TED2017
Tom Gruber: How AI can enhance our memory, work and social lives
湯姆葛魯柏: 人工智慧如何強化我們的記憶、工作、及社交生活
Filmed:
Readability: 4.6
2,030,318 views
我們的機器能讓我們變多聰明？ Siri 的共同創造者湯姆葛魯柏想要做的是「人性化的人工智慧」，它會增強我們的能力並與我們合作，而不是與我們競爭或取代我們。他分享了他對未來的願景，在未來，人工智慧會協助我們在感知、創造力、認知功能上有超人般的表現，從大大增強我們的設計技巧，並協助我們記住我們所有閱讀過的內容及見過的每個人。「我們正在人工智慧的文藝復興時代當中，」葛魯柏說：「每次當機器變更聰明，我們也會變更聰明。」
Tom Gruber - AI product designer
As co-creator of Siri, Tom Gruber helped redefine the role of machine intelligence in our lives and transformed the way we interact with our devices. Full bio
As co-creator of Siri, Tom Gruber helped redefine the role of machine intelligence in our lives and transformed the way we interact with our devices. Full bio
Double-click the English transcript below to play the video.
00:13
I'm here to offer you
a new way to think about my field,
a new way to think about my field,
0
1040
3751
今天來到這裡，是想邀請大家
從一個新的角度來看我熟悉的領域：
從一個新的角度來看我熟悉的領域：
00:16
artificial intelligence.
1
4815
1768
人工智慧。
00:18
I think the purpose of AI
2
6607
2014
我認為人工智慧的目的，
00:20
is to empower humans
with machine intelligence.
with machine intelligence.
3
8645
3434
是要用機器智慧來增強人類的能力。
00:25
And as machines get smarter,
4
13286
2450
而隨著機器變得更聰明，
00:27
we get smarter.
5
15760
1253
我們也會更聰明。
00:29
I call this "humanistic AI" --
6
17907
2512
我稱這個為「人性化人工智慧」──
00:32
artificial intelligence
designed to meet human needs
designed to meet human needs
7
20443
3671
透過與人工智慧合作並
強化人類的設計方式
強化人類的設計方式
00:36
by collaborating and augmenting people.
8
24138
2931
來滿足人類的需求。
00:40
Now, today I'm happy to see
9
28149
1910
如今，我很高興能看到
00:42
that the idea of an intelligent assistant
10
30083
2177
智慧助理的這個想法
00:44
is mainstream.
11
32284
1469
能成為主流。
00:45
It's the well-accepted metaphor
for the interface between humans and AI.
for the interface between humans and AI.
12
33777
5060
這象徵了人類與人工智慧之間的
互動介面已被廣泛接受。
互動介面已被廣泛接受。
00:51
And the one I helped create
is called Siri.
is called Siri.
13
39556
2697
而其中一個
我協助創造的就是 Siri。
我協助創造的就是 Siri。
00:54
You know Siri.
14
42957
1152
各位都認識 Siri。
00:56
Siri is the thing that knows your intent
15
44133
2645
Siri 會知道你的意圖，
00:58
and helps you do it for you,
16
46802
2042
並協助你
01:00
helps you get things done.
17
48868
1501
把一切搞定。
01:02
But what you might not know
is that we designed Siri
is that we designed Siri
18
50393
2474
但各位可能不知道，
當初我們在設計 Siri 時，
當初我們在設計 Siri 時，
01:05
as humanistic AI,
19
53645
1966
就把它定位在
「人性化的人工智慧」，
「人性化的人工智慧」，
01:07
to augment people
with a conversational interface
with a conversational interface
20
55635
3171
透過對話介面的方式來強化人類，
01:10
that made it possible for them
to use mobile computing,
to use mobile computing,
21
58830
3453
讓使用行動計算成為可能，
01:14
regardless of who they were
and their abilities.
and their abilities.
