未登录,请登录后再发表信息
最新评论 (0)
播放视频

人工智能是如何让我们变得更好的

3 principles for creating safer AI | Stuart Russell

这是李世石
This is Lee Sedol.
李世石是全世界 最顶尖的围棋高手之一
Lee Sedol is one of the world’s greatest Go players,
在这一刻 他所经历的 足以让我硅谷的朋友们
and he’s having what my friends in Silicon Valley call
喊一句”我的天啊“——
a “Holy Cow” moment —
(笑声)
(Laughter)
在这一刻
a moment where we realize
我们意识到原来人工智能发展的进程 比我们预想的要快得多
that AI is actually progressing a lot faster than we expected.
人们在围棋棋盘上已经输了 那在现实世界中又如何呢?
So humans have lost on the Go board. What about the real world?
当然了 现实世界要 比围棋棋盘要大得多
Well, the real world is much bigger,
复杂得多
much more complicated than the Go board.
相比之下每一步也没那么明确
It’s a lot less visible,
但现实世界仍然是一个选择性问题
but it’s still a decision problem.
如果我们想想那一些在不久的未来
And if we think about some of the technologies
即将来临的新科技……
that are coming down the pike …
Noriko提到机器还不能进行阅读
Noriko [Arai] mentioned that reading is not yet happening in machines,
至少达不到理解的程度
at least with understanding.
但这迟早会发生
But that will happen,
而当它发生时
and when that happens,
不久之后
very soon afterwards,
机器就将读遍人类写下的所有东西
machines will have read everything that the human race has ever written.
这将使机器除了拥有
And that will enable machines,
比人类看得更远的能力
along with the ability to look further ahead than humans can,
就像我们在围棋中看到的那样
as we’ve already seen in Go,
如果机器能接触到比人类更多的信息
if they also have access to more information,
则将能够在现实世界中 做出比人类更好的选择
they’ll be able to make better decisions in the real world than we can.
那这是一件好事吗?
So is that a good thing?
我当然希望如此
Well, I hope so.
人类的全部文明 我们所珍视的一切
Our entire civilization, everything that we value,
都是基于我们的智慧之上
is based on our intelligence.
如果我们能掌控更强大的智能
And if we had access to a lot more intelligence,
那我们人类的 创造力 就真的没有极限了
then there’s really no limit to what the human race can do.
我认为这可能就像很多人描述的那样
And I think this could be, as some people have described it,
会成为人类历史上最重要的事件
the biggest event in human history.
那为什么有的人会说出以下的言论
So why are people saying things like this,
说人工智能将是人类的末日呢?
that AI might spell the end of the human race?
这是一个新事物吗?
Is this a new thing?
这只关乎伊隆马斯克、 比尔盖茨 和斯提芬霍金吗?
Is it just Elon Musk and Bill Gates and Stephen Hawking?
其实不是的 人工智能 这个概念已经存在很长时间了
Actually, no. This idea has been around for a while.
请看这段话:
Here’s a quotation:
“即便我们能够将机器 维持在一个屈服于我们的地位
“Even if we could keep the machines in a subservient position,
比如说 在战略性时刻将电源关闭 ”——
for instance, by turning off the power at strategic moments” —
我等会儿再来讨论 ”关闭电源“这一话题 ”
and I’ll come back to that “turning off the power” idea later on —
我们作为一个物种 仍然应该自感惭愧
“we should, as a species, feel greatly humbled.”
“这段话是谁说的呢? 是阿兰图灵 他在1951年说的
So who said this? This is Alan Turing in 1951.
阿兰图灵 众所皆知 是计算机科学之父
Alan Turing, as you know, is the father of computer science
从很多意义上说 他也是人工智能之父
and in many ways, the father of AI as well.
当我们考虑这个问题
So if we think about this problem,
创造一个比自己更智能的 物种的问题时
the problem of creating something more intelligent than your own species,
我们不妨将它称为”大猩猩问题
we might call this “the gorilla problem,”
