ABOUT THE SPEAKER
Anthony Goldbloom - Machine learning expert
Anthony Goldbloom crowdsources solutions to difficult problems using machine learning.

Why you should listen

Anthony Goldbloom is the co-founder and CEO of Kaggle. Kaggle hosts machine learning competitions, where data scientists download data and upload solutions to difficult problems. Kaggle has a community of over 600,000 data scientists and has worked with companies ranging Facebook to GE on problems ranging from predicting friendships to flight arrival times.

Before Kaggle, Anthony worked as an econometrician at the Reserve Bank of Australia, and before that the Australian Treasury. In 2011 and 2012, Forbes named Anthony one of the 30 under 30 in technology; in 2013 the MIT Tech Review named him one of top 35 innovators under the age of 35, and the University of Melbourne awarded him an Alumni of Distinction Award. He holds a first call honors degree in Econometrics from the University of Melbourne.  

More profile about the speaker
Anthony Goldbloom | Speaker | TED.com
TED2016

Anthony Goldbloom: The jobs we'll lose to machines -- and the ones we won't

安東尼.葛博倫: 即將要被機器取代以及無法取代的工作

Filmed:
2,568,213 views

機器學習不再僅止於簡單任務,好比評估信用風險、分類郵件。如今,它能夠承擔更複雜的工作,像是幫作文打分數、診斷疾病。這些進步帶來了一個令人不安的問題:未來機器人會搶走你的工作嗎?
- Machine learning expert
Anthony Goldbloom crowdsources solutions to difficult problems using machine learning. Full bio

