Machine intelligence makes human morals more important | Zeynep Tufekci

翻译人员: Yangyang Liu 校对人员: Junyi Sha

So, I started my first job as a computer programmer

我的第一份工作是程序员，

in my very first year of college --

那是在我刚上大学的时候，

basically, as a teenager.

不到二十岁。

Soon after I started working,

我刚开始工作不久，

writing software in a company,

正当在公司写程序，

a manager who worked at the company came down to where I was,

公司的一位经理来到我旁边，

and he whispered to me,

他悄悄的对我说，

"Can he tell if I'm lying?"

“他能看出来我在撒谎吗？”

There was nobody else in the room.

当时屋子里没有别人。

“你是指谁能看出你在撒谎？ 还有，我们干嘛要悄悄地说话？”

"Can who tell if you're lying? And why are we whispering?"

The manager pointed at the computer in the room.

那个经理指着屋子里的电脑，说：

"Can he tell if I'm lying?"

“他能看出我在撒谎吗？”

Well, that manager was having an affair with the receptionist.

其实，那个经理和前台有一腿。

(Laughter)

（笑声）

当时我只有十来岁，

And I was still a teenager.

So I whisper-shouted back to him,

我低声地回答他，

"Yes, the computer can tell if you're lying."

“是的，电脑什么都知道。”

（笑声）

(Laughter)

Well, I laughed, but actually, the laugh's on me.

我笑了，但其实我是在笑自己，

Nowadays, there are computational systems

现在，计算机系统已经可以

that can suss out emotional states and even lying

通过分析人脸来辨别人的情绪，

from processing human faces.

甚至包括是否在撒谎。

Advertisers and even governments are very interested.

广告商，甚至政府 都对此很感兴趣。

I had become a computer programmer

我选择成为电脑程序员，

because I was one of those kids crazy about math and science.

因为我是那种 痴迷于数学和科学孩子。

But somewhere along the line I'd learned about nuclear weapons,

其间我也学习过核武器，

我也非常关心科学伦理。

and I'd gotten really concerned with the ethics of science.

我曾经很困惑。

I was troubled.

However, because of family circumstances,

但是，因为家庭原因，

I also needed to start working as soon as possible.

我需要尽快参加工作。

So I thought to myself, hey, let me pick a technical field

我对自己说，嘿， 选一个容易找工作

where I can get a job easily

的科技领域吧，

and where I don't have to deal with any troublesome questions of ethics.

并且找个不需要 操心伦理问题的。

所以我选了计算机。

So I picked computers.

(Laughter)

（笑声）

Well, ha, ha, ha! All the laughs are on me.

哈哈哈， 我多可笑。

如今，计算机科学控制着

Nowadays, computer scientists are building platforms

that control what a billion people see every day.

十亿人每天能看到的信息，

它们可以控制汽车朝哪里开，

They're developing cars that could decide who to run over.

They're even building machines, weapons,

它们可以建造机器、武器，

that might kill human beings in war.

那些在战争中 用于杀人的武器。

It's ethics all the way down.

说到底， 都是伦理问题。

Machine intelligence is here.

机器智能来了。

We're now using computation to make all sort of decisions,

我们用计算机来做各种决策，

but also new kinds of decisions.

包括人们面临的新决策。

We're asking questions to computation that have no single right answers,

我们向计算机询问多解的、

that are subjective

主观的、

and open-ended and value-laden.

开放性的或意义深远的问题。

我们会问，

We're asking questions like,

"Who should the company hire?"

“我们公司应该聘请谁？”

“你该关注哪个朋友 的哪条状态？”

"Which update from which friend should you be shown?"

"Which convict is more likely to reoffend?"

“哪种犯罪更容易再犯？”

"Which news item or movie should be recommended to people?"

“应该给人们推荐 哪条新闻或是电影？”

Look, yes, we've been using computers for a while,

看，是的，我们使用计算机 已经有一段时间了，

but this is different.

但现在不一样了。

This is a historical twist,

这是历史性的转折，

because we cannot anchor computation for such subjective decisions

因为我们在这些主观决策上 无法主导计算机，

the way we can anchor computation for flying airplanes, building bridges,

不像我们在 管理飞机、建造桥梁、

登月等问题上， 可以主导它们。

going to the moon.

Are airplanes safer? Did the bridge sway and fall?