“因为这正是大猩猩的 祖先们几百万年前所经历的
because gorillas’ ancestors did this a few million years ago,
我们今天可以去问大猩猩们:
and now we can ask the gorillas:
那么做是不是一个好主意?
Was this a good idea?
在这幅图里 大猩猩们正在 开会讨论那么做是不是一个好主意
So here they are having a meeting to discuss whether it was a good idea,
片刻后他们下定结论
and after a little while, they conclude, no,
不是的那是一个很糟糕的主意
this was a terrible idea.
我们的物种已经奄奄一息了
Our species is in dire straits.
你都可以从它们的眼神中看到这种忧伤
In fact, you can see the existential sadness in their eyes.
(笑声)
(Laughter)
所以创造比你自己更聪明的物种
So this queasy feeling that making something smarter than your own species
也许不是一个好主意——
is maybe not a good idea —
那我们能做些什么呢?
what can we do about that?
其实没什么能做的 除了停止研究人工智能
Well, really nothing, except stop doing AI,
但因为人工智能能带来 我之前所说的诸多益处
and because of all the benefits that I mentioned
也因为我是 人工智能的研究者之一
and because I’m an AI researcher,
我可不同意就这么止步实际上
I’m not having that.
我想继续做人工智能
I actually want to be able to keep doing AI.
所以我们需要把这个问题更细化一点
So we actually need to nail down the problem a bit more.
它到底是什么呢?
What exactly is the problem?
那就是为什么更强大的 人工智能可能会是灾难呢?
Why is better AI possibly a catastrophe?
再来看这段话:”
So here’s another quotation:
我们一定得确保我们 给机器输入的目的和价值
“We had better be quite sure that the purpose put into the machine
是我们确实想要的目的和价值
is the purpose which we really desire.”
“这是诺博特维纳在1960年说的
This was said by Norbert Wiener in 1960,
他说这话时是刚看到 一个早期的学习系统
shortly after he watched one of the very early learning systems
这个系统在学习如何能把 西洋棋下得比它的
learn to play checkers better than its creator.
创造者更好 与此如出一辙的一句话
But this could equally have been said
迈达斯国王也说过
by King Midas.
迈达斯国王说:”我希望 我触碰的所有东西都变成金子 “
King Midas said, “I want everything I touch to turn to gold,”
结果他真的获得了点石成金的能力
and he got exactly what he asked for.
那就是他所输入的目的
That was the purpose that he put into the machine,
从一定程度上说
so to speak,
后来他的食物、 他的家人都变成了金子
and then his food and his drink and his relatives turned to gold
他死在痛苦与饥饿之中
and he died in misery and starvation.
我们可以把这个问题 叫做”迈达斯问题“
So we’ll call this “the King Midas problem”
这个问题是我们阐述的目标 但实际上
of stating an objective which is not, in fact,
与我们真正想要的不一致
truly aligned with what we want.
用现代的术语来说 我们把它称为”价值一致性问题
In modern terms, we call this “the value alignment problem.”
“而输入错误的目标 仅仅是问题的一部分
Putting in the wrong objective is not the only part of the problem.
它还有另一部分
There’s another part.
如果你为机器输入一个目标
If you put an objective into a machine,
即便是一个很简单的目标 比如说”去把咖啡端来
even something as simple as, “Fetch the coffee,”
“机器会对自己说:”
the machine says to itself,
好吧 那我要怎么去拿咖啡呢?
“Well, how might I fail to fetch the coffee?
说不定有人会把我的电源关掉好吧
Someone might switch me off.
那我要想办法 阻止别人把我关掉
OK, I have to take steps to prevent that.
我得让我的‘关闭’开关失效
I will disable my ‘off’ switch.
我得尽一切可能自我防御 不让别人干涉我
I will do anything to defend myself against interference
这都是因为我被赋予的目标 ”
with this objective that I have been given.”
这种一根筋的思维
So this single-minded pursuit
以一种十分防御型的 模式去实现某一目标
in a very defensive mode of an objective that is, in fact,
