ABOUT THE SPEAKER
Jeff Hawkins - Computer designer, brain researcher
Jeff Hawkins pioneered the development of PDAs such as the Palm and Treo. Now he's trying to understand how the human brain really works, and adapt its method -- which he describes as a deep system for storing memory -- to create new kinds of computers and tools.

Why you should listen

Jeff Hawkins' Palm PDA became such a widely used productivity tool during the 1990s that some fanatical users claimed it replaced their brains. But Hawkins' deepest interest was in the brain itself. So after the success of the Palm and Treo, which he brought to market at Handspring, Hawkins delved into brain research at the Redwood Center for Theoretical Neuroscience in Berkeley, Calif., and a new company called Numenta.

Hawkins' dual goal is to achieve an understanding of how the human brain actually works -- and then develop software to mimic its functionality, delivering true artificial intelligence. In his book On Intelligence (2004) he lays out his compelling, controversial theory: Contrary to popular AI wisdom, the human neocortex doesn't work like a processor; rather, it relies on a memory system that stores and plays back experiences to help us predict, intelligently, what will happen next. He thinks that "hierarchical temporal memory" computer platforms, which mimic this functionality (and which Numenta might pioneer), could enable groundbreaking new applications that could powerfully extend human intelligence.

More profile about the speaker
Jeff Hawkins | Speaker | TED.com
TED2003

Jeff Hawkins: How brain science will change computing

Jeff Hawkins - 大脑研究将改变计算机科学

Filmed:
1,674,773 views

手提微型电脑 Treo 的发明者 Jeff Hawkins 建议学者们从另一个角度去研究大脑,传统看法把大脑看成中央处理器,他建议把大脑看成一个记忆体,用于贮存和重演经历来有智慧地预计将会发生的事情
- Computer designer, brain researcher
Jeff Hawkins pioneered the development of PDAs such as the Palm and Treo. Now he's trying to understand how the human brain really works, and adapt its method -- which he describes as a deep system for storing memory -- to create new kinds of computers and tools. Full bio