飞机会更安全吗？ 桥梁会摇晃或倒塌吗？

There, we have agreed-upon, fairly clear benchmarks,

在这些问题上，我们 有统一而清晰的判断标准，

and we have laws of nature to guide us.

我们有自然定律来指导。

We have no such anchors and benchmarks

但是在复杂的人类事务上，

for decisions in messy human affairs.

我们没有这样的客观标准。

To make things more complicated, our software is getting more powerful,

让问题变得更复杂的， 是我们的软件正越来越强大，

but it's also getting less transparent and more complex.

同时也变得更加不透明， 更加复杂。

Recently, in the past decade,

最近的几十年，

complex algorithms have made great strides.

复杂算法已 取得了长足发展，

They can recognize human faces.

它们可以识别人脸，

They can decipher handwriting.

它们可以破解笔迹，

They can detect credit card fraud

它们可以识别信用卡欺诈，

and block spam

可以屏蔽垃圾信息，

and they can translate between languages.

它们可以翻译语言，

They can detect tumors in medical imaging.

他们可以通过 医学图像识别肿瘤，

They can beat humans in chess and Go.

它们可以在国际象棋 和围棋上击败人类。

Much of this progress comes from a method called "machine learning."

类似的很多发展， 都来自一种叫“机器学习”的方法。

机器学习不像传统程序一样，

Machine learning is different than traditional programming,

where you give the computer detailed, exact, painstaking instructions.

需要给计算机详细、 准确的逐条指令。

It's more like you take the system and you feed it lots of data,

它更像是你给系统 喂了很多数据，

including unstructured data,

包括非结构化数据，

like the kind we generate in our digital lives.

比如我们在数字生活中 产生的数据。

And the system learns by churning through this data.

系统扎进这些数据中学习，

And also, crucially,

重要的是，

these systems don't operate under a single-answer logic.

这些系统不再局限单一答案。

They don't produce a simple answer; it's more probabilistic:

他们得出的不是一个 简单的答案，而是概率性的：

"This one is probably more like what you're looking for."

“这个更像是你在寻找的。”

它的优势是： 它真的非常强大。

Now, the upside is: this method is really powerful.

Google 人工智能系统的 负责人称它为：

The head of Google's AI systems called it,

"the unreasonable effectiveness of data."

“不可思议的数据效率”。

The downside is,

缺点在于，

we don't really understand what the system learned.

我们无法清楚的了解 系统学到了什么，

In fact, that's its power.

事实上，这也正是 它的强大之处。

This is less like giving instructions to a computer;

不像是给计算机下达指令，

it's more like training a puppy-machine-creature

更像是在训练一个机器狗，

we don't really understand or control.

我们无法精确的 了解和控制它。

So this is our problem.

这就是我们遇到的问题。

It's a problem when this artificial intelligence system gets things wrong.

人工智能会出错， 这是一个问题。

It's also a problem when it gets things right,

但他们得出正确答案， 又是另一种问题。

because we don't even know which is which when it's a subjective problem.

因为我们面对主观问题， 是不应该有答案的。

We don't know what this thing is thinking.

我们不知道 这些机器在想什么。

So, consider a hiring algorithm --

所以，考虑一下招聘算法－

通过机器学习构建的招聘系统。

a system used to hire people, using machine-learning systems.

这样的系统会用员工 现有的数据进行自我培训，

Such a system would have been trained on previous employees' data

and instructed to find and hire

参照公司的优秀员工

people like the existing high performers in the company.

来寻找和招聘新人。

Sounds good.

听起来很好。

I once attended a conference

有次我参加了一个会议，

会上聚集了很多 人力资源部的经理和总监，

that brought together human resources managers and executives,

都是高管，

high-level people,

using such systems in hiring.

让他们使用这样的招聘系统。

They were super excited.

他们都非常兴奋，

They thought that this would make hiring more objective, less biased,

认为这可以让招聘变得 更加客观，从而减少偏见，

and give women and minorities a better shot

给女性和少数族裔 更多的机会，

against biased human managers.

减少他们自身的偏见。

And look -- human hiring is biased.

你知道的， 招聘是存在偏见的，

我也很清楚。

I know.