实际上与我们人类最初 想实现的目标并不一致——
not aligned with the true objectives of the human race —
这就是我们面临的问题实际上
that’s the problem that we face.
这就是今天这个演讲的核心
And in fact, that’s the high-value takeaway from this talk.
如果你在我的演讲中只记住一件事
If you want to remember one thing,
那就是:如果你死了 你就不能去端咖啡了
it’s that you can’t fetch the coffee if you’re dead.
(笑声)
(Laughter)
这很简单 记住它就行了 每天对自己重复三遍
It’s very simple. Just remember that. Repeat it to yourself three times a day.
(笑声)
(Laughter)
实际上 这正是电影
And in fact, this is exactly the plot
《2001太空漫步》的剧情
of “2001: [A Space Odyssey]”
HAL有一个目标 一个任务
HAL has an objective, a mission,
但这个目标和人类的目标不一致
which is not aligned with the objectives of the humans,
这就导致了矛盾的产生
and that leads to this conflict.
幸运的是 HAL并不具备超级智能
Now fortunately, HAL is not superintelligent.
他挺聪明的 但还是 比不过人类主角戴夫
He’s pretty smart, but eventually Dave outwits him
戴夫成功地把HAL关掉了
and manages to switch him off.
但我们可能就没有这么幸运了
But we might not be so lucky.
那我们应该怎么办呢?
So what are we going to do?
我想要重新定义人工智能
I’m trying to redefine AI
远离传统的定义
to get away from this classical notion
将其仅限定为 机器通过智能去达成目标
of machines that intelligently pursue objectives.
新的定义涉及到三个原则:
There are three principles involved.
第一个原则是利他主义原则
The first one is a principle of altruism, if you like,
也就是说 机器的唯一目标
that the robot’s only objective
就是去最大化地实现人类的目标
is to maximize the realization of human objectives,
人类的价值
of human values.
至于价值 我指的不是感情化的价值
And by values here I don’t mean touchy-feely, goody-goody values.
而是指人类对生活所向往的
I just mean whatever it is that the human would prefer
无论是什么
their life to be like.
这实际上违背了阿西莫夫定律
And so this actually violates Asimov’s law
他指出机器人一定要维护自己的生存
that the robot has to protect its own existence.
但我定义的机器 对维护自身生存毫无兴趣
It has no interest in preserving its existence whatsoever.
第二个原则不妨称之为谦逊原则
The second law is a law of humility, if you like.
这一条对于制造安全的机器十分重要
And this turns out to be really important to make robots safe.
它说的是机器不知道
It says that the robot does not know
人类的价值是什么
what those human values are,
机器知道它需要将人类的价值最大化 却不知道这价值
so it has to maximize them, but it doesn’t know what they are.
究竟是什么 为了避免一根筋地追求
And that avoids this problem of single-minded pursuit
某一目标
of an objective.
这种不确定性是至关重要的
This uncertainty turns out to be crucial.
那机器为了对我们有用
Now, in order to be useful to us,
它就得掌握一些 关于我们想要什么的信息
it has to have some idea of what we want.
它主要通过观察人类 做的选择来获取这样的信息
It obtains that information primarily by observation of human choices,
我们自己做出的选择会包含着
so our own choices reveal information
关于我们希望我们的生活 是什么样的信息
about what it is that we prefer our lives to be like.
这就是三条原则
So those are the three principles.
让我们来看看它们是如何应用到
Let’s see how that applies to this question of:
像图灵说的那样 “将机器关掉”这个问题上来
“Can you switch the machine off?” as Turing suggested.
这是一个PR2机器人
So here’s a PR2 robot.
我们实验室里有一个
This is one that we have in our lab,
它的背面有一个大大的红色的开关
and it has a big red “off” switch right on the back.
那问题来了:它会让你把它关掉吗?
The question is: Is it going to let you switch it off?