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

00:25
I do two things: I design设计 mobile移动 computers电脑 and I study研究 brains大脑.
0
0
3000
我有两个专业,设计微型电脑和研究大脑
00:29
And today's今天的 talk is about brains大脑 and,
1
4000
2000
今天的演说是关于大脑的
00:31
yay好极了, somewhere某处 I have a brain fan风扇 out there.
2
6000
2000
嘿,我们听众里面好像有大脑研究的粉丝
00:33
(Laughter笑声)
3
8000
2000
(笑声)
00:35
I'm going to, if I can have my first slide滑动 up here,
4
10000
2000
请把我演说的首页播放
00:37
and you'll你会 see the title标题 of my talk and my two affiliations隶属关系.
5
12000
4000
你们可以看到我演说的标题和我的两个专业资格
00:41
So what I'm going to talk about is why we don't have a good brain theory理论,
6
16000
4000
我会先说为什么我们没有一个好的大脑理论
00:45
why it is important重要 that we should develop发展 one and what we can do about it.
7
20000
3000
研究出一个大脑理论的重要性和怎么应用
00:48
And I'll try to do all that in 20 minutes分钟. I have two affiliations隶属关系.
8
23000
3000
我会尝试在20分钟内完成。我有两个职业
00:51
Most of you know me from my Palm棕榈 and Handspring翻筋斗 days,
9
26000
3000
你们可能认识我其中的职业和我的发明,Palm 和 Handspring 掌上电脑
00:54
but I also run a nonprofit非营利性 scientific科学 research研究 institute研究所
10
29000
3000
但我还有一个非盈利的研究院 :
00:57
called the Redwood红木 Neuroscience神经科学 Institute研究所 in Menlo门罗 Park公园,
11
32000
2000
位于 美国 Menlo Park的Redwood(红木)神经系统科学研究院
00:59
and we study研究 theoretical理论 neuroscience神经科学,
12
34000
2000
在那里我们研究神经系统科学理论
01:01
and we study研究 how the neocortex新皮层 works作品.
13
36000
2000
和研究 新(大脑)皮层 是怎么运作的
01:03
I'm going to talk all about that.
14
38000
2000
我将会讲解有关的研究
01:05
I have one slide滑动 on my other life, the computer电脑 life, and that's the slide滑动 here.
15
40000
3000
我有一页演说是关于我电脑方面的工作,这张就是
01:08
These are some of the products制品 I've worked工作 on over the last 20 years年份,
16
43000
3000
这些是我在近 20 年来设计过的电子产品
01:11
starting开始 back from the very original原版的 laptop笔记本电脑 to some of the first tablet片剂 computers电脑
17
46000
4000
由最早的笔记本到第一台手写笔记本
01:15
and so on, and ending结尾 up most recently最近 with the Treo的Treo,
18
50000
2000
到最近的 微型笔记本 Treo
01:17
and we're continuing继续 to do this.
19
52000
2000
而我们会继续这方面的工作
01:19
And I've doneDONE this because I really believe that mobile移动 computing计算
20
54000
2000
我干这些是因为我深信移动计算技术
01:21
is the future未来 of personal个人 computing计算, and I'm trying to make the world世界
21
56000
3000
是个人计算系统的未来,而我会尝试通过这些工作
01:24
a little bit better by working加工 on these things.
22
59000
3000
来造福人群
01:27
But this was, I have to admit承认, all an accident事故.
23
62000
2000
但我得承认这些都是巧合
01:29
I really didn't want to do any of these products制品
24
64000
2000
我其实没有想过时间这些产品
01:31
and very early in my career事业 I decided决定
25
66000
2000
而在我刚刚开始工作的时候我决定
01:33
I was not going to be in the computer电脑 industry行业.
26
68000
3000
我不会从事计算机行业
01:36
And before I tell you about that, I just have to tell you
27
71000
2000
在说那个之前,让我先告诉你
01:38
this one little picture图片 of graffiti涂鸦 there I picked采摘的 off the web卷筒纸 the other day.
28
73000
2000
我在网上找到这个小图片,
01:40
I was looking for a picture图片 of graffiti涂鸦, little text文本 input输入 language语言,
29
75000
3000
我在网上找有关涂鸦的图片,
01:43
and I found发现 the website网站 dedicated专用 to teachers教师 who want to make these,
30
78000
3000
而发现这专为教师们而设的网站
01:46
you know, the script脚本 writing写作 things across横过 the top最佳 of their blackboard黑板,
31
81000
3000
他们教学中在黑板上写的,
01:49
and they had added添加 graffiti涂鸦 to it, and I'm sorry about that.
32
84000
3000
而他们却把这涂鸦上了,真可惜,
01:52
(Laughter笑声)
33
87000
2000
(听众的笑声)
01:54
So what happened发生 was, when I was young年轻 and got out of engineering工程 school学校
34
89000
5000
经过是这样的,我还年轻的时候,刚刚从Cornell 康奈尔大学工程学院毕业
01:59
at Cornell康奈尔 in '79, I decided决定 -- I went to work for Intel英特尔 and
35
94000
4000
是 1979 年, 我决定去 Intel 英特尔工作
02:03
I was in the computer电脑 industry行业 -- and three months个月 into that,
36
98000
3000
我在从事计算机行业,3 个月后
02:06
I fell下跌 in love with something else其他, and I said, "I made制作 the wrong错误 career事业 choice选择 here,"
37
101000
4000
我爱上另一个东西,我发现我选错了行业
02:10
and I fell下跌 in love with brains大脑.
38
105000
3000
而我爱上了大脑
02:13
This is not a real真实 brain. This is a picture图片 of one, a line线 drawing画画.
39
108000
3000
这不是真的大脑, 这是一张大脑的图画
02:16
But I don't remember记得 exactly究竟 how it happened发生,
40
111000
3000
但我不太记得是怎么发生的
02:19
but I have one recollection回忆, which哪一个 was pretty漂亮 strong强大 in my mind心神.
41
114000
3000
但我还记得一段挺强烈的记忆
02:22
In September九月 1979, Scientific科学 American美国 came来了 out
42
117000
3000
1979 年 9 月,Scientific America(美国科学杂志)发表了
02:25
with a single topic话题 issue问题 about the brain. And it was quite相当 good.
43
120000
3000
一本关于大脑研究的特刊
02:28
It was one of the best最好 issues问题 ever. And they talked about the neuron神经元
44
123000
3000
那是该杂志中最好的一期。那特刊讨论脑细胞
02:31
and development发展 and disease疾病 and vision视力 and all the things
45
126000
2000
的发展,疾病,视觉和其它
02:33
you might威力 want to know about brains大脑. It was really quite相当 impressive有声有色.
46
128000
3000
关于大脑的课题。真的是很棒的
02:36
And one might威力 have the impression印象 that we really knew知道 a lot about brains大脑.
47
131000
3000
你可能认为我们对大脑很了解
02:39
But the last article文章 in that issue问题 was written书面 by Francis弗朗西斯 Crick克里克 of DNA脱氧核糖核酸 fame名誉.
48
134000
4000
特邗里最后有 Francis Crick 写有关 DNA 的文章
02:43
Today今天 is, I think, the 50th anniversary周年 of the discovery发现 of DNA脱氧核糖核酸.
49
138000
3000
今天应该是发现 DNA 的 50 周年
02:46
And he wrote a story故事 basically基本上 saying,
50
141000
2000
他(Francis Crick)写了一段,
02:48
well, this is all well and good, but you know what,
51
143000
3000
大概意思是,
02:51
we don't know diddleydiddley squat about brains大脑
52
146000
2000
我们基本上对大脑一点都不认识
02:53
and no one has a clue线索 how these things work,
53
148000
2000
而没有人知道它怎么运作
02:55
so don't believe what anyone任何人 tells告诉 you.
54
150000
2000
所以别随便相信别人说的(以为我们很了了解大脑)
02:57
This is a quote引用 from that article文章. He said, "What is conspicuously显着 lacking不足,"
55
152000
3000
他在文章里提到 ”我们现在显著地缺少的是 。。。“
03:00
he's a very proper正确 British英国的 gentleman绅士 so, "What is conspicuously显着 lacking不足
56
155000
4000
他是一个很传统的英国绅士, ”现在显著地缺少的是,
03:04
is a broad广阔 framework骨架 of ideas思路 in which哪一个 to interpret these different不同 approaches方法."
57
159000
3000
一个可以融入对大脑已经的不同想法和不同解释方式的框架“
03:07
I thought the word framework骨架 was great.
58
162000
2000
我认为’框架‘这词用的很好
03:09
He didn't say we didn't even have a theory理论. He says,
59
164000
2000
他甚至没有提到’理论‘,他说,
03:11
we don't even know how to begin开始 to think about it --
60
166000
2000
我们根本不知道怎么开始去想
03:13
we don't even have a framework骨架.
61
168000
2000
我们连框架都没有
03:15
We are in the pre-paradigm前范式 days, if you want to use Thomas托马斯 Kuhn库恩.
62
170000
3000
我们正处于 Thomas Kuhn 所说的规范前时期
03:18
And so I fell下跌 in love with this, and said look,
63
173000
3000
后来我就爱上大脑研究了,我想,
03:21
we have all this knowledge知识 about brains大脑. How hard can it be?
64
176000
3000
我们有这么多关于大脑的知识,能有多难呢?
03:24
And this is something we can work on my lifetime一生. I felt I could make a difference区别,
65
179000
3000
后来这成为我毕生的工作, 我觉得我可以有所贡献,
03:27
and so I tried试着 to get out of the computer电脑 business商业, into the brain business商业.
66
182000
4000
我尝试离开计算机行业而专注大脑研究
03:31
First, I went to MITMIT, the AIAI lab实验室 was there,
67
186000
2000
首先我去了 MIT(麻省理工学院)的人工智能研究院,
03:33
and I said, well, I want to build建立 intelligent智能 machines, too,
68
188000
2000
我想,我也想设计和制作聪明的机器,
03:35
but the way I want to do it is to study研究 how brains大脑 work first.
69
190000
3000
但我的想法是先研究大脑怎么运作
03:38
And they said, oh, you don't need to do that.
70
193000
3000
而他们说,呃,你不需要这样做
03:41
We're just going to program程序 computers电脑; that's all we need to do.
71
196000
2000
我们只需要计算机编程
03:43
And I said, no, you really ought应该 to study研究 brains大脑. They said, oh, you know,
72
198000
3000
而我说,不,你应该先研究大脑。 他们说,呃,你知道吗,
03:46
you're wrong错误. And I said, no, you're wrong错误, and I didn't get in.
73
201000
2000
你错了。而我说,不,你们错了,最后我没被取录
03:48
(Laughter笑声)
74
203000
1000
(笑声)
03:50