22
62307
3284
不論他們是誰、
能力如何，都可以使用。
能力如何，都可以使用。
01:18
Now for most of us,
23
66814
1577
對我們大多數人而言，
01:20
the impact of this technology
24
68415
2781
受到這項技術的影響
01:23
is to make things
a little bit easier to use.
a little bit easier to use.
25
71220
2283
是讓我們在使用東西上
變得比較容易。
變得比較容易。
01:25
But for my friend Daniel,
26
73916
2179
但對我朋友丹尼爾而言，
01:29
the impact of the AI
in these systems is a life changer.
in these systems is a life changer.
27
77352
3171
卻是改變了他一生的人工智慧系統。
01:33
You see, Daniel is a really social guy,
28
81217
4833
丹尼爾是個很愛社交的人，
01:38
and he's blind and quadriplegic,
29
86074
1759
他也是個四肢麻痺的盲人，
01:40
which makes it hard to use those devices
that we all take for granted.
that we all take for granted.
30
88614
3859
我們認為理所當然的裝置設備，
對他來說卻很難用。
對他來說卻很難用。
01:44
The last time I was at his house,
his brother said,
his brother said,
31
92497
2442
上次我去他家，他哥哥說：
「等等，丹尼爾還沒準備好。
01:46
"Hang on a second, Daniel's not ready.
32
94963
1858
他正在和他線上認識的女生講電話。」
01:48
He's on the phone with a woman
he met online."
he met online."
33
96845
2342
01:52
I'm like, "That's cool, how'd he do it?"
34
100250
2010
我說：「好酷，他怎麼辦到的？」
01:54
Well, Daniel uses Siri
to manage his own social life --
to manage his own social life --
35
102284
3494
丹尼爾用 Siri 來管理
他自己的社交生活──
他自己的社交生活──
01:58
his email, text and phone --
36
106731
1940
他的電子郵件、訊息、電話──
02:00
without depending on his caregivers.
37
108695
2476
不用依賴他的照護者。
02:04
This is kind of interesting, right?
38
112501
1738
這挺有趣的，對吧？
02:06
The irony here is great.
39
114263
1317
這也相當諷刺。
02:07
Here's the man whose relationship with AI
40
115604
2896
這個人與人工智慧的關係
02:10
helps him have relationships
with genuine human beings.
with genuine human beings.
41
118524
3177
協助他與真正的人類建立了關係。
02:14
And this is humanistic AI.
42
122805
3130
這就是人性化的人工智慧。
02:19
Another example with
life-changing consequences
life-changing consequences
43
127404
2398
還有一個可以
改變人生結果的例子，
改變人生結果的例子，
02:21
is diagnosing cancer.
44
129826
1257
也就是癌症診斷。
02:24
When a doctor suspects cancer,
45
132795
2155
當醫生懷疑你有癌症時，
02:26
they take a sample
and send it to a pathologist,
and send it to a pathologist,
46
134974
2703
他們會先取樣交給病理學家，
02:29
who looks at it under a microscope.
47
137701
1952
病理學家再利用顯微鏡來看樣本。
02:32
Now, pathologists look at
hundreds of slides
hundreds of slides
48
140145
2525
病理學家每天要看數百片的載玻片
02:34
and millions of cells every day.
49
142694
2324
及數百萬個細胞。
02:37
So to support this task,
50
145630
1387
所以為了支援這項工作，
02:39
some researchers made an AI classifier.
51
147041
3442
有幾位研究人員做出了一種
人工智慧分類器。
人工智慧分類器。
02:43
Now, the classifier says,
"Is this cancer or is this not cancer?"
"Is this cancer or is this not cancer?"
52
151548
4553
現在，分類器已經會看著圖片說：
「這是不是癌症？」
02:48
looking at the pictures.
53
156125
1305
02:50
The classifier was pretty good,
54
158355
1564
分類器挺厲害的，
02:52
but not as good as the person,
55
160967
1685
但沒有人類那麼厲害，
02:55
who got it right most of the time.