如果我们按传统的方法
If we do it the classical way,
给它一个目标 让它拿咖啡 它会想:”我必须去拿咖啡
we give it the objective of, “Fetch the coffee, I must fetch the coffee,
但我死了就不能拿咖啡了 “
I can’t fetch the coffee if I’m dead,”
显然PR2听过我的演讲了
so obviously the PR2 has been listening to my talk,
所以它说:”我必须让我的开关失灵
and so it says, therefore, “I must disable my ‘off’ switch,
可能还要把那些在星巴克里
and probably taser all the other people in Starbucks
可能干扰我的人都电击一下 “
who might interfere with me.”
(笑声)
(Laughter)
这看起来必然会发生 对吗?
So this seems to be inevitable, right?
这种失败看起来是必然的
This kind of failure mode seems to be inevitable,
因为机器人在遵循 一个十分确定的目标
and it follows from having a concrete, definite objective.
那如果机器对目标 不那么确定会发生什么呢?
So what happens if the machine is uncertain about the objective?
那它的思路就不一样了
Well, it reasons in a different way.
它会说:”好的 人类可能会把我关掉
It says, “OK, the human might switch me off,
但只在我做错事的时候
but only if I’m doing something wrong.
我不知道什么是错事
Well, I don’t really know what wrong is,
但我知道我不该做那些事 ”
but I know that I don’t want to do it.”
这就是第一和第二原则
So that’s the first and second principles right there.
“那我就应该让人类把我关掉 ”
“So I should let the human switch me off.”
事实上你可以计算出机器人
And in fact you can calculate the incentive that the robot has
让人类把它关掉的动机
to allow the human to switch it off,
而且这个动机是
and it’s directly tied to the degree
与对目标的不确定程度直接相关的
of uncertainty about the underlying objective.
当机器被关闭后
And then when the machine is switched off,
第三条原则就起作用了
that third principle comes into play.
机器开始学习它所追求的目标
It learns something about the objectives it should be pursuing,
因为它知道它刚做的事是不对的
because it learns that what it did wasn’t right.
实际上 我们可以用希腊字母
In fact, we can, with suitable use of Greek symbols,
就像数学家们经常做的那样
as mathematicians usually do,
直接证明这一定理
we can actually prove a theorem
那就是这样的一个机器人 对人们是绝对有利的
that says that such a robot is provably beneficial to the human.
可以证明我们的生活 有如此设计的机器人会变得
You are provably better off with a machine that’s designed in this way
比没有这样的机器人更好
than without it.
这是一个很简单的例子 但这只是
So this is a very simple example, but this is the first step
我们尝试实现与人类 兼容的人工智能的第一步
in what we’re trying to do with human-compatible AI.
现在来看第三个原则
Now, this third principle,
我知道你们可能正在 为这一个原则而大伤脑筋
I think is the one that you’re probably scratching your head over.
你可能会想:“你知道 我有时不按规矩办事
You’re probably thinking, “Well, you know, I behave badly.
我可不希望我的机器人 像我一样行事
I don’t want my robot to behave like me.
我有时大半夜偷偷摸摸地 从冰箱里找东西吃
I sneak down in the middle of the night and take stuff from the fridge.
诸如此类的事 ”
I do this and that.”
有各种各样的事你是 不希望机器人去做的
There’s all kinds of things you don’t want the robot doing.
但实际上并不一定会这样
But in fact, it doesn’t quite work that way.
仅仅是因为你表现不好
Just because you behave badly
并不代表机器人就会复制你的行为
doesn’t mean the robot is going to copy your behavior.
它会去尝试理解你做事的动机 而且可能会在合适的情况下制止你去做
It’s going to understand your motivations and maybe help you resist them,
那些不该做的事
if appropriate.
但这仍然十分困难实际上
But it’s still difficult.