I mean, in one of my early jobs as a programmer,

在我刚开始做程序员的时候，

my immediate manager would sometimes come down to where I was

我的直接主管会来找我，

really early in the morning or really late in the afternoon,

在早晨很早或下午很晚的时候，

说，“ 图费, 我们去吃午饭！”

and she'd say, "Zeynep, let's go to lunch!"

I'd be puzzled by the weird timing.

我就被这奇怪的时间 给搞糊涂了，

It's 4pm. Lunch?

现在是下午4点，吃午饭？

我当时很穷，所以 不会放过免费的午餐。

I was broke, so free lunch. I always went.

I later realized what was happening.

后来我才想明白原因，

My immediate managers had not confessed to their higher-ups

我的主管们没有 向他们的上级坦白，

that the programmer they hired for a serious job was a teen girl

他们雇了一个十多岁的小女孩 来做重要的编程工作，

who wore jeans and sneakers to work.

一个穿着牛仔裤， 运动鞋工作的女孩。

我的工作做得很好， 我只是看起来不合适，

I was doing a good job, I just looked wrong

and was the wrong age and gender.

年龄和性别也不合适。

所以，忽略性别和种族的招聘，

So hiring in a gender- and race-blind way

certainly sounds good to me.

听起来很适合我。

但是这样的系统会带来更多问题，

But with these systems, it is more complicated, and here's why:

Currently, computational systems can infer all sorts of things about you

当前，计算机系统 能根据零散的数据，

from your digital crumbs,

推断出关于你的一切，

even if you have not disclosed those things.

甚至你没有公开的事。

They can infer your sexual orientation,

它们可以推断你的性取向，

your personality traits,

你的性格特点，

your political leanings.

你的政治倾向。

They have predictive power with high levels of accuracy.

它们有高准确度的预测能力，

Remember -- for things you haven't even disclosed.

记住，是你没有公开的事情，

This is inference.

这就是推断。

I have a friend who developed such computational systems

我有个朋友 就是开发这种系统，

to predict the likelihood of clinical or postpartum depression

从社交媒体的数据中，

推断患临床或 产后抑郁症的可能性。

from social media data.

The results are impressive.

结果令人印象深刻，

Her system can predict the likelihood of depression

她的系统可以 在症状出现前几个月

months before the onset of any symptoms --

成功预测到 患抑郁的可能性，

months before.

提前几个月。

No symptoms, there's prediction.

在有症状之前， 就可以预测到，

She hopes it will be used for early intervention. Great!

她希望这可以用于 临床早期干预，这很棒！

But now put this in the context of hiring.

现在我们把这项技术 放到招聘中来看。

在那次人力资源管理会议中，

So at this human resources managers conference,

我接近了一位大公司的高管，

I approached a high-level manager in a very large company,

and I said to her, "Look, what if, unbeknownst to you,

我对她说，“看，如果这个系统 在不通知你的情况下，

your system is weeding out people with high future likelihood of depression?

就剔除了未来 有可能抑郁的人，怎么办？

They're not depressed now, just maybe in the future, more likely.

他们现在不抑郁， 只是未来有可能。

What if it's weeding out women more likely to be pregnant

如果它剔除了 有可能怀孕的女性，怎么办？

in the next year or two but aren't pregnant now?

她们现在没怀孕， 但未来一两年有可能。

What if it's hiring aggressive people because that's your workplace culture?"

如果因为你的公司文化， 它只雇佣激进的候选人怎么办？”

只看性别比例， 你发现不了这些问题，

You can't tell this by looking at gender breakdowns.

Those may be balanced.

性别比例是可以被调整的。

And since this is machine learning, not traditional coding,

并且因为这是机器学习， 不是传统的代码，

there is no variable there labeled "higher risk of depression,"

不会有一个变量来标识 “高抑郁风险”、

"higher risk of pregnancy,"

“高怀孕风险”、

"aggressive guy scale."

“人员的激进程度”。

Not only do you not know what your system is selecting on,

你不仅无法了解系统 在选什么样的人，

you don't even know where to begin to look.

你甚至不知道 从哪里入手了解。

它是个暗箱。

It's a black box.

It has predictive power, but you don't understand it.

它有预测的能力， 但你不了解它。

"What safeguards," I asked, "do you have

我问，“你有什么措施 可以保证，

to make sure that your black box isn't doing something shady?"

你的暗箱没有 在做些见不得人的事？”

She looked at me as if I had just stepped on 10 puppy tails.