56
163945
1768
人類大部份都能判讀正確。
02:58
But when they combine the ability
of the machine and the human together,
of the machine and the human together,
57
166427
6180
但，當機器與人類的能力
結合在一起時，
結合在一起時，
03:04
accuracy went to 99.5 percent.
58
172631
2995
正確率可以達到 99.5%。
03:09
Adding that AI to a partnership
eliminated 85 percent of the errors
eliminated 85 percent of the errors
59
177150
5001
加上人工智慧的協作，
人類病理學家
人類病理學家
03:14
that the human pathologist
would have made working alone.
would have made working alone.
60
182175
3686
能減少個人判斷時 85% 的錯誤。
03:18
That's a lot of cancer
that would have otherwise gone untreated.
that would have otherwise gone untreated.
61
186820
3159
這相當多啊，很多癌症患者，
當初就是因誤判而耽擱了治療。
當初就是因誤判而耽擱了治療。
03:23
Now, for the curious, it turns out
62
191097
1930
人類因為好奇的原因，所以
03:25
that the human was better
at rejecting false positives,
at rejecting false positives,
63
193051
3474
比較擅長判斷錯誤的假陽性案例，
03:28
and the machine was better
at recognizing those hard-to-spot cases.
at recognizing those hard-to-spot cases.
64
196549
3572
而機器比較擅長判斷
難以辨別的案例。
難以辨別的案例。
03:32
But the lesson here isn't about
which agent is better
which agent is better
65
200145
3237
但我們學到的並不是
在比誰在圖像分類上比較厲害。
03:35
at this image-classification task.
66
203406
2078
03:37
Those things are changing every day.
67
205888
1839
那些東西每天都會變。
03:39
The lesson here
68
207751
1568
這裡我們學到的是，
03:41
is that by combining the abilities
of the human and machine,
of the human and machine,
69
209343
3360
藉由結合人類與機器的能力，
03:44
it created a partnership
that had superhuman performance.
that had superhuman performance.
70
212727
4951
其創造出的合作關係
會有超人般的表現。
會有超人般的表現。
03:50
And that is humanistic AI.
71
218543
2940
而這就是人性化的人工智慧 。
03:54
Now let's look at another example
72
222787
1986
我們再舉個例子，
03:56
with turbocharging performance.
73
224797
2287
這方面的進步也是突飛猛進。
03:59
This is design.
74
227108
1403
那就是「設計。」
04:00
Now, let's say you're an engineer.
75
228535
2041
假設你是個工程師，
04:02
You want to design a new frame for a drum.
76
230600
2117
你想要設計一個新的無人機骨架。
04:04
You get out your favorite
software tools, CAD tools,
software tools, CAD tools,
77
232741
3064
你拿出你最愛的
軟體工具，電腦繪圖輔助工具，
軟體工具，電腦繪圖輔助工具，
04:07
and you enter the form and the materials,
and then you analyze performance.
and then you analyze performance.
78
235829
4215
你輸入了外形和材料，
然後按下分析表現。
然後按下分析表現。
04:12
That gives you one design.
79
240068
1344
它只會跑出一種設計方案。
04:14
If you give those same tools to an AI,
80
242162
3677
如果你把這些工具
拿給人工智慧使用，
拿給人工智慧使用，
04:17
it can generate thousands of designs.
81
245863
3124
它會幫你跑出數千種設計。
04:22
This video by Autodesk is amazing.
82
250193
1934
歐特克做的影片非常震撼。
04:24
This is real stuff.
83
252151
1300
這是真實案例。
04:26
So this transforms how we do design.
84
254092
3071
這改變了我們設計的方式。
04:29
The human engineer now
85
257187
1428
現在，人類工程師
04:30
says what the design should achieve,
86
258639
3315
只要說出想要達成的設計，
04:33
and the machine says,
87
261978
1156
機器就會說：
04:35
"Here's the possibilities."