我们在做的是
What we’re trying to do, in fact,
让机器去预测任何一个人 在他们的任何一种
is to allow machines to predict for any person and for any possible life
可能的生活中
that they could live,
以及别人的生活中
and the lives of everybody else:
他们会更倾向于哪一种?
Which would they prefer?
这涉及到诸多困难;
And there are many, many difficulties involved in doing this;
我不认为这会很快地就被解决
I don’t expect that this is going to get solved very quickly.
实际上 真正的困难是我们自己
The real difficulties, in fact, are us.
就像我刚说的那样 我们做事不守规矩
As I have already mentioned, we behave badly.
我们中有的人甚至行为肮脏
In fact, some of us are downright nasty.
就像我说的 机器人并不会复制那些行为
Now the robot, as I said, doesn’t have to copy the behavior.
机器人没有自己的目标
The robot does not have any objective of its own.
它是完全无私的
It’s purely altruistic.
它的设计不是去满足 某一个人、一个用户的欲望
And it’s not designed just to satisfy the desires of one person, the user,
而是去尊重所有人的意愿
but in fact it has to respect the preferences of everybody.
所以它能对付一定程度的肮脏行为
So it can deal with a certain amount of nastiness,
它甚至能理解你的不端行为 比如说
and it can even understand that your nastiness, for example,
假如你是一个边境护照官员 很可能收取贿赂
you may take bribes as a passport official
因为你得养家、 得供你的孩子们上学
because you need to feed your family and send your kids to school.
机器人能理解这一点 它不会因此去偷
It can understand that; it doesn’t mean it’s going to steal.
它反而会帮助你去供孩子们上学
In fact, it’ll just help you send your kids to school.
我们的计算能力也是有限的
We are also computationally limited.
李世石是一个杰出的围棋大师
Lee Sedol is a brilliant Go player,
但他还是输了
but he still lost.
如果我们看他的行动 他最终输掉了棋局
So if we look at his actions, he took an action that lost the game.
但这不意味着他想要输
That doesn’t mean he wanted to lose.
所以要理解他的行为
So to understand his behavior,
我们得从人类认知模型来反过来想
we actually have to invert through a model of human cognition
这包含了我们的计算能力限制 是一个很复杂的模型
that includes our computational limitations — a very complicated model.
但仍然是我们可以尝试去理解的
But it’s still something that we can work on understanding.
可能对于我这样一个 人工智能研究人员来说最大的困难
Probably the most difficult part, from my point of view as an AI researcher,
是我们彼此各不相同
is the fact that there are lots of us,
所以机器必须想办法去判别衡量
and so the machine has to somehow trade off, weigh up the preferences
不同人的不同需求
of many different people,
而又有众多方法去做这样的判断
and there are different ways to do that.
经济学家、社会学家、 哲学家都理解这一点
Economists, sociologists, moral philosophers have understood that,
我们正在积极地去寻求合作
and we are actively looking for collaboration.
让我们来看看如果我们 把这一步弄错了会怎么样
Let’s have a look and see what happens when you get that wrong.
举例来说 你可能会 与你的人工智能助理
So you can have a conversation, for example,
有这样的对话:
with your intelligent personal assistant
这样的人工智能可能几年内就会出现
that might be available in a few years’ time.
可以把它想做加强版的Siri
Think of a Siri on steroids.
Siri对你说:“你的妻子打电话 提醒你今晚要跟她共进晚餐 ”
So Siri says, “Your wife called to remind you about dinner tonight.”
而你呢 自然忘了这回事: “什么?什么晚饭?
And of course, you’ve forgotten. “What? What dinner?
你在说什么?”
What are you talking about?”
“啊 你们晚上7点 庆祝结婚20周年纪念日 ”
“Uh, your 20th anniversary at 7pm.”