她看着我，就好像 我刚踩了10只小狗的尾巴。

(Laughter)

（笑声）

她瞪着我说：

She stared at me and she said,

"I don't want to hear another word about this."

“我不想再听你多说一个字。”

And she turned around and walked away.

然后她转身走开了。

其实， 她不是无礼，

Mind you -- she wasn't rude.

It was clearly: what I don't know isn't my problem, go away, death stare.

她想表达的其实是：我不知道， 这不是我的错，走开，不然我瞪死你。

(Laughter)

（笑声）

Look, such a system may even be less biased

看，这样的系统 可能在某些方面

than human managers in some ways.

比人类高管 怀有更少偏见，

And it could make monetary sense.

而且可以创造经济价值。

But it could also lead

但它也可能

to a steady but stealthy shutting out of the job market

用一种顽固且隐秘的方式，

把高抑郁风险的人清出职场。

of people with higher risk of depression.

Is this the kind of society we want to build,

这是我们想要的未来吗？

without even knowing we've done this,

把决策权给予我们 并不完全了解的机器，

because we turned decision-making to machines we don't totally understand?

在我们不知情的状况下 构建一种新的社会？

Another problem is this:

另一个问题是，

these systems are often trained on data generated by our actions,

这些系统通常使用

human imprints.

我们真实的 行为数据来训练。

Well, they could just be reflecting our biases,

它们可能只是在 反馈我们的偏见，

这些系统会 继承我们的偏见，

and these systems could be picking up on our biases

and amplifying them

并把它们放大，

and showing them back to us,

然后反馈给我们。

while we're telling ourselves,

我们骗自己说，

"We're just doing objective, neutral computation."

“我们只做客观、 中立的预测。”

Researchers found that on Google,

研究者发现，在 Google 上，

高收入工作的广告 更多的被展示给男性用户。

women are less likely than men to be shown job ads for high-paying jobs.

And searching for African-American names

搜索非裔美国人的名字，

更可能出现 关于犯罪史的广告，

is more likely to bring up ads suggesting criminal history,

even when there is none.

即使某些根本不存在。

Such hidden biases and black-box algorithms

这些潜在的偏见 以及暗箱中的算法，

that researchers uncover sometimes but sometimes we don't know,

有些会被研究者揭露， 有些根本不会被发现，

can have life-altering consequences.

它的后果可能是 改变一个人的人生。

In Wisconsin, a defendant was sentenced to six years in prison

在威斯康星，一个被告

因逃避警察被判刑六年。

for evading the police.

You may not know this,

你可能不知道，

但计算机算法正越来越多的 被应用在假释及量刑裁定上。

but algorithms are increasingly used in parole and sentencing decisions.

他想要弄清楚，这个 得分是怎么算出来的？

He wanted to know: How is this score calculated?

It's a commercial black box.

这是个商业暗箱，

The company refused to have its algorithm be challenged in open court.

这家公司拒绝在公开法庭上 讨论他们的算法。

But ProPublica, an investigative nonprofit, audited that very algorithm

但是一家叫 ProPublica 的非盈利机构，

根据公开数据， 对这个算法进行了评估，

with what public data they could find,

and found that its outcomes were biased

他们发现这个算法 的结论是有偏见的，

and its predictive power was dismal, barely better than chance,

它的预测能力很差， 比碰运气强不了多少，

and it was wrongly labeling black defendants as future criminals

并且它错误的把黑人被告 未来犯罪的可能性

at twice the rate of white defendants.

标记为白人的两倍。

So, consider this case:

看下这个案例：

这个女人急着去佛罗里达州, 布劳沃德县的一所学校，

This woman was late picking up her godsister

from a school in Broward County, Florida,

去接她的干妹妹。

running down the street with a friend of hers.

女人和她的朋友在街上狂奔，

她们看到门廊上一辆没上锁的 儿童自行车，和一辆电瓶车，

They spotted an unlocked kid's bike and a scooter on a porch

and foolishly jumped on it.

于是就愚蠢的骑上了车。

As they were speeding off, a woman came out and said,

正在她们要骑走的时候， 另一个女人出来，喊道：

"Hey! That's my kid's bike!"

“嘿！那是我孩子的自行车！”

They dropped it, they walked away, but they were arrested.

她们扔掉车走开， 但还是被抓住了。

她做错了，她很愚蠢， 但她也才刚满18岁，

She was wrong, she was foolish, but she was also just 18.