But I was a little disappointed失望 -- pretty漂亮 young年轻 -- but I went back again
75
205000
2000
但我真的有点失望,那时候年轻,但我再尝试
03:52
a few少数 years年份 later后来 and this time was in California加州, and I went to Berkeley伯克利.
76
207000
3000
几年后在加州的 Berkley(加州大学伯克利分校)
03:55
And I said, I'll go in from the biological生物 side.
77
210000
4000
这次我尝试去学习生物研究方面
03:59
So I got in -- in the Ph博士.D. program程序 in biophysics生物物理学, and I was, all right,
78
214000
3000
我开始攻读生物物理博士课程
04:02
I'm studying研究 brains大脑 now, and I said, well, I want to study研究 theory理论.
79
217000
3000
我在学习大脑了,而我想学理论
04:05
And they said, oh no, you can't study研究 theory理论 about brains大脑.
80
220000
2000
而他们说,不,你不可以学大脑的理论
04:07
That's not something you do. You can't get funded资助 for that.
81
222000
2000
这是不可以的,你不会拿到研究经费
04:09
And as a graduate毕业 student学生, you can't do that. So I said, oh my gosh天哪.
82
224000
4000
而作为研究生,没有研究经费是不可以的。我的天
04:13
I was very depressed郁闷. I said, but I can make a difference区别 in this field领域.
83
228000
2000
我很沮丧但我还坚信我可以在这一研究领域作出贡献
04:15
So what I did is I went back in the computer电脑 industry行业
84
230000
3000
最后我回到计算机行业
04:18
and said, well, I'll have to work here for a while, do something.
85
233000
2000
对自己说,我先工作,做些有意义的
04:20
That's when I designed设计 all those computer电脑 products制品.
86
235000
3000
就是那时候我设计了你们认识的一系列的微型电子产品
04:23
(Laughter笑声)
87
238000
1000
(笑声)
04:24
And I said, I want to do this for four years年份, make some money,
88
239000
3000
我计划干四年,挣点钱,
04:27
like I was having a family家庭, and I would mature成熟 a bit,
89
242000
4000
组织自己的家庭,我可能会成熟点
04:31
and maybe the business商业 of neuroscience神经科学 would mature成熟 a bit.
90
246000
3000
也可能那时候神经系统科学也会成熟一点了
04:34
Well, it took longer than four years年份. It's been about 16 years年份.
91
249000
3000
结果干了比四年长多了,已经大概十六年
04:37
But I'm doing it now, and I'm going to tell you about it.
92
252000
2000
但我终于做到了,而我现在告诉你们
04:39
So why should we have a good brain theory理论?
93
254000
3000
那为什么我们需要有一个好的大脑理论呢?
04:42
Well, there's lots of reasons原因 people do science科学.
94
257000
3000
嗯, 科学研究有很多目的
04:45
One is -- the most basic基本 one is -- people like to know things.
95
260000
3000
其中比较简单的是,我们喜欢了解各种的事物
04:48
We're curious好奇, and we just go out and get knowledge知识, you know?
96
263000
2000
我们好奇,而我们渴求知识
04:50
Why do we study研究 ants蚂蚁? Well, it's interesting有趣.
97
265000
2000
我们为什么研究蚂蚁?因为这个有趣
04:52
Maybe we'll learn学习 something really useful有用 about it, but it's interesting有趣 and fascinating迷人.
98
267000
3000
可能我们从中会学到一些很有用的知识,但本质上这研究很有趣
04:55
But sometimes有时, a science科学 has some other attributes属性
99
270000
2000
有时候,科学有其他本质
04:57
which哪一个 makes品牌 it really, really interesting有趣.
100
272000
2000
令它很有趣
04:59
Sometimes有时 a science科学 will tell something about ourselves我们自己,
101
274000
3000
有时候科学会告诉我们一些关于我们自己的,
05:02
it'll它会 tell us who we are.
102
277000
1000
告诉我们,我们到底是什么
05:03
Rarely很少, you know: evolution演化 did this and Copernicus哥白尼 did this,
103
278000
3000
这很罕有的,例如,进化论,哥白尼(Copernicus)
05:06
where we have a new understanding理解 of who we are.
104
281000
2000
都让我们对自身有新一层的理解
05:08
And after all, we are our brains大脑. My brain is talking to your brain.
105
283000
4000
毕竟,我们就是我们的大脑。我的大脑正在跟你们的大脑沟通
05:12
Our bodies身体 are hanging along沿 for the ride, but my brain is talking to your brain.
106
287000
3000
我们的身体只是随行的部分,但我的大脑正在跟你们的大脑沟通
05:15
And if we want to understand理解 who we are and how we feel and perceive感知,
107
290000
3000
如果我们想了解我们是什么和我们怎么去感受和察觉
05:18
we really understand理解 what brains大脑 are.
108
293000
2000
我们就先要明白大脑是什么
05:20
Another另一个 thing is sometimes有时 science科学
109
295000
2000
又有时候科学会
05:22
leads引线 to really big societal社会的 benefits好处 and technologies技术,
110
297000
2000
让我们有新的科技和为社会带来很大好处
05:24
or businesses企业, or whatever随你, that come out of it. And this is one, too,
111
299000
2000
甚至商业,和其它。 而大脑科学研究也会有这些好处
05:26
because when we understand理解 how brains大脑 work, we're going to be able能够
112
301000
3000
因为如果我们明白了大脑怎么运作,我们就可以
05:29
to build建立 intelligent智能 machines, and I think that's actually其实 a good thing on the whole整个,
113
304000
3000
制作有智能的机器,而这总体来说是好的
05:32
and it's going to have tremendous巨大 benefits好处 to society社会,
114
307000
2000
而且对社会带来好处
05:34
just like a fundamental基本的 technology技术.
115
309000
2000
就好像很基本的科技一样
05:36
So why don't we have a good theory理论 of brains大脑?
116
311000
2000
那为什么我们没有一个好的大脑理论?
05:38
And people have been working加工 on it for 100 years年份.
117
313000
3000
虽然人们已经研究了大概100多年了
05:41
Well, let's first take a look at what normal正常 science科学 looks容貌 like.
118
316000
2000
我们先看看一般的科学研究是怎么进行的
05:43
This is normal正常 science科学.
119
318000
2000
这是一般的科学
05:45
Normal正常 science科学 is a nice不错 balance平衡 between之间 theory理论 and experimentalists实验者.
120
320000
4000
一般的科学是平衡于理论和实验的
05:49
And so the theorist理论家 guys say, well, I think this is what's going on,
121
324000
2000
比方说,理论家先认为是这样的,
05:51
and the experimentalist实验者 says, no, you're wrong错误.
122
326000
2000
而实验家说,不,你错了
05:53
And it goes back and forth向前, you know?
123
328000
2000
反复的验证,你们明白吗?
05:55
This works作品 in physics物理. This works作品 in geology地质学. But if this is normal正常 science科学,
124
330000
2000
物理学是这样研究的,地质学也是这样研究的,但这是一般的科学
05:57
what does neuroscience神经科学 look like? This is what neuroscience神经科学 looks容貌 like.
125
332000
3000
那神经系统科学研究又怎样进行呢?我们看看
06:00
We have this mountain of data数据, which哪一个 is anatomy解剖学, physiology生理 and behavior行为.
126
335000
5000
我们有巨多的数据,包括:解剖学的,生理学的和行为学的
06:05
You can't imagine想像 how much detail详情 we know about brains大脑.
127
340000
3000
你们很难想象我们已经有多少数据
06:08
There were 28,000 people who went to the neuroscience神经科学 conference会议 this year,
128
343000
4000
今年的神经系统科学研讨会我们有 28000 个专家参与
06:12
and every一切 one of them is doing research研究 in brains大脑.
129
347000
2000
而每一个都在研究大脑
06:14
A lot of data数据. But there's no theory理论. There's a little, wimpy懦弱 box on top最佳 there.
130
349000
4000
很多的数据,但没有理论,可能有一点点,就像最上边的那小的可怜的箱子
06:18
And theory理论 has not played发挥 a role角色 in any sort分类 of grand盛大 way in the neurosciences神经科学.
131
353000
5000
而在神经系统科学研究领域当中,理论从没有像它们在一般科学里的主导地位
06:23
And it's a real真实 shame耻辱. Now why has this come about?
132
358000
3000
这是很可惜的,为什么会这样呢?
06:26
If you ask neuroscientists神经学家, why is this the state of affair事务,
133
361000
2000
如果你问神经系统科学专家,为什么情况会这样?
06:28
they'll他们会 first of all admit承认 it. But if you ask them, they'll他们会 say,
134
363000
3000
他们会同意情况是这样,但如果你问为什么,他们会说
06:31
well, there's various各个 reasons原因 we don't have a good brain theory理论.
135
366000
3000
有很多原因导致我们没有一个好的大脑理论
06:34
Some people say, well, we don't still have enough足够 data数据,
136
369000
2000
有些专家会说,我们还没有足够的数据
06:36
we need to get more information信息, there's all these things we don't know.
137
371000
3000
我们要拿更多的数据,我们还有很多不明白的
06:39
Well, I just told you there's so much data数据 coming未来 out your ears耳朵.
138
374000
3000
嗯, 我刚刚告诉过你们了
06:42
We have so much information信息, we don't even know how to begin开始 to organize组织 it.
139
377000
3000
我们有太多的数据但不知道怎么去组织
06:45
What good is more going to do?
140
380000
2000
那就算有更多的数据又有何用?
06:47
Maybe we'll be lucky幸运 and discover发现 some magic魔法 thing, but I don't think so.
141
382000
3000
可能我们会幸运的突然发现谜底,但我不认为会发生
06:50
This is actually其实 a symptom症状 of the fact事实 that we just don't have a theory理论.
142
385000
3000
种种证据都在说明我们根本没有一个好的理论
06:53
We don't need more data数据 -- we need a good theory理论 about it.
143
388000
3000
我们不需要更多的数据,我们只需要一个好的理论
06:56
Another另一个 one is sometimes有时 people say, well, brains大脑 are so complex复杂,
144
391000
3000
另一些专家会说,大脑太复杂了
06:59
it'll它会 take another另一个 50 years年份.
145
394000
2000
这研究会再花 50 年
07:01
I even think Chris克里斯 said something like this yesterday昨天.
146
396000
2000
我想 Chris 昨天也说过类似的话
07:03
I'm not sure what you said, Chris克里斯, but something like,
147
398000
2000
我不肯定 Chris 你所说的内容,但大概是,
07:05
