and became an editor at one of the UK's early computer magazines. A year later he founded Future Publishing with a $25,000 bank loan. The new company initially focused on specialist computer publications but eventually expanded into other areas such as cycling, music, video games, technology and design, doubling in size every year for seven years. In 1994, Chris moved to the United States where he built Imagine Media, publisher of Business 2.0 magazine and creator of the popular video game users website IGN. Chris eventually merged Imagine and Future, taking the combined entity public in London in 1999, under the Future name. At its peak, it published 150 magazines and websites and employed 2,000 people.

This success allowed Chris to create a private nonprofit organization, the Sapling Foundation, with the hope of finding new ways to tackle tough global issues through media, technology, entrepreneurship and, most of all, ideas. In 2001, the foundation acquired the TED Conference, then an annual meeting of luminaries in the fields of Technology, Entertainment and Design held in Monterey, California, and Chris left Future to work full time on TED.

He expanded the conference's remit to cover all topics, including science, business and key global issues, while adding a Fellows program, which now has some 300 alumni, and the TED Prize, which grants its recipients "one wish to change the world." The TED stage has become a place for thinkers and doers from all fields to share their ideas and their work, capturing imaginations, sparking conversation and encouraging discovery along the way.

In 2006, TED experimented with posting some of its talks on the Internet. Their viral success encouraged Chris to begin positioning the organization as a global media initiative devoted to 'ideas worth spreading,' part of a new era of information dissemination using the power of online video. In June 2015, the organization posted its 2,000th talk online. The talks are free to view, and they have been translated into more than 100 languages with the help of volunteers from around the world. Viewership has grown to approximately one billion views per year.

Continuing a strategy of 'radical openness,' in 2009 Chris introduced the TEDx initiative, allowing free licenses to local organizers who wished to organize their own TED-like events. More than 8,000 such events have been held, generating an archive of 60,000 TEDx talks. And three years later, the TED-Ed program was launched, offering free educational videos and tools to students and teachers.

More profile about the speaker
Chris Anderson | Speaker | TED.com