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

00:12
So this is my niece侄女.
0
968
1262
這是我的姪女,
00:14
Her name名稱 is YahliYahli.
1
2644
1535
她的名字是雅莉,
00:16
She is nine months個月 old.
2
4203
1511
她現在九個月大,
00:18
Her mum沉默 is a doctor醫生,
and her dad is a lawyer律師.
3
6201
2528
媽媽是位醫生、爸爸是位律師;
00:21
By the time YahliYahli goes to college學院,
4
9269
2006
不過等到她上大學的時候
00:23
the jobs工作 her parents父母 do
are going to look dramatically顯著 different不同.
5
11299
3253
她父母親的工作將會迥然不同了。
00:27
In 2013, researchers研究人員 at Oxford牛津 University大學
did a study研究 on the future未來 of work.
6
15347
5073
2013年,牛津大學的研究人員
做了一個對未來工作的研究,
00:32
They concluded總結 that almost幾乎 one
in every一切 two jobs工作 have a high risk風險
7
20766
4139
他們得出結論:差不多將近一半的工作
00:36
of being存在 automated自動化 by machines.
8
24929
1824
都有被機器自動化取代的危險。
00:40
Machine learning學習 is the technology技術
9
28388
1905
而「機器學習」
要對這種顛覆負主要的責任。
00:42
that's responsible主管 for most
of this disruption瓦解.
10
30317
2278
00:44
It's the most powerful強大 branch
of artificial人造 intelligence情報.
11
32619
2790
它是人工智慧最呼風喚雨的分支,
00:47
It allows允許 machines to learn學習 from data數據
12
35433
1882
它讓機器得以從數據中學習,
00:49
and mimic模仿者 some of the things
that humans人類 can do.
13
37339
2592
並模仿一些人類可以做到的事情。
00:51
My company公司, KaggleKaggle, operates操作
on the cutting切割 edge邊緣 of machine learning學習.
14
39955
3415
我的公司「Kaggle」算是能操控
機器學習的尖端科技公司。
00:55
We bring帶來 together一起
hundreds數以百計 of thousands數千 of experts專家
15
43394
2386
我們召集了成千上萬的專家
00:57
to solve解決 important重要 problems問題
for industry行業 and academia學術界.
16
45804
3118
為產、學界解決重要的難題。
01:01
This gives us a unique獨特 perspective透視
on what machines can do,
17
49279
3222
所以我們可以從獨特的角度
來觀察機器可以做什麽、
01:04
what they can't do
18
52525
1235
不可以做什麽。
01:05
and what jobs工作 they might威力
automate自動化 or threaten威脅.
19
53784
2939
哪些工作可以被自動化或者受到威脅。
01:09
Machine learning學習 started開始 making製造 its way
into industry行業 in the early '90s.
20
57316
3550
機器學習是在90年代初
進入產業界的,
01:12
It started開始 with relatively相對 simple簡單 tasks任務.
21
60890
2124
一開始,它只是執行一些
簡單的任務。
01:15
It started開始 with things like assessing評估
credit信用 risk風險 from loan貸款 applications應用,
22
63406
4115
像評估貸款申請的信用風險、
01:19
sorting排序 the mail郵件 by reading
handwritten手寫 characters人物 from zip壓縮 codes代碼.
23
67545
4053
查看郵遞區號的手寫字碼
來分類郵件。
01:24
Over the past過去 few少數 years年份, we have made製作
dramatic戲劇性 breakthroughs突破.
24
72036
3169
過去幾年來我們已經做出
多項重大的突破,
01:27
Machine learning學習 is now capable
of far, far more complex複雜 tasks任務.
25
75586
3916
機器學習現在已經可以完成
非常覆雜的任務。
01:31
In 2012, KaggleKaggle challenged挑戰 its community社區
26
79860
3231
在 2012 年 Kaggle
給自家社群出了一道難題,
01:35
to build建立 an algorithm算法
that could grade年級 high-school中學 essays隨筆.
27
83115
3189
要大家設計出一個演算法
來評判高中作文。
01:38
The winning勝利 algorithms算法
were able能夠 to match比賽 the grades等級
28
86328
2604
獲勝的演算法給出的分數居然
和真正老師給出的分數相符
01:40
given特定 by human人的 teachers教師.
29
88956
1665
去年,我們出了一道更難的題目:
01:43
Last year, we issued發行
an even more difficult challenge挑戰.
30
91092
2984
01:46
Can you take images圖片 of the eye
and diagnose診斷 an eye disease疾病
31
94100
2953
你可不可以藉由眼球的影像
診斷出一種叫「糖尿病視網膜病變」的眼疾?
01:49
called diabetic糖尿病患者 retinopathy視網膜病變?
32
97077
1694
01:51
Again, the winning勝利 algorithms算法
were able能夠 to match比賽 the diagnoses診斷
33
99164
4040
果然,獲勝的演算法給出的診斷
可以和人類眼科醫師的診斷相媲美。
01:55
given特定 by human人的 ophthalmologists眼科醫生.
34
103228
1825
01:57
Now, given特定 the right data數據,
machines are going to outperform跑贏大市 humans人類
35
105561
3212
只要給定正確的數據 ,
機器在類似的任務中
02:00
at tasks任務 like this.
36
108797
1165
將完全超越人類。
02:01
A teacher老師 might威力 read 10,000 essays隨筆
over a 40-year-年 career事業.
37
109986
3992
一位老師,在他的40年職業生涯中
也許只能審閱10000篇作文
02:06
An ophthalmologist眼科醫生 might威力 see 50,000 eyes眼睛.
38
114407
2360
一名眼科醫生,大概可以看50,000隻眼睛
02:08
A machine can read millions百萬 of essays隨筆
or see millions百萬 of eyes眼睛
39
116791
3913
但一部機器可以在短短幾分鐘內
讀完上百萬篇文章
02:12
within minutes分鐘.
40
120728
1276
或是看完上百萬顆眼睛。
02:14
We have no chance機會 of competing競爭
against反對 machines
41
122456
2858
在頻繁、大批量的任務上
我們無法與機器抗衡。
02:17
on frequent頻繁, high-volume高音量 tasks任務.
42
125338
2321
02:20
But there are things we can do
that machines can't do.
43
128665
3724
不過還是有我們能做
而機器做不到的事情,
02:24
Where machines have made製作
very little progress進展
44
132791
2200
機器在解決複雜的新情況方面
02:27
is in tackling搶斷 novel小說 situations情況.
45
135015
1854
進展甚微。
02:28
They can't handle處理 things
they haven't沒有 seen看到 many許多 times before.
46
136893
3899
它們對還沒看到很多次的事情無法掌握。
02:33
The fundamental基本的 limitations限制
of machine learning學習
47
141321
2584
機器學習的先天限制就是:
02:35
is that it needs需求 to learn學習
from large volumes of past過去 data數據.
48
143929
3394
它需要從大量的過往資料中學習。
02:39
Now, humans人類 don't.
49
147347
1754
人類就不一樣了,
02:41
We have the ability能力 to connect
seemingly似乎 disparate不同 threads線程
50
149125
3030
我們有一種能把看似毫不相關的事物
聯系起來的能力,