well, it's one of the most complicated复杂 things in the universe宇宙. That's not true真正.
148
400000
3000
(大脑研究)是宇宙中最复杂的。我不认同
07:08
You're more complicated复杂 than your brain. You've got a brain.
149
403000
2000
你们都比大脑复杂,你们都有大脑
07:10
And it's also, although虽然 the brain looks容貌 very complicated复杂,
150
405000
2000
而且,大脑只是看似复杂,
07:12
things look complicated复杂 until直到 you understand理解 them.
151
407000
3000
所以事物在弄明白前都是复杂的
07:15
That's always been the case案件. And so all we can say, well,
152
410000
3000
我们可以说,
07:18
my neocortex新皮层, which哪一个 is the part部分 of the brain I'm interested有兴趣 in, has 30 billion十亿 cells细胞.
153
413000
4000
新大脑皮层(neocortex),大脑里面我们最感兴趣的部分,有 300 亿细胞
07:22
But, you know what? It's very, very regular定期.
154
417000
2000
但你们知道吗,它(新大脑皮层)非常有规律
07:24
In fact事实, it looks容貌 like it's the same相同 thing repeated重复 over and over and over again.
155
419000
3000
实际上,它就像同样的组织不停的重覆
07:27
It's not as complex复杂 as it looks容貌. That's not the issue问题.
156
422000
3000
它不像想象中复杂,那不是问题
07:30
Some people say, brains大脑 can't understand理解 brains大脑.
157
425000
2000
有些人说,大脑不能明白大脑
07:32
Very Zen-like禅宗般. Whoo. (Laughter笑声)
158
427000
3000
很玄,喔
07:35
You know,
159
430000
1000
(笑声)
07:36
it sounds声音 good, but why? I mean, what's the point?
160
431000
3000
听起来挺好,但有什么用?
07:39
It's just a bunch of cells细胞. You understand理解 your liver.
161
434000
3000
它只是一堆细胞,就好像你了解你的肝脏
07:42
It's got a lot of cells细胞 in it too, right?
162
437000
2000
肝脏也是一堆细胞是吗
07:44
So, you know, I don't think there's anything to that.
163
439000
2000
所以,我不见得大脑有什么分别的
07:46
And finally最后, some people say, well, you know,
164
441000
2000
还有一些人说
07:48
I don't feel like a bunch of cells细胞, you know. I'm conscious意识.
165
443000
4000
“我不认为自己只是一堆细胞,我是神志清醒的
07:52
I've got this experience经验, I'm in the world世界, you know.
166
447000
2000
我又很多经历,我处在一世界,明白不,
07:54
I can't be just a bunch of cells细胞. Well, you know,
167
449000
2000
我不可能只是一堆细胞”
07:56
people used to believe there was a life force to be living活的,
168
451000
3000
人们曾经相信有‘生命力’
07:59
and we now know that's really not true真正 at all.
169
454000
2000
我们现在已经知道那根本不正确
08:01
And there's really no evidence证据 that says -- well, other than people
170
456000
3000
而且根本就没有证据证明,除了人类之外
08:04
just have disbelief怀疑 that cells细胞 can do what they do.
171
459000
2000
只是不相信一堆细胞能做人能做的事
08:06
And so, if some people have fallen堕落 into the pit of metaphysical抽象的 dualism二元论,
172
461000
3000
有些人沉迷于形而上学唯物论
08:09
some really smart聪明 people, too, but we can reject拒绝 all that.
173
464000
3000
包括一些很聪明的人,但我们可以全否定
08:12
(Laughter笑声)
174
467000
2000
(笑声)
08:14
No, I'm going to tell you there's something else其他,
175
469000
3000
不,我会告诉你们另外的
08:17
and it's really fundamental基本的, and this is what it is:
176
472000
2000
很基础很根本的
08:19
there's another另一个 reason原因 why we don't have a good brain theory理论,
177
474000
2000
原因导致我们无法拥有一个好的大脑理论
08:21
and it's because we have an intuitive直观的, strongly-held强举行,
178
476000
3000
因为我们有很根深蒂固
08:24
but incorrect不正确 assumption假设 that has prevented防止 us from seeing眼看 the answer回答.
179
479000
5000
但错误的假设,这阻止了我们去寻找答案
08:29
There's something we believe that just, it's obvious明显, but it's wrong错误.
180
484000
3000
我们相信这个明显的假设,但它是错的
08:32
Now, there's a history历史 of this in science科学 and before I tell you what it is,
181
487000
4000
这在科学研究中是有先例的,但在说那之前,
08:36
I'm going to tell you a bit about the history历史 of it in science科学.
182
491000
2000
我先告诉你一些科学的历史
08:38
You look at some other scientific科学 revolutions革命,
183
493000
2000
看看其它的科学革命
08:40
and this case案件, I'm talking about the solar太阳能 system系统, that's Copernicus哥白尼,
184
495000
2000
比方说哥白尼的天体运行学说
08:42
Darwin's达尔文 evolution演化, and tectonic构造 plates, that's Wegener韦格纳.
185
497000
3000
达尔文的进化论,和魏格纳的大陆漂移学说
08:45
They all have a lot in common共同 with brain science科学.
186
500000
3000
它们跟大脑理论有很多共同点
08:48
First of all, they had a lot of unexplained原因不明 data数据. A lot of it.
187
503000
3000
第一,很多无法解析的数据
08:51
But it got more manageable管理 once一旦 they had a theory理论.
188
506000
3000
但有理论后就变的容易处理了
08:54
The best最好 minds头脑 were stumped难倒 -- really, really smart聪明 people.
189
509000
3000
那时候众多很聪明的学者都被困惑
08:57
We're not smarter聪明 now than they were then.
190
512000
2000
我们并不比他们聪明,
08:59
It just turns out it's really hard to think of things,
191
514000
2000
只是想出理论是很困难的,
09:01
but once一旦 you've thought of them, it's kind of easy简单 to understand理解 it.
192
516000
2000
但一想到了,就很容易明白
09:03
My daughters女儿 understood了解 these three theories理论
193
518000
2000
我的女儿都明白那三个理论
09:05
in their basic基本 framework骨架 by the time they were in kindergarten幼儿园.
194
520000
3000
的大概,在幼儿园的时候就明白
09:08
And now it's not that hard, you know, here's这里的 the apple苹果, here's这里的 the orange橙子,
195
523000
3000
所以并不是那么困难,像这有一苹果,这一橘子,
09:11
you know, the Earth地球 goes around, that kind of stuff东东.
196
526000
3000
地球围着走,等等
09:14
Finally最后, another另一个 thing is the answer回答 was there all along沿,
197
529000
2000
还有,答案早就存在
09:16
but we kind of ignored忽视 it because of this obvious明显 thing, and that's the thing.
198
531000
3000
我们只是忽视了而已
09:19
It was an intuitive直观的, strong-held强举行 belief信仰 that was wrong错误.
199
534000
3000
第二,有很根深蒂固但错的想法
09:22
In the case案件 of the solar太阳能 system系统, the idea理念 that the Earth地球 is spinning纺织
200
537000
3000
天体运行学的比方,地球在自转
09:25
and the surface表面 of the Earth地球 is going like a thousand miles英里 an hour小时,
201
540000
3000
地球表面在以千多英里在移动,
09:28
and the Earth地球 is going through通过 the solar太阳能 system系统 about a million百万 miles英里 an hour小时.
202
543000
3000
同时地球在太阳系里的轨道以百万多英里运行
09:31
This is lunacy疯狂行为. We all know the Earth地球 isn't moving移动.
203
546000
2000
疯了吧,我们都知道地球不在动
09:33
Do you feel like you're moving移动 a thousand miles英里 an hour小时?
204
548000
2000
你感觉到我们在以千多英里移动吗?
09:35
Of course课程 not. You know, and someone有人 who said,
205
550000
2000
肯定没有,还有人说
09:37
well, it was spinning纺织 around in space空间 and it's so huge巨大,
206
552000
2000
它(地球)在太空里自转而它很大
09:39
they would lock you up, and that's what they did back then.
207
554000
2000
会把你锁上,他们当时是这样想的
09:41
(Laughter笑声)
208
556000
1000
(笑声)
09:42
So it was intuitive直观的 and obvious明显. Now what about evolution演化?
209
557000
3000
这是显而易见的,我们再看看进化论
09:45
Evolution's进化的 the same相同 thing. We taught our kids孩子, well, the Bible圣经 says,
210
560000
3000
我们教孩子圣经里面说
09:48
you know, God created创建 all these species种类, cats are cats, dogs小狗 are dogs小狗,
211
563000
2000
上帝创造万物,猫是猫,狗是狗
09:50
people are people, plants植物 are plants植物, they don't change更改.
212
565000
3000
人是人,植物是植物,他们都不会变的
09:53
Noah诺亚 put them on the Ark方舟 in that order订购, blah胡说, blah胡说, blah胡说. And, you know,
213
568000
4000
诺亚(Noah) 把他们都放进方舟,等等
09:57
the fact事实 is, if you believe in evolution演化, we all have a common共同 ancestor祖先,
214
572000
4000
事实上,如果你相信进化论,我们都有共同的祖先,
10:01
and we all have a common共同 ancestry祖先 with the plant in the lobby前厅.
215
576000
3000
我们跟大厅里的植物也有共同的祖先
10:04
This is what evolution演化 tells告诉 us. And, it's true真正. It's kind of unbelievable难以置信的.
216
579000
3000
进化论是这样说的,而这是这真的,虽然有点难以置信,
10:07
And the same相同 thing about tectonic构造 plates, you know?
217
582000
3000
大陆漂移学说也一样,
10:10
All the mountains and the continents大陆 are kind of floating漂浮的 around
218
585000
2000
所有高山和大洲都在浮动
10:12
on top最佳 of the Earth地球, you know? It's like, it doesn't make any sense.
219
587000
4000
于地球上,听起来好像不合情理
10:16
So what is the intuitive直观的, but incorrect不正确 assumption假设,
220
591000
4000
那什么是直觉但错的假设
10:20
that's kept不停 us from understanding理解 brains大脑?
221
595000
2000
阻止我们理解大脑呢?
10:22
Now I'm going to tell it to you, and it's going to seem似乎 obvious明显 that that is correct正确,
222
597000
2000
我现在就告诉你们,而且很明显是正确的,
10:24
and that's the point, right? Then I'm going to have to make an argument论据
223
599000
2000
那才是重点对吗?然后我会提出论据,
10:26
why you're incorrect不正确 about the other assumption假设.