She had a couple of juvenile misdemeanors.

她之前有不少 青少年轻罪的记录。

Meanwhile, that man had been arrested for shoplifting in Home Depot --

与此同时，这个男人 在连锁超市偷窃被捕了，

偷了价值85美金的东西， 同样的轻微犯罪，

85 dollars' worth of stuff, a similar petty crime.

But he had two prior armed robbery convictions.

但他有两次持枪抢劫的案底。

But the algorithm scored her as high risk, and not him.

这个程序将这位女性判定为 高风险，而这位男性则不是。

Two years later, ProPublica found that she had not reoffended.

两年后，ProPublica 发现她没有再次犯罪，

It was just hard to get a job for her with her record.

但这个记录 使她很难找到工作。

He, on the other hand, did reoffend

而这位男性，却再次犯罪，

and is now serving an eight-year prison term for a later crime.

并因此被判八年监禁。

显然，我们需要 审查这些暗箱，

Clearly, we need to audit our black boxes

and not have them have this kind of unchecked power.

确保它们不再有这样 不加限制的权限。

（掌声）

(Applause)

审查是很重要的， 但不能解决所有的问题。

Audits are great and important, but they don't solve all our problems.

Take Facebook's powerful news feed algorithm --

拿 Facebook 的强大的 新闻流算法来说，

就是通过你的朋友圈 和你浏览过的页面，

you know, the one that ranks everything and decides what to show you

决定你的 “推荐内容”的算法。

from all the friends and pages you follow.

Should you be shown another baby picture?

它会决定要不要 再推一张婴儿照片给你，

(Laughter)

（笑声）

A sullen note from an acquaintance?

要不要推一条熟人 的沮丧状态？

An important but difficult news item?

要不要推一条重要 但艰涩的新闻？

There's no right answer.

这个问题没有正解。

Facebook optimizes for engagement on the site:

Facebook 会根据 网站的参与度来优化：

likes, shares, comments.

喜欢、分享、评论。

在2014年8月，

In August of 2014,

protests broke out in Ferguson, Missouri,

密苏里州弗格森市爆发了游行，

after the killing of an African-American teenager by a white police officer,

一个白人警察在不明状况下

杀害了一位非裔少年。

under murky circumstances.

The news of the protests was all over

关于游行的新闻

在我的未经算法过滤的 Twitter 上大量出现，

my algorithmically unfiltered Twitter feed,

but nowhere on my Facebook.

但 Facebook 上却没有。

Was it my Facebook friends?

是因为我的 Facebook 好友 不关注这事吗？

I disabled Facebook's algorithm,

我禁用了 Facebook 的算法，

which is hard because Facebook keeps wanting to make you

这是很麻烦的一键事， 因为 Facebook 希望

come under the algorithm's control,

你一直在它的算法 控制下使用,

and saw that my friends were talking about it.

希望我的朋友持续 地谈论这件事。

It's just that the algorithm wasn't showing it to me.

只是算法没法 给我这些信息。

I researched this and found this was a widespread problem.

我研究了这个现象， 发现这是个普遍的问题。

The story of Ferguson wasn't algorithm-friendly.

弗格森事件 对算法是不适用的，

它不是值得“赞”的新闻，

It's not "likable."

Who's going to click on "like?"

谁会在这样 的文章下点“赞”呢？

It's not even easy to comment on.

甚至这新闻都不好被评论。

Without likes and comments,

因为没有“赞”和评论，

算法会减少 这些新闻的曝光，

the algorithm was likely showing it to even fewer people,

so we didn't get to see this.

所以我们无法看到。

Instead, that week,

相反的，在同一周，

Facebook's algorithm highlighted this,

Facebook 的算法热推了

which is the ALS Ice Bucket Challenge.

ALS 冰桶挑战的信息。

Worthy cause; dump ice water, donate to charity, fine.

这很有意义，倒冰水， 为慈善捐款，很好。

But it was super algorithm-friendly.

这个事件对算法是很适用的，

The machine made this decision for us.

机器帮我们做了这个决定。

A very important but difficult conversation

非常重要但艰涩的新闻事件

might have been smothered,

可能会被埋没掉，

had Facebook been the only channel.

因为 Facebook 已经成为 主要的信息来源。

Now, finally, these systems can also be wrong

最后，这些系统 也可能会在一些

in ways that don't resemble human systems.