88
263158
1370
「有這些可能性方案。」
04:37
Now in her job, the engineer's job
89
265758
2940
現在，她的工作，工程師的工作，
04:40
is to pick the one that best meets
the goals of the design,
the goals of the design,
90
268722
3100
就是選出一種最符合目標的設計，
04:43
which she knows as a human
better than anyone else,
better than anyone else,
91
271846
3339
這是她身為人類最擅長的，
04:47
using human judgment and expertise.
92
275209
1794
用的就是人類的判斷和專業。
04:49
In this case, the winning form
93
277414
1580
在這案例中，脫穎而出的設計
04:51
looks kind of like something
nature would have designed,
nature would have designed,
94
279018
2711
像是自然界設計出來的，
04:53
minus a few million years of evolution
95
281753
2115
但這樣的設計，
不需要幾百萬年的演化
不需要幾百萬年的演化
04:55
and all that unnecessary fur.
96
283892
2381
也不需有那些皮毛。
05:00
Now let's see where this idea
of humanistic AI might lead us
of humanistic AI might lead us
97
288424
4672
我們現在來想想，
如果我們跟著人性化人工智慧走，
如果我們跟著人性化人工智慧走，
05:05
if we follow it into
the speculative beyond.
the speculative beyond.
98
293120
2393
它會把我們帶到那些
意想不到的境界？
意想不到的境界？
05:08
What's a kind of augmentation
that we would all like to have?
that we would all like to have?
99
296118
3004
我們想要的強化，有哪些方面？
05:11
Well, how about cognitive enhancement?
100
299529
2906
認知強化如何？
05:15
Instead of asking,
101
303904
1479
別再問：
05:17
"How smart can we make our machines?"
102
305407
2061
「我們能把機器做到多聰明？」
05:19
let's ask
103
307492
1173
我們要改問：
05:20
"How smart can our machines make us?"
104
308689
2705
「我們的機器能把我們變多聰明？」
05:25
I mean, take memory for example.
105
313105
1780
我的意思是，例如「記憶力。」
05:27
Memory is the foundation
of human intelligence.
of human intelligence.
106
315511
2946
記憶是人類智慧的基礎。
05:31
But human memory is famously flawed.
107
319767
2971
但人類的記憶力是出名的差。
05:35
We're great at telling stories,
108
323541
2074
我們很會說故事，
05:37
but not getting the details right.
109
325639
1951
但無法把細節說正確。
05:39
And our memories -- they decay over time.
110
327614
3206
我們的記憶力會隨時間衰退。
05:42
I mean, like, where did the '60s go,
and can I go there, too?
and can I go there, too?
111
330844
3232
就像六十年代去哪了？
我也能去那兒嗎？
我也能去那兒嗎？
05:46
(Laughter)
112
334100
1467
（笑聲）
05:48
But what if you could have a memory
that was as good as computer memory,
that was as good as computer memory,
113
336648
4304
但如果你能擁有和電腦
一樣好的記憶力，
一樣好的記憶力，
05:54
and was about your life?
114
342158
2859
而且這些記憶與你的人生有關，
那會如何呢？
那會如何呢？
05:57
What if you could remember
every person you ever met,
every person you ever met,
115
345714
2789
如果你能記得你見過的每一個人、
06:00
how to pronounce their name,
116
348527
1946
叫得出每一個人的名字、
06:02
their family details,
their favorite sports,
their favorite sports,
117
350497
2092
家庭狀況的細節、他們最愛的運動、
06:04
the last conversation you had with them?
118
352613
1963
你和他們上次的對話，那會如何呢？
06:06
If you had this memory all your life,
119
354600
2655
如果你一生都有這樣的記憶，
06:09
you could have the AI look
at all the interactions
at all the interactions
120
357279
3298
你就可以要求人工智慧
去看看你過去
去看看你過去
06:12
you had with people over time
121
360601
1417
與人們的所有互動，
06:14
and help you reflect on the long arc
of your relationships.
of your relationships.
122
362042
3793
並協助你反思你的人際關係弧線。
06:18
What if you could have the AI read
everything you've ever read
everything you've ever read
123
366530
5066
如果我們能讓人工智慧去閱讀
所有你閱讀過的東西、
所有你閱讀過的東西、
06:23
and listen to every song
you've ever heard?
you've ever heard?