224
601000
2000
为什么其它的假设是错误的
10:28
The intuitive直观的 but obvious明显 thing is that somehow不知何故 intelligence情报
225
603000
3000
直觉告诉我们智慧
10:31
is defined定义 by behavior行为,
226
606000
2000
界定于行为
10:33
that we are intelligent智能 because of the way that we do things
227
608000
2000
我们聪明因为我们做事的方法
10:35
and the way we behave表现 intelligently智能, and I'm going to tell you that's wrong错误.
228
610000
3000
和我们行为上表现聪明,我会告诉你们这想法是错的
10:38
What it is is intelligence情报 is defined定义 by prediction预测.
229
613000
2000
智慧应该界定于推测能力
10:40
And I'm going to work you through通过 this in a few少数 slides幻灯片 here,
230
615000
3000
我会用这几张笔记
10:43
give you an example of what this means手段. Here's这里的 a system系统.
231
618000
4000
给你们看看一例子, 这是一系统,
10:47
Engineers工程师 like to look at systems系统 like this. Scientists科学家们 like to look at systems系统 like this.
232
622000
3000
工程师喜欢看系统,科学家也喜欢,
10:50
They say, well, we have a thing in a box, and we have its inputs输入 and its outputs输出.
233
625000
3000
我们有一箱子,我们有输入和输出
10:53
The AIAI people said, well, the thing in the box is a programmable可编程的 computer电脑
234
628000
3000
人工智能专家会说,那箱子里面是可编程计算机
10:56
because that's equivalent当量 to a brain, and we'll feed饲料 it some inputs输入
235
631000
2000
因为它等同大脑,而我们输入数据,
10:58
and we'll get it to do something, have some behavior行为.
236
633000
2000
我们会得到输出的行为
11:00
And Alan艾伦 Turing图灵 defined定义 the Turing图灵 test测试, which哪一个 is essentially实质上 saying,
237
635000
3000
艾伦.图灵(Alan Turing)的图灵测试说,
11:03
we'll know if something's什么是 intelligent智能 if it behaves的行为 identical相同 to a human人的.
238
638000
3000
如果行为跟人类接近就是有智慧的
11:06
A behavioral行为的 metric of what intelligence情报 is,
239
641000
3000
这是测度智慧的行为指标,
11:09
and this has stuck卡住 in our minds头脑 for a long period of time.
240
644000
3000
而我们被这想法困住了很长时间
11:12
Reality现实 though虽然, I call it real真实 intelligence情报.
241
647000
2000
实际上,我称这为真正智慧,
11:14
Real真实 intelligence情报 is built内置 on something else其他.
242
649000
2000
真正智慧是建筑于其它层面上的
11:16
We experience经验 the world世界 through通过 a sequence序列 of patterns模式, and we store商店 them,
243
651000
4000
我们通过一系列的模式来感受世界环境,然后贮存,
11:20
and we recall召回 them. And when we recall召回 them, we match比赛 them up
244
655000
3000
再回想,当我们回想时,我们会比较和对应
11:23
against反对 reality现实, and we're making制造 predictions预测 all the time.
245
658000
4000
实际情况,就这样我们不断的推测
11:27
It's an eternal永恒 metric. There's an eternal永恒 metric about us sort分类 of saying,
246
662000
3000
这是永恒的指标,一个测度我们对世界环境了解的指标和
11:30
do we understand理解 the world世界? Am I making制造 predictions预测? And so on.
247
665000
3000
我是否在推测环境,等等
11:33
You're all being存在 intelligent智能 right now, but you're not doing anything.
248
668000
2000
你们都在表现有智慧会的行为中,虽然你们什么对没有做
11:35
Maybe you're scratching搔抓 yourself你自己, or picking选择 your nose鼻子,
249
670000
2000
可能你在搔痒,可能在挖鼻子 ,
11:37
I don't know, but you're not doing anything right now,
250
672000
2000
但没有在做什么特别的,
11:39
but you're being存在 intelligent智能; you're understanding理解 what I'm saying.
251
674000
3000
但你们还是有理性有智慧的,你们明白我在说什么,
11:42
Because you're intelligent智能 and you speak说话 English英语,
252
677000
2000
因为你们都有智慧,而你们都会英语,
11:44
you know what word is at the end结束 of this -- (Silence安静)
253
679000
1000
你们都知道我说这句 --
11:45
sentence句子.
254
680000
2000
11:47
The word came来了 into you, and you're making制造 these predictions预测 all the time.
255
682000
3000
你们都猜到那字,因为你们不断的推测,
11:50
And then, what I'm saying is,
256
685000
2000
而我想说,
11:52
is that the eternal永恒 prediction预测 is the output产量 in the neocortex新皮层.
257
687000
2000
新大脑皮层的输出就是不断的推测,
11:54
And that somehow不知何故, prediction预测 leads引线 to intelligent智能 behavior行为.
258
689000
3000
推测导致有理性有智慧的行为
11:57
And here's这里的 how that happens发生. Let's start开始 with a non-intelligent非智能 brain.
259
692000
3000
而过程是这样的,我们从大脑里没有智慧的部分开始,
12:00
Well I'll argue争论 a non-intelligent非智能 brain, we got hold保持 of an old brain,
260
695000
4000
我认为我们脑里面有部分是没有智慧的,是古老的,
12:04
and we're going to say it's like a non-mammal非哺乳动物, like a reptile爬虫,
261
699000
3000
它甚至不属于哺乳类的,是属于爬行类年代的,
12:07
so I'll say, an alligator鳄鱼; we have an alligator鳄鱼.
262
702000
2000
比方说,鳄鱼,
12:09
And the alligator鳄鱼 has some very sophisticated复杂的 senses感官.
263
704000
3000
鳄鱼有很复杂强大的感官系统,
12:12
It's got good eyes眼睛 and ears耳朵 and touch触摸 senses感官 and so on,
264
707000
3000
有很好的眼睛,耳朵,触觉,等等
12:15
a mouth and a nose鼻子. It has very complex复杂 behavior行为.
265
710000
4000
还有口和鼻, 也有很复杂的行为,
12:19
It can run and hide隐藏. It has fears恐惧 and emotions情绪. It can eat you, you know.
266
714000
4000
会走会躲,会害怕会有情绪,会吃人,
12:23
It can attack攻击. It can do all kinds of stuff东东.
267
718000
4000
会攻击, 等等
12:27
But we don't consider考虑 the alligator鳄鱼 very intelligent智能, not like in a human人的 sort分类 of way.
268
722000
5000
但我们不会视鳄鱼为很有智慧,不像人类的智慧,
12:32
But it has all this complex复杂 behavior行为 already已经.
269
727000
2000
虽然它已拥有很复杂的行为,
12:34
Now, in evolution演化, what happened发生?
270
729000
2000
进化论里怎么说的?
12:36
First thing that happened发生 in evolution演化 with mammals哺乳动物,
271
731000
3000
哺乳类的进化,
12:39
we started开始 to develop发展 a thing called the neocortex新皮层.
272
734000
2000
从开发新大脑皮层开始,
12:41
And I'm going to represent代表 the neocortex新皮层 here,
273
736000
2000
我们用这个来代表新大脑皮层,
12:43
by this box that's sticking症结 on top最佳 of the old brain.
274
738000
2000
这个在(老)小脑上面的箱子,
12:45
Neocortex新皮层 means手段 new layer. It is a new layer on top最佳 of your brain.
275
740000
3000
新大脑皮层的解释是大脑上面的新一层,
12:48
If you don't know it, it's the wrinkly thing on the top最佳 of your head that,
276
743000
3000
它看上去是皱褶着的
12:51
it's got wrinkly because it got shoved in there and doesn't fit适合.
277
746000
3000
因为它被挤进去而没有空间了,
12:54
(Laughter笑声)
278
749000
1000
(笑声)
12:55
No, really, that's what it is. It's about the size尺寸 of a table napkin餐巾.
279
750000
2000
是真的!它大概是一张台布的大小
12:57
And it doesn't fit适合, so it gets得到 all wrinkly. Now look at how I've drawn this here.
280
752000
3000
而放不下,所以就皱褶起来了,现在我们看看我画 的这个,
13:00
The old brain is still there. You still have that alligator鳄鱼 brain.
281
755000
4000
(旧)小脑还在这里,那鳄鱼的脑袋还在,
13:04
You do. It's your emotional情绪化 brain.
282
759000
2000
你们都有,是你脑里情绪和感官的部分
13:06
It's all those things, and all those gut肠道 reactions反应 you have.
283
761000
3000
它负责所有直觉,本能反应,
13:09
And on top最佳 of it, we have this memory记忆 system系统 called the neocortex新皮层.
284
764000
3000
在它上面,是我们说的新大脑皮层,
13:12
And the memory记忆 system系统 is sitting坐在 over the sensory感觉的 part部分 of the brain.
285
767000
4000
它是包围着脑里感官系统的记忆系统,
13:16
And so as the sensory感觉的 input输入 comes in and feeds供稿 from the old brain,
286
771000
3000
感官输入先进小脑,
13:19
it also goes up into the neocortex新皮层. And the neocortex新皮层 is just memorizing记忆.
287
774000
4000
再走上新大脑皮层,而新大脑皮层只是记忆着,
13:23
It's sitting坐在 there saying, ah, I'm going to memorize记忆 all the things that are going on:
288
778000
4000
它记着所以发生的事情,
13:27
where I've been, people I've seen看到, things I've heard听说, and so on.
289
782000
2000
像去了哪里,见过的人,听过的事,等等,
13:29
And in the future未来, when it sees看到 something similar类似 to that again,
290
784000
4000
在以后见到类似的情况,
13:33
so in a similar类似 environment环境, or the exact精确 same相同 environment环境,
291
788000
3000
类似的环境,或一样的环境,
13:36
it'll它会 play it back. It'll它会 start开始 playing播放 it back.
292
791000
2000
它会把记忆‘重播’,
13:38
Oh, I've been here before. And when you've been here before,
293
793000
2000
就会发现以前来过这地方,而如果你曾经来过这里,
13:40
this happened发生 next下一个. It allows允许 you to predict预测 the future未来.
294
795000
3000
你记得什么会发生,让你可以猜测将来
13:43
It allows允许 you to, literally按照字面 it feeds供稿 back the signals信号 into your brain;
295
798000
4000
就好象,外界的信号传入大脑,
13:47
they'll他们会 let you see what's going to happen发生 next下一个,
296
802000
2000
让你看到什么将会发生,
13:49
will let you hear the word "sentence句子" before I said it.
297
804000
3000
就像刚才你们会知道我准备会说的词
13:52
And it's this feeding馈送 back into the old brain
298
807000
3000
正是这个信号的传递回小脑