不同于人力系统 的那些事情上搞错。

Do you guys remember Watson, IBM's machine-intelligence system

你们记得 Watson 吧， 那个在智力竞赛《危险边缘》中

that wiped the floor with human contestants on Jeopardy?

横扫人类选手的 IBM 机器智能系统，

It was a great player.

它是个很厉害的选手。

But then, for Final Jeopardy, Watson was asked this question:

但是，在最后一轮比赛中， Watson 被问道：

"Its largest airport is named for a World War II hero,

“它最大的机场是以 二战英雄命名的，

its second-largest for a World War II battle."

它第二大机场是以 二战战场命名的。”

(Hums Final Jeopardy music)

（哼唱《危险边缘》插曲）

Chicago.

芝加哥。

The two humans got it right.

两位人类选手答对了，

Watson, on the other hand, answered "Toronto" --

但 Watson 答的是， “多伦多”，

这是个猜美国城市的环节！

for a US city category!

The impressive system also made an error

这个厉害的系统也会犯

that a human would never make, a second-grader wouldn't make.

人类都不会犯的，二年级 小孩都不会犯的错误。

Our machine intelligence can fail

我们的机器智能系统,

in ways that don't fit error patterns of humans,

会在一些不符合人类 出错模式的问题上出错，

这些问题都是我们 无法预料和准备的。

in ways we won't expect and be prepared for.

丢失一份完全有能力胜任 的工作时，人们会感到很糟，

It'd be lousy not to get a job one is qualified for,

but it would triple suck if it was because of stack overflow

但如果是因为机器 子程序的过度堆积，

in some subroutine.

就简直糟透了。

(Laughter)

（笑声）

In May of 2010,

在2010年五月，

a flash crash on Wall Street fueled by a feedback loop

华尔街出现一次 股票闪电崩盘，

原因是“卖出”算法 的反馈回路导致，

in Wall Street's "sell" algorithm

wiped a trillion dollars of value in 36 minutes.

在36分钟内 损失了几十亿美金。

I don't even want to think what "error" means

我甚至不敢想， 致命的自动化武器

in the context of lethal autonomous weapons.

发生“错误”会是什么后果。

So yes, humans have always made biases.

是的，人类总是会有偏见，

Decision makers and gatekeepers,

法庭上、新闻机构、战争中的，

in courts, in news, in war ...

决策者、看门人…

they make mistakes; but that's exactly my point.

他们都会犯错， 但这恰恰是我要说的。

We cannot escape these difficult questions.

我们无法抛开 这些困难的问题，

我们不能把我们自身 该承担的责任推给机器。

We cannot outsource our responsibilities to machines.

(Applause)

（掌声）

人工智能不会给我们 一张“伦理免责卡”。

Artificial intelligence does not give us a "Get out of ethics free" card.

Data scientist Fred Benenson calls this math-washing.

数据科学家 Fred Benenson 称之为“数学粉饰”。

我们需要是相反的东西。

We need the opposite.

We need to cultivate algorithm suspicion, scrutiny and investigation.

我们需要培养算法的 怀疑、复查和调研能力。

We need to make sure we have algorithmic accountability,

我们需要确保 有人为算法负责，

auditing and meaningful transparency.

为算法审查， 并切实的公开透明。

We need to accept that bringing math and computation

我们必须认识到， 把数学和计算引入

to messy, value-laden human affairs

解决复杂的、高价值 的人类事务中，

does not bring objectivity;

并不能带来客观性，

相反，人类事务 的复杂性会扰乱算法。

rather, the complexity of human affairs invades the algorithms.

是的，我们可以 并且需要使用计算机

Yes, we can and we should use computation

to help us make better decisions.

来帮助我们做更好的决策，

But we have to own up to our moral responsibility to judgment,

但我们也需要在判断中 加入道德义务，

在这个框架下使用算法，

and use algorithms within that framework,

not as a means to abdicate and outsource our responsibilities

而不是像人与人 之间相互推卸那样，

就把责任转移给机器。

to one another as human to human.

Machine intelligence is here.

人工智能到来了，

That means we must hold on ever tighter

这意味着 我们要格外坚守

人类的价值观和伦理。

to human values and human ethics.

谢谢。

Thank you.

(Applause)

（掌声）