124
371620
2514
去聽你聽過的每首歌，會如何呢？
06:27
From the tiniest clue,
it could help you retrieve
it could help you retrieve
125
375344
3739
從最微小的線索，它就能協助你找回
06:31
anything you've ever seen or heard before.
126
379107
2113
你以前看過或聽過的任何東西。
06:33
Imagine what that would do
for the ability to make new connections
for the ability to make new connections
127
381931
3250
想像一樣，那對於
建立新連結以及產生新點子的
能力會有什麼影響。
能力會有什麼影響。
06:37
and form new ideas.
128
385205
1580
06:38
And what about our bodies?
129
386809
1988
那麼，我們的身體呢？
06:41
What if we could remember the consequences
of every food we eat,
of every food we eat,
130
389732
4278
如果我們能記得
我們吃過的每樣食物、
我們吃過的每樣食物、
06:46
every pill we take,
131
394034
1501
每顆藥物、
06:48
every all-nighter we pull?
132
396485
1835
記得每次熬夜 K 書的內容，
那會如何？
那會如何？
06:51
We could do our own science
on our own data
on our own data
133
399048
2710
我們可以用自己的資料
來做自己的科學，
來做自己的科學，
06:53
about what makes us feel
good and stay healthy.
good and stay healthy.
134
401782
3860
找出什麼會讓我們
感覺很好並保持健康。
感覺很好並保持健康。
06:57
And imagine how this could revolutionize
135
405666
2476
想像一下，這會為我們管理
07:00
the way we manage
allergies and chronic disease.
allergies and chronic disease.
136
408166
2690
過敏及慢性病帶來怎樣的革命。
07:04
I believe that AI will make
personal memory enhancement a reality.
personal memory enhancement a reality.
137
412909
4974
我相信，人工智慧能夠
實現個人記憶的強化。
實現個人記憶的強化。
07:11
I can't say when or what
form factors are involved,
form factors are involved,
138
419122
2824
我不知道何時或需要那些機緣，
07:14
but I think it's inevitable,
139
422627
1963
但我認為它是一定會到來的，
07:17
because the very things
that make AI successful today --
that make AI successful today --
140
425923
4827
因為人工智慧在現今
能夠成功的原因，
能夠成功的原因，
07:22
the availability of comprehensive data
141
430774
2493
就是這些綜合性資料的可取得性，
07:25
and the ability for machines
to make sense of that data --
to make sense of that data --
142
433291
3548
以及機器能從這些資料中
理解出意義的能力，
理解出意義的能力，
07:28
can be applied to the data of our lives.
143
436863
3059
能為我們生活上的資料所用。
07:32
And those data are here today,
available for all of us,
available for all of us,
144
440505
3487
我們現今就已經能夠
取得那些資料了，
取得那些資料了，
07:36
because we lead digitally mediated lives,
145
444016
3868
因為我們可以直接
用手機或上網的方式，
用手機或上網的方式，
07:40
in mobile and online.
146
448570
2135
將我們的生活與數位科技做連結。
07:43
In my view, a personal memory
is a private memory.
is a private memory.
147
451833
3890
依我所見，個人的記憶
是私人的記憶。
是私人的記憶。
07:47
We get to choose what is and is not
recalled and retained.
recalled and retained.
148
455747
3937
我們可以選擇要或不要，
回憶或保留哪些記憶。
回憶或保留哪些記憶。
07:52
It's absolutely essential
that this be kept very secure.
that this be kept very secure.
149
460517
3664
保持資料的安全絕對是有必要的。
07:57
Now for most of us,
150
465294
1462
對大部份人來說，
07:58
the impact of augmented personal memory
151
466780
2893
增強個人記憶希望產生的影響會是：
08:01
will be a more improved mental gain,
152
469697
2735
可以獲得心智上的成長
08:04
maybe, hopefully, a bit more social grace.