13:55
that'll那会 allow允许 you to make very more intelligent智能 decisions决定.
299
810000
3000
让你们去作出很理性的决定
13:58
This is the most important重要 slide滑动 of my talk, so I'll dwell on it a little bit.
300
813000
3000
这是我演说里最重要的一点,
14:01
And so, all the time you say, oh, I can predict预测 the things.
301
816000
4000
所以,你们不断的在猜测食物,
14:05
And if you're a rat and you go through通过 a maze迷宫, and then you learn学习 the maze迷宫,
302
820000
3000
如果我们像白老鼠一样在走迷宫,那就学习那个迷宫,
14:08
the next下一个 time you're in a maze迷宫, you have the same相同 behavior行为,
303
823000
2000
下次再走,行为一样,
14:10
but all of a sudden突然, you're smarter聪明
304
825000
2000
但会变聪明了,
14:12
because you say, oh, I recognize认识 this maze迷宫, I know which哪一个 way to go,
305
827000
3000
因为会认得那迷宫,知道怎么走,
14:15
I've been here before, I can envision预见 the future未来. And that's what it's doing.
306
830000
3000
曾经走过,可以预想,
14:18
In humans人类 -- by the way, this is true真正 for all mammals哺乳动物;
307
833000
3000
人类和其他哺乳类动物都会这样,
14:21
it's true真正 for other mammals哺乳动物 -- and in humans人类, it got a lot worse更差.
308
836000
2000
人类的情况会更极端,
14:23
In humans人类, we actually其实 developed发达 the front面前 part部分 of the neocortex新皮层
309
838000
3000
我们会发展新大脑皮层的前端,
14:26
called the anterior前面的 part部分 of the neocortex新皮层. And nature性质 did a little trick.
310
841000
4000
然后大自然会弄一小把戏
14:30
It copied复制 the posterior part部分, the back part部分, which哪一个 is sensory感觉的,
311
845000
2000
将新大脑皮层后端,感官的部分,拷贝
14:32
and put it in the front面前 part部分.
312
847000
2000
到前端
14:34
And humans人类 uniquely独特地 have the same相同 mechanism机制 on the front面前,
313
849000
2000
人类大脑前端有独特的构造,跟后端一样
14:36
but we use it for motor发动机 control控制.
314
851000
2000
但我们用来控制运动
14:38
So we are now able能够 to make very sophisticated复杂的 motor发动机 planning规划, things like that.
315
853000
3000
所以我们可以进行很复杂的计划运动,
14:41
I don't have time to get into all this, but if you want to understand理解 how a brain works作品,
316
856000
3000
这个我们先不说,要理解大脑怎么运作,
14:44
you have to understand理解 how the first part部分 of the mammalian哺乳动物 neocortex新皮层 works作品,
317
859000
3000
我们先了解第一代哺乳类动物新大脑皮层的运作,
14:47
how it is we store商店 patterns模式 and make predictions预测.
318
862000
2000
和怎么去贮存资料样式和作出猜测
14:49
So let me give you a few少数 examples例子 of predictions预测.
319
864000
3000
我先列几个猜测的例子
14:52
I already已经 said the word "sentence句子." In music音乐,
320
867000
2000
我已说过句子了,在音乐中,
14:54
if you've heard听说 a song歌曲 before, if you heard听说 Jill吉尔 sing those songs歌曲 before,
321
869000
3000
如果你听过一首歌,如果你听过吉尔(Jill)唱歌,
14:57
when she sings them, the next下一个 note注意 pops持久性有机污染物 into your head already已经 --
322
872000
3000
当她唱的时候,下一个音符就会在你脑海中了,
15:00
you anticipate预料 it as you're going. If it was an album专辑 of music音乐,
323
875000
2000
你会有预感,如果是一张音乐专辑,
15:02
the end结束 of one album专辑, the next下一个 song歌曲 pops持久性有机污染物 into your head.
324
877000
3000
听完一首歌,下一首已在你脑海出现,
15:05
And these things happen发生 all the time. You're making制造 these predictions预测.
325
880000
2000
这情况经常发生,你在不断的猜测,
15:07
I have this thing called the altered改变 door thought experiment实验.
326
882000
3000
我有一个用门的实验,
15:10
And the altered改变 door thought experiment实验 says, you have a door at home,
327
885000
3000
是这样的,你家有一门,
15:13
and when you're here, I'm changing改变 it, I've got a guy
328
888000
3000
当你在这里的时候,我去把它改了,我们有一个人,
15:16
back at your house right now, moving移动 the door around,
329
891000
2000
现在在你家,把门改过来,
15:18
and they're going to take your doorknob门把手 and move移动 it over two inches英寸.
330
893000
2000
他会把你的门把手移 2 寸,
15:20
And when you go home tonight今晚, you're going to put your hand out there,
331
895000
2000
当你今晚回家,找把手开门,
15:22
and you're going to reach达到 for the doorknob门把手 and you're going to notice注意
332
897000
2000
你会发现把手
15:24
it's in the wrong错误 spot, and you'll你会 go, whoa, something happened发生.
333
899000
3000
在错的位置,你会感觉,有点问题,
15:27
It may可能 take a second第二 to figure数字 out what it was, but something happened发生.
334
902000
2000
可能等一秒才发现问题,但感觉到不对劲,
15:29
Now I could change更改 your doorknob门把手 in other ways方法.
335
904000
2000
我也可以用别的方法改变门把手,
15:31
I can make it larger or smaller, I can change更改 its brass黄铜 to silver,
336
906000
2000
弄大一点或小一点,从铜改为银的,
15:33
I could make it a lever杠杆. I can change更改 your door, put colors颜色 on;
337
908000
2000
也可以把门改了,改颜色,
15:35
I can put windows视窗 in. I can change更改 a thousand things about your door,
338
910000
3000
加玻璃,有很多方法去改,
15:38
and in the two seconds you take to open打开 your door,
339
913000
2000
而就在你开门的两秒钟,
15:40
you're going to notice注意 that something has changed.
340
915000
3000
你会发现不对劲,
15:43
Now, the engineering工程 approach途径 to this, the AIAI approach途径 to this,
341
918000
2000
那传统的工程或人工智能对这问题的方法是,
15:45
is to build建立 a door database数据库. It has all the door attributes属性.
342
920000
3000
起一个门的数据库,有所以关于门的参数,
15:48
And as you go up to the door, you know, let's check them off one at time.
343
923000
3000
当你到了门前,便进数据库一个一个比较,
15:51
Door, door, door, you know, color颜色, you know what I'm saying.
344
926000
2000
所以样式的门,不同颜色的,
15:53
We don't do that. Your brain doesn't do that.
345
928000
2000
我们人类肯定不会这样做的,你们的大脑不会这样运作,
15:55
What your brain is doing is making制造 constant不变 predictions预测 all the time
346
930000
2000
你的大脑会不停的作出猜测,
15:57
about what is going to happen发生 in your environment环境.
347
932000
2000
对你附近环境有可能会发生的作出猜测,
15:59
As I put my hand on this table, I expect期望 to feel it stop.
348
934000
3000
当我把手放着桌上,我预料手会停在上面,
16:02
When I walk步行, every一切 step, if I missed错过 it by an eighth第八 of an inch英寸,
349
937000
3000
当我走路的时候,每一步,如果只是差了八分之一寸,
16:05
I'll know something has changed.
350
940000
2000
我会知道有情况改变,
16:07
You're constantly经常 making制造 predictions预测 about your environment环境.
351
942000
2000
你们不停的对身边环境作出猜测,
16:09
I'll talk about vision视力 here briefly简要地. This is a picture图片 of a woman女人.
352
944000
3000
让我们看看视觉系统,这是一张女人的图片,
16:12
And when you look at people, your eyes眼睛 are caught抓住
353
947000
2000
当你看人的时候,你的眼神会停留,
16:14
over at two to three times a second第二.
354
949000
1000
大概两到三秒,
16:15
You're not aware知道的 of this, but your eyes眼睛 are always moving移动.
355
950000
2000
你应该意识不到,但你的眼球不停在动,
16:17
And so when you look at someone's谁家 face面对,
356
952000
2000
所以当你看一个人的脸,
16:19
you'd typically一般 go from eye to eye to eye to nose鼻子 to mouth.
357
954000
2000
你通常会从看着眼到鼻到口,
16:21
Now, when your eye moves移动 from eye to eye,
358
956000
2000
如果你在看眼的位置的时候,
16:23
if there was something else其他 there like, a nose鼻子,
359
958000
2000
出现像鼻子的东西,
16:25
you'd see a nose鼻子 where an eye is supposed应该 to be,
360
960000
2000
你看见鼻子长在眼睛的地方,
16:27
and you'd go, oh shit拉屎, you know --
361
962000
3000
你会吓一跳
16:30
(Laughter笑声)
362
965000
1000
(笑声)
16:31
There's something wrong错误 about this person.
363
966000
2000
这个人有点问题,
16:33
And that's because you're making制造 a prediction预测.
364
968000
2000
这都以为你在推测,
16:35
It's not like you just look over there and say, what am I seeing眼看 now?
365
970000
2000
不会是因为你在看东西而在想着到底是什么,
16:37
A nose鼻子, that's okay. No, you have an expectation期望 of what you're going to see.
366
972000
3000
你不会预料看到一鼻子在眼睛的位置,
16:40
(Laughter笑声)
367
975000
1000
(笑声)
16:41
Every一切 single moment时刻. And finally最后, let's think about how we test测试 intelligence情报.
368
976000
4000
现在我们看看我们怎样测试智慧的,
16:45
We test测试 it by prediction预测. What is the next下一个 word in this, you know?
369
980000
3000
我们用猜测能力来测试的,下一个词是什么?
16:48
This is to this as this is to this. What is the next下一个 number in this sentence句子?
370
983000
3000
这个配这个,那个配那个,下一个数是什么?
16:51
Here's这里的 three visions愿景 of an object目的.
371
986000
2000
这是这东西的三个看法,
16:53
What's the fourth第四 one? That's how we test测试 it. It's all about prediction预测.
372
988000
4000
第四个是什么?我们就是这样测试猜测能力
16:57
So what is the recipe食谱 for brain theory理论?
373
992000
3000
那什么是大脑理论的秘诀?
17:00
First of all, we have to have the right framework骨架.
374
995000
3000