02:44
to solve解決 problems問題 we've我們已經 never seen看到 before.
51
152179
2238
從而解決我們先前還不曾見過的難題。
02:46
Percy珀西 Spencer斯賓塞 was a physicist物理學家
working加工 on radar雷達 during World世界 War戰爭 IIII,
52
154808
4411
波西‧史賓塞是二次世界大戰期間,
從事雷達研究的物理學家,
02:51
when he noticed注意到 the magnetron磁控
was melting融化 his chocolate巧克力 bar酒吧.
53
159243
3013
當他注意到磁控管不斷融化
他的巧克力棒時,
02:54
He was able能夠 to connect his understanding理解
of electromagnetic電磁 radiation輻射
54
162970
3295
他能夠把他對電磁波的認知
02:58
with his knowledge知識 of cooking烹飪
55
166289
1484
與烹飪的知識做結合,
02:59
in order訂購 to invent發明 -- any guesses猜測? --
the microwave微波 oven烤箱.
56
167797
3258
因此發明了--各位猜猜是什麽?
微波爐。
03:03
Now, this is a particularly尤其 remarkable卓越
example of creativity創造力.
57
171444
3073
這是個特別傑出的創新例子
03:06
But this sort分類 of cross-pollination異花受粉
happens發生 for each of us in small ways方法
58
174541
3664
但是這種跨領域的碰撞,
每天在我們的周遭會上演好幾千回。
03:10
thousands數千 of times per day.
59
178229
1828
03:12
Machines cannot不能 compete競爭 with us
60
180501
1661
在解決新的棘手問題方面
機器無法與我們媲美,
03:14
when it comes to tackling搶斷
novel小說 situations情況,
61
182186
2251
03:16
and this puts看跌期權 a fundamental基本的 limit限制
on the human人的 tasks任務
62
184461
3117
而這使機器自動化取代人工
03:19
that machines will automate自動化.
63
187602
1717
受到了限制。
03:22
So what does this mean
for the future未來 of work?
64
190041
2405
那麽這對未來的工作意味著什麽呢?
03:24
The future未來 state of any single job工作 lies
in the answer回答 to a single question:
65
192804
4532
未來工作的狀態完全取決於一個問題:
03:29
To what extent程度 is that job工作 reducible還原
to frequent頻繁, high-volume高音量 tasks任務,
66
197360
4981
「該工作有多少程度可以縮減成
經常性、高產量的任務,
03:34
and to what extent程度 does it involve涉及
tackling搶斷 novel小說 situations情況?
67
202365
3253
以及有多少程度是在解決新的棘手問題?」
03:37
On frequent頻繁, high-volume高音量 tasks任務,
machines are getting得到 smarter聰明 and smarter聰明.
68
205975
4035
對於那些頻繁,大批量的任務,
機器變得越來越聰明。
03:42
Today今天 they grade年級 essays隨筆.
They diagnose診斷 certain某些 diseases疾病.
69
210034
2714
今天它們能給作文打分數、
診斷特定疾病,
03:44
Over coming未來 years年份,
they're going to conduct進行 our audits審計,
70
212772
3157
過了幾年後,它們將可以進行審計、
03:47
and they're going to read boilerplate樣板
from legal法律 contracts合同.
71
215953
2967
從法律合約中解讀法律語言。
03:50
Accountants會計師 and lawyers律師 are still needed需要.
72
218944
1997
盡管會計師和律師還是需要的
03:52
They're going to be needed需要
for complex複雜 tax structuring結構,
73
220965
2682
但僅能研究覆雜的稅務結構及
無例可循的法律問題,
03:55
for pathbreaking開創性 litigation訴訟.
74
223671
1357
03:57
But machines will shrink收縮 their ranks行列
75
225052
1717
不過機器將會減少他們的就業機會,
增加就業難度。
03:58
and make these jobs工作 harder更難 to come by.
76
226793
1872
04:00
Now, as mentioned提到,
77
228689
1151
如同我說過的:
04:01
machines are not making製造 progress進展
on novel小說 situations情況.
78
229864
2949
機器在處理複雜的新情境上
沒有進步!
04:04
The copy複製 behind背後 a marketing營銷 campaign運動
needs需求 to grab consumers'消費者 attention注意.
79
232837
3457
行銷推案的文宣必須擄獲消費者的青睞,
04:08
It has to stand out from the crowd人群.
80
236318
1715
它必須脫俗出眾。
04:10
Business商業 strategy戰略 means手段
finding發現 gaps空白 in the market市場,
81
238057
2444
商業策略必須在市場上找到一些
04:12
things that nobody沒有人 else其他 is doing.
82
240525
1756
其它人還沒開始做的領域。
04:14
It will be humans人類 that are creating創建
the copy複製 behind背後 our marketing營銷 campaigns活動,
83
242305
4118
人類才是營銷文案的創造者,
04:18
and it will be humans人類 that are developing發展
our business商業 strategy戰略.
84
246447
3517
人類才是商業戰略的拓展人
04:21
So YahliYahli, whatever隨你 you decide決定 to do,
85
249988
2817
所以,雅莉,不管妳決定要做什麼,
04:24
let every一切 day bring帶來 you a new challenge挑戰.
86
252829
2361
讓每一天帶給妳新的挑戰,
04:27
If it does, then you will stay
ahead of the machines.
87
255587
2809
如果是這樣,那麼妳將永遠領先機器一步。
04:31
Thank you.
88
259126
1176
謝謝大家!
04:32
(Applause掌聲)
89
260326
3104
(掌聲)
Translated by Liv Ran
Reviewed by Yi-Fan Yu

▲Back to top

ABOUT THE SPEAKER
Anthony Goldbloom - Machine learning expert
Anthony Goldbloom crowdsources solutions to difficult problems using machine learning.

Why you should listen

Anthony Goldbloom is the co-founder and CEO of Kaggle. Kaggle hosts machine learning competitions, where data scientists download data and upload solutions to difficult problems. Kaggle has a community of over 600,000 data scientists and has worked with companies ranging Facebook to GE on problems ranging from predicting friendships to flight arrival times.

Before Kaggle, Anthony worked as an econometrician at the Reserve Bank of Australia, and before that the Australian Treasury. In 2011 and 2012, Forbes named Anthony one of the 30 under 30 in technology; in 2013 the MIT Tech Review named him one of top 35 innovators under the age of 35, and the University of Melbourne awarded him an Alumni of Distinction Award. He holds a first call honors degree in Econometrics from the University of Melbourne.  

More profile about the speaker
Anthony Goldbloom | Speaker | TED.com