153
472456
2455
或是希望可以有更高雅的社交能力。
08:08
But for the millions who suffer
from Alzheimer's and dementia,
from Alzheimer's and dementia,
154
476658
4647
但對於數百萬受阿茲海默症
及癡呆症所苦的人而言，
及癡呆症所苦的人而言，
08:15
the difference that augmented
memory could make
memory could make
155
483209
2203
增加記憶能造成的改變，
08:17
is a difference
between a life of isolation
between a life of isolation
156
485436
2461
就是讓那孤獨的人生，
08:21
and a life of dignity and connection.
157
489016
1779
變成有尊嚴及連結的人生。
08:24
We are in the middle of a renaissance
in artificial intelligence right now.
in artificial intelligence right now.
158
492166
5647
我們現在正在人工智慧的
文藝復興時期當中，
文藝復興時期當中，
08:30
I mean, in just the past few years,
159
498652
1971
僅僅在過去幾年，
08:32
we're beginning to see
solutions to AI problems
solutions to AI problems
160
500647
3631
我們就開始看到
人工智慧問題的解決方法，
人工智慧問題的解決方法，
08:36
that we have struggled
with literally for decades:
with literally for decades:
161
504302
3970
這在過去數十年來，
一直是無法解決的問題，像是：
一直是無法解決的問題，像是：
08:41
speech understanding, text understanding,
162
509226
2667
口語理解、文字理解、
08:43
image understanding.
163
511917
1320
圖像理解。
08:46
We have a choice in how we use
this powerful technology.
this powerful technology.
164
514177
4075
我們可以選擇要
如何使用這強大的技術。
如何使用這強大的技術。
08:51
We can choose to use AI
to automate and compete with us,
to automate and compete with us,
165
519268
4864
我們可以選擇用人工智慧
來做自動化並與我們競爭、
來做自動化並與我們競爭、
08:56
or we can use AI to augment
and collaborate with us,
and collaborate with us,
166
524156
3977
或我們可以選擇用人工智慧
來做增強化並與我們合作，
來做增強化並與我們合作，
09:00
to overcome our cognitive limitations
167
528157
3071
來克服我們的認知限制，
09:03
and to help us do what we want to do,
168
531252
3162
並協助我們去做我們想要做的事，
09:06
only better.
169
534438
1291
而且做得更好。
09:08
And as we discover new ways
to give machines intelligence,
to give machines intelligence,
170
536339
6504
隨著我們發掘出新的方式
來將智慧賦予機器，
來將智慧賦予機器，
09:14
we can distribute that intelligence
171
542867
3069
我們可以把那智慧移轉到
09:17
to all of the AI assistants in the world,
172
545960
2381
全世界所有的人工智慧助理身上，
09:20
and therefore to every person,
regardless of circumstance.
regardless of circumstance.
173
548365
3534
且不論如何，都要確保
每個人都能因此受惠。
每個人都能因此受惠。
09:25
And that is why,
174
553189
1529
那就是為什麼
09:26
every time a machine gets smarter,
175
554742
3161
每次當機器變更聰明，
09:30
we get smarter.
176
558628
1404
我們也會變更聰明。
09:33
That is an AI worth spreading.
177
561028
3733
而這就是人工智慧值得推廣的原因。
09:36
Thank you.
178
564785
1182
謝謝大家。
09:37
(Applause)
179
565991
3973
（掌聲）
ABOUT THE SPEAKER
Tom Gruber - AI product designerAs co-creator of Siri, Tom Gruber helped redefine the role of machine intelligence in our lives and transformed the way we interact with our devices.
Why you should listen
By connecting humans and machines with AI, designer, inventor and polymath Tom Gruber is opening up new ways to improve our lives and augment human intelligence.
Gruber led the team that revolutionized human-machine interaction with Siri, the intelligent personal assistant that can understand your spoken language and help you get things done. Launched in 2010, Siri is now used billions of times a week in more than 30 countries around the world.
Tom Gruber | Speaker | TED.com