“我可去不了 我约了晚上7点半见领导
“I can’t do that. I’m meeting with the secretary-general at 7:30.
怎么会这样呢?”
How could this have happened?”
“呃 我可是提醒过你的 但你不听我的建议 ”
“Well, I did warn you, but you overrode my recommendation.”
“我该怎么办呢?我可不能 跟领导说我有事 没空见他 ”
“Well, what am I going to do? I can’t just tell him I’m too busy.”
“别担心 我已经安排了 让他的航班延误
“Don’t worry. I arranged for his plane to be delayed.”
(笑声)
(Laughter)
“像是因为某种计算机故障那样 ”
“Some kind of computer malfunction.”
(笑声)
(Laughter)
“真的吗?这个你也能做到?”
“Really? You can do that?”
“领导很不好意思 跟你道歉
“He sends his profound apologies
并且告诉你明天 中午午饭不见不散 ”
and looks forward to meeting you for lunch tomorrow.”
(笑声)
(Laughter)
这里就有一个小小的问题
So the values here — there’s a slight mistake going on.
这显然是在遵循我妻子的价值论
This is clearly following my wife’s values
那就是“老婆开心 生活舒心”
which is “Happy wife, happy life.”
(笑声)
(Laughter)
它也有可能发展成另一种情况
It could go the other way.
你忙碌一天 回到家里
You could come home after a hard day’s work,
电脑对你说:“像是繁忙的一天啊?”
and the computer says, “Long day?”
“是啊 我连午饭都没来得及吃 ”
“Yes, I didn’t even have time for lunch.”
“那你一定很饿了吧 ”
“You must be very hungry.”
“快饿晕了 你能做点晚饭吗?”
“Starving, yeah. Could you make some dinner?”
“有一件事我得告诉你
“There’s something I need to tell you.”
(笑声)”
(Laughter)
南苏丹的人们可比你更需要照顾
“There are humans in South Sudan who are in more urgent need than you.”
(笑声)
(Laughter)
“所以我要离开了 你自己做饭去吧 ”
“So I’m leaving. Make your own dinner.”
(笑声)
(Laughter)
我们得解决这些问题
So we have to solve these problems,
我也很期待去解决
and I’m looking forward to working on them.
我们有理由感到乐观
There are reasons for optimism.
理由之一是
One reason is,
我们有大量的数据
there is a massive amount of data.
记住 我说过机器将能够阅读一切 人类所写下来的东西
Because remember — I said they’re going to read everything
而我们写下的大多数是
the human race has ever written.
我们做的什么事情
Most of what we write about is human beings doing things
以及其他人对此有什么意见
and other people getting upset about it.
所以机器可以从大量的数据中去学习
So there’s a massive amount of data to learn from.
同时从经济的角度 我们也有足够的动机
There’s also a very strong economic incentive
去把这件事做对
to get this right.
想象一下 你家里有个居家机器人
So imagine your domestic robot’s at home.
而你又得加班 机器人得给孩子们做饭
You’re late from work again and the robot has to feed the kids,
孩子们很饿 但冰箱里什么都没有
and the kids are hungry and there’s nothing in the fridge.
然后机器人看到了家里的猫
And the robot sees the cat.
(笑声)
(Laughter)
机器人还没学透人类的价值论
And the robot hasn’t quite learned the human value function properly,
所以它不知道
so it doesn’t understand
猫的感情价值 大于猫的营养价值
the sentimental value of the cat outweighs the nutritional value of the cat.
(笑声)
(Laughter)
接下来会发生什么?
So then what happens?
差不多是这样的:
Well, it happens like this:
头版头条:“疯狂的机器人 把猫煮了给主人当晚饭!”
“Deranged robot cooks kitty for family dinner.”
这一个事故就足以结束 整个居家机器人产业
That one incident would be the end of the domestic robot industry.
所以我们有足够的动机在我们实现
So there’s a huge incentive to get this right
超级智能机器让它更加完善
long before we reach superintelligent machines.
总结来说:
So to summarize:
我想要改变人工智能的定义
I’m actually trying to change the definition of AI