第一,我们需要合适的框架,
17:03
And the framework骨架 is a memory记忆 framework骨架,
375
998000
2000
一个记忆的框架,
17:05
not a computation计算 or behavior行为 framework骨架. It's a memory记忆 framework骨架.
376
1000000
2000
不是计算的或行为的框架,是一个记忆的框架,
17:07
How do you store商店 and recall召回 these sequences序列 or patterns模式? It's spatio-temporal时空 patterns模式.
377
1002000
4000
你怎么贮存和回忆有关联的样式组合?这是时间空间样式,
17:11
Then, if in that framework骨架, you take a bunch of theoreticians理论家.
378
1006000
3000
然后,如果在框架里,我们找一群理论研究者,
17:14
Now biologists生物学家 generally通常 are not good theoreticians理论家.
379
1009000
2000
生物学者一般不是好的理论学者,
17:16
It's not always true真正, but in general一般, there's not a good history历史 of theory理论 in biology生物学.
380
1011000
4000
不一定,但历史里没有好的生物理论,
17:20
So I found发现 the best最好 people to work with are physicists物理学家,
381
1015000
3000
我觉得物理学者,
17:23
engineers工程师 and mathematicians数学家, who tend趋向 to think algorithmically算法.
382
1018000
3000
工程师和数学家都适合,他们想法都很规则很系统的,
17:26
Then they have to learn学习 the anatomy解剖学, and they've他们已经 got to learn学习 the physiology生理.
383
1021000
3000
然后他们要学解剖学和生理学,
17:29
You have to make these theories理论 very realistic实际 in anatomical解剖 terms条款.
384
1024000
4000
我们需要让这理论非常的实在,从解剖学角度来看,
17:33
Anyone任何人 who gets得到 up and tells告诉 you their theory理论 about how the brain works作品
385
1028000
4000
如果有人解释大脑理论时,
17:37
and doesn't tell you exactly究竟 how it's working加工 in the brain
386
1032000
2000
而不告诉你大脑里面怎么运作,
17:39
and how the wiring接线 works作品 in the brain, it is not a theory理论.
387
1034000
2000
和大脑各部分的联系,那就不是真正的理论了
17:41
And that's what we're doing at the Redwood红木 Neuroscience神经科学 Institute研究所.
388
1036000
3000
而我们的研究院正是研究这方面的,
17:44
I would love to have more time to tell you we're making制造 fantastic奇妙 progress进展 in this thing,
389
1039000
4000
我很希望有更多的时间告诉你们最近的研究成果
17:48
and I expect期望 to be back up on this stage阶段,
390
1043000
2000
我以后会再回来
17:50
maybe this will be some other time in the not too distant遥远 future未来 and tell you about it.
391
1045000
2000
在不久的将来,来告诉大家
17:52
I'm really, really excited兴奋. This is not going to take 50 years年份 at all.
392
1047000
3000
我真的很兴奋,这肯定不会花 50 年,
17:55
So what will brain theory理论 look like?
393
1050000
2000
那大脑理论像什么呢?
17:57
First of all, it's going to be a theory理论 about memory记忆.
394
1052000
2000
首先,它会是一个关于记忆的理论,
17:59
Not like computer电脑 memory记忆. It's not at all like computer电脑 memory记忆.
395
1054000
3000
不是计算机的记忆
18:02
It's very, very different不同. And it's a memory记忆 of these very
396
1057000
2000
很不一样
18:04
high-dimensional高维 patterns模式, like the things that come from your eyes眼睛.
397
1059000
3000
是多维样式的记忆,就像从你们眼睛输出的,
18:07
It's also memory记忆 of sequences序列.
398
1062000
2000
它也会是很多组有关联记忆,
18:09
You cannot不能 learn学习 or recall召回 anything outside of a sequence序列.
399
1064000
2000
你不会学习或回忆没有关联的东西,
18:11
A song歌曲 must必须 be heard听说 in sequence序列 over time,
400
1066000
3000
就像一首歌在时间上是有先后的记忆
18:14
and you must必须 play it back in sequence序列 over time.
401
1069000
3000
要回忆起来也是一连串的回忆,
18:17
And these sequences序列 are auto-associatively自动关联方式 recalled回顾, so if I see something,
402
1072000
3000
这些关联记忆组群会在回忆时会自动联系连结,所以当我们看到,
18:20
I hear something, it reminds提醒 me of it, and then it plays播放 back automatically自动.
403
1075000
3000
听到一些类似的东西,就会把记忆重播,
18:23
It's an automatic自动 playback回放. And prediction预测 of future未来 inputs输入 is the desired期望 output产量.
404
1078000
4000
是自动的重播,最后输出是未来的猜测,
18:27
And as I said, the theory理论 must必须 be biologically生物 accurate准确,
405
1082000
3000
我们提过,这理论在生物学上合理,
18:30
it must必须 be testable可测试, and you must必须 be able能够 to build建立 it.
406
1085000
2000
能测试的,可推理出来的
18:32
If you don't build建立 it, you don't understand理解 it. So, one more slide滑动 here.
407
1087000
4000
如果你不推理出来,你不会明白,还有一张笔记,
18:36
What is this going to result结果 in? Are we going to really build建立 intelligent智能 machines?
408
1091000
4000
这研究结果有什么作用呢?我们真的会制造有智慧的机器?
18:40
Absolutely绝对. And it's going to be different不同 than people think.
409
1095000
4000
这是肯定的,而会跟我们想像的不一样,
18:44
No doubt怀疑 that it's going to happen发生, in my mind心神.
410
1099000
3000
我绝不怀疑,
18:47
First of all, it's going to be built内置 up, we're going to build建立 the stuff东东 out of silicon.
411
1102000
4000
首先,我们会用硅来制造,
18:51
The same相同 techniques技术 we use for building建造 silicon computer电脑 memories回忆,
412
1106000
3000
制造计算机内存的方法,
18:54
we can use for here.
413
1109000
1000
我们可以用上,
18:55
But they're very different不同 types类型 of memories回忆.
414
1110000
2000
但是将会是很不一样的记忆体,
18:57
And we're going to attach连接 these memories回忆 to sensors传感器,
415
1112000
2000
我们会把感应器和这些记忆体连接上,
18:59
and the sensors传感器 will experience经验 real-live实时直播, real-world真实世界 data数据,
416
1114000
3000
感应器会接受真实环境的数据,
19:02
and these things are going to learn学习 about their environment环境.
417
1117000
2000
而这些机器会学习它们的环境,
19:04
Now it's very unlikely不会 the first things you're going to see are like robots机器人.
418
1119000
3000
一开始发展出来就像机器人的可能比较低,
19:07
Not that robots机器人 aren't useful有用 and people can build建立 robots机器人.
419
1122000
3000
不是说机器人没有用处或我们制造不出来,
19:10
But the robotics机器人 part部分 is the hardest最难 part部分. That's the old brain. That's really hard.
420
1125000
4000
但是机器人硬件是最难制造的,那像旧(小)脑,
19:14
The new brain is actually其实 kind of easier更轻松 than the old brain.
421
1129000
2000
(新)大脑比小脑容易,
19:16
So the first thing we're going to do are the things that don't require要求 a lot of robotics机器人.
422
1131000
3000
所以刚开始我们会造一些不需要太多机器人硬件的,
19:19
So you're not going to see C-C-3POPO.
423
1134000
2000
应该不会见到 C-3PO
19:21
You're going to more see things like, you know, intelligent智能 cars汽车
424
1136000
2000
你会见到比较多类似,智能车,
19:23
that really understand理解 what traffic交通 is and what driving主动 is
425
1138000
3000
会理解交通情况和驾驶,
19:26
and have learned学到了 that certain某些 types类型 of cars汽车 with the blinkers护目镜 on for half a minute分钟
426
1141000
3000
和懂得比方说,有些车的转向显示灯亮了半分钟
19:29
probably大概 aren't going to turn, things like that.
427
1144000
2000
应该不是真的想转向,
19:31
(Laughter笑声)
428
1146000
1000
(笑声)
19:32
We can also do intelligent智能 security安全 systems系统.
429
1147000
2000
我们也可以制造智能保安系统
19:34
Anywhere随地 where we're basically基本上 using运用 our brain, but not doing a lot of mechanics机械学.
430
1149000
4000
任何需要很多大脑分析但不需要很多的机械的领域,
19:38
Those are the things that are going to happen发生 first.
431
1153000
2000
都会是在初期有发展的。
19:40
But ultimately最终, the world's世界 the limit限制 here.
432
1155000
2000
但最终,会发展到各方面,
19:42
I don't know how this is going to turn out.
433
1157000
2000
我也不知道会发展成怎样,
19:44
I know a lot of people who invented发明 the microprocessor微处理器
434
1159000
2000
我认识很多发明微处理器的专家,
19:46
and if you talk to them, they knew知道 what they were doing was really significant重大,
435
1161000
5000
你如果问他们,那时候他们知道正在做很有意义的事,
19:51
but they didn't really know what was going to happen发生.
436
1166000
3000
但也不知道会发展成什么,
19:54
They couldn't不能 anticipate预料 cell细胞 phones手机 and the Internet互联网 and all this kind of stuff东东.
437
1169000
5000
他们也没有预计手机,互联网等等的发展,
19:59
They just knew知道 like, hey, they were going to build建立 calculators计算器
438
1174000
2000
他们只知道,会制造计算机,
20:01
and traffic交通 light controllers控制器. But it's going to be big.
439
1176000
2000
交通灯的控制器等等,但都感觉到是很重大的,
20:03
In the same相同 way, this is like brain science科学 and these memories回忆
440
1178000
3000
同样地,大脑的研究和记忆体,
20:06
are going to be a very fundamental基本的 technology技术, and it's going to lead
441
1181000
3000
将会成为很基本的科技,它将会在未来100年带领着
20:09
to very unbelievable难以置信的 changes变化 in the next下一个 100 years年份.
442
1184000
3000
一些很难想象的发展
20:12
And I'm most excited兴奋 about how we're going to use them in science科学.
443
1187000
4000
而令我最兴奋的是我们怎样利用这科技,
20:16
So I think that's all my time, I'm over it, and I'm going to end结束 my talk
444
1191000
3000
我想我已经超过限时了,我的演讲就在这
20:19
right there.
445
1194000
1000
结束
Translated by 陆政宜 Lawrence Lok
Reviewed by Ken Zheng