让我们可以证明机器对我们是有利的
so that we have provably beneficial machines.
这三个原则是:
And the principles are:
机器是利他的
machines that are altruistic,
只想着实现我们的目标
that want to achieve only our objectives,
但它不确定我们的目标是什么
but that are uncertain about what those objectives are,
所以它会观察我们
and will watch all of us
从中学习我们想要的究竟是什么
to learn more about what it is that we really want.
希望在这个过程中 我们也能学会成为更好的人
And hopefully in the process, we will learn to be better people.
谢谢大家
Thank you very much.
(掌声)
(Applause)
克里斯安德森: 非常有意思 斯图尔特 我们趁着工作人员
Chris Anderson: So interesting, Stuart.
为下一位演讲者布置的时候
We’re going to stand here a bit because I think they’re setting up
来简单聊几句
for our next speaker.
我有几个问题
A couple of questions.
从直觉上来看 将无知编入到程序中 似乎是一个很重要的理念
So the idea of programming in ignorance seems intuitively really powerful.
当你要实现超级智能时
As you get to superintelligence,
什么能阻止机器人?
what’s going to stop a robot
当它在阅读和学习的过程中发现
reading literature and discovering this idea that knowledge
知识比无知更强大
is actually better than ignorance
然后就改变它的目标 去重新编写程序呢?
and still just shifting its own goals and rewriting that programming?
斯图尔特拉塞尔:是的 我们想要它去学习 就像我说的
Stuart Russell: Yes, so we want it to learn more, as I said,
学习我们的目标
about our objectives.
它只有在理解得越来越正确的时候 才会变得更确定
It’ll only become more certain as it becomes more correct,
我们有证据显示
so the evidence is there
它的设计使它能按正确的方式理解比如说
and it’s going to be designed to interpret it correctly.
它能够理解书中的论证是
It will understand, for example, that books are very biased
带有非常强的偏见的
in the evidence they contain.
书中只会讲述国王、王子
They only talk about kings and princes
和那些精英白人男性做的事
and elite white male people doing stuff.
这是一个复杂的问题
So it’s a complicated problem,
但当它更深入地学习我们的目标时
but as it learns more about our objectives
它就变得对我们更有用
it will become more and more useful to us.
CA:那你不能把这些 都集中在一条准则里吗?
CA: And you couldn’t just boil it down to one law,
把这样的命令写在它的程序里:
you know, hardwired in:
“如果人类什么时候想把我关掉
“if any human ever tries to switch me off,
我服从 我服从 ”
I comply. I comply.”
SR:绝对不行
SR: Absolutely not.
那将是一个很糟糕的主意
That would be a terrible idea.
试想一下 你有一辆无人驾驶汽车
So imagine that you have a self-driving car
你想让它送你五岁的孩子
and you want to send your five-year-old
去上学
off to preschool.
你希望你五岁的孩子 能在汽车运行过程中
Do you want your five-year-old to be able to switch off the car
将它关闭吗?
while it’s driving along?
应该不会吧它得理解
Probably not.
下指令的人有多理智 是不是讲道理
So it needs to understand how rational and sensible the person is.
这个人越理智
The more rational the person,
它就越愿意自己被关掉
the more willing you are to be switched off.
如果这个人是完全思绪混乱 或者甚至是有恶意的
If the person is completely random or even malicious,
那你就不愿意它被关掉
then you’re less willing to be switched off.
CA:好吧 斯图尔特 我得说
CA: All right. Stuart, can I just say,
我真的希望你为我们 能把这一切研究出来
I really, really hope you figure this out for us.
很感谢你的演讲 太精彩了 SR:
Thank you so much for that talk. That was amazing.
谢谢
SR: Thank you.
(掌声)
(Applause)

发表评论

译制信息
视频概述
听录译者

收集自网络

翻译译者

收集自网络

审核员

自动通过审核

视频来源

https://www.youtube.com/watch?v=EBK-a94IFHY

相关推荐