▲Back to top

ABOUT THE SPEAKER
Jeff Hawkins - Computer designer, brain researcher
Jeff Hawkins pioneered the development of PDAs such as the Palm and Treo. Now he's trying to understand how the human brain really works, and adapt its method -- which he describes as a deep system for storing memory -- to create new kinds of computers and tools.

Why you should listen

Jeff Hawkins' Palm PDA became such a widely used productivity tool during the 1990s that some fanatical users claimed it replaced their brains. But Hawkins' deepest interest was in the brain itself. So after the success of the Palm and Treo, which he brought to market at Handspring, Hawkins delved into brain research at the Redwood Center for Theoretical Neuroscience in Berkeley, Calif., and a new company called Numenta.

Hawkins' dual goal is to achieve an understanding of how the human brain actually works -- and then develop software to mimic its functionality, delivering true artificial intelligence. In his book On Intelligence (2004) he lays out his compelling, controversial theory: Contrary to popular AI wisdom, the human neocortex doesn't work like a processor; rather, it relies on a memory system that stores and plays back experiences to help us predict, intelligently, what will happen next. He thinks that "hierarchical temporal memory" computer platforms, which mimic this functionality (and which Numenta might pioneer), could enable groundbreaking new applications that could powerfully extend human intelligence.

More profile about the speaker
Jeff Hawkins | Speaker | TED.com