War, AI and the New Global Arms Race | Alexandr Wang | TED

翻译人员: Yip Yan Yeung 校对人员: suya f.

（注：本视频立场 不代表 TED 译者意见）

Artificial intelligence and warfare.

人工智能与战争。

Let's talk about what this really could look like.

我们来聊一聊 这究竟会是什么样的。

Swarms of lethal drones with facial recognition

成群的致命性无人机 携带人脸识别功能，

that know your every move.

对你的一举一动了如指掌。

Or unmanned armed robots that are near impossible to defeat.

无人操纵的武装机器人 几乎立于不败之地。

Autonomous fighter jets that can travel at supersonic speeds

自动驾驶战斗机以超音速飞行，

and can withstand greater gravitational force

可以承受大于 人类飞行员能够存活的重力。

than a human pilot could survive.

Cyberattacks that incapacitate critical port infrastructure

让关键港口设施 无法正常运行的网络攻击

or disinformation campaigns and deepfakes that throw presidential elections.

或充满虚假信息的宣传 和扰乱总统选举的深度伪造。

Or even foreign adversaries taking out satellites,

甚至还有境外敌对势力破坏卫星——

our eyes and ears in space,

我们在太空中的眼睛和耳朵，

rendering us blind to global events.

让我们无从知晓全球事件。

All superintelligent weapons of terror.

这些都是超智能的恐怖武器。

We are at the dawn of a new age of warfare.

我们已经走到了战争新时代的黎明。

I grew up in the birthplace of a technology

我长大的地方是一项

that defined the last era of warfare,

定义了上一个战争时代的科技——

the atomic bomb.

原子弹的诞生地。

I was keenly aware of how this technology had fundamentally shaped geopolitics

我深知这项科技 如何彻底改变了地缘政治

and the nature of war.

和战争的本质。

My parents were both scientists at Los Alamos National Laboratory.

我的父母都是洛斯·阿拉莫斯 国家实验室的科学家。

My dad’s a physicist, and my mom’s an astrophysicist.

我的父亲是一名物理学家， 母亲是一名天体物理学家。

Their scientific work in plasma fluid dynamics

他们在等离子体流体动力学中的科研成果

will have deep implications for how we understand our universe.

会对我们理解宇宙的方式 产生深远的影响。

So naturally, I knew I wanted to work on something just as impactful.

所以，我自然而然知道自己 也想从事如此影响深远的工作。

I decided to become a programmer and study artificial intelligence.

我决定成为一名程序员， 研究人工智能。

AI is one of the most critical technologies of our time

AI 是当今最重要的科技之一，

and with deep implications for national security

会对国家安全和全球民主 产生巨大的影响。

and democracy globally.

As we saw in World War II with the atomic bomb,

我们在第二次世界大战中 见识了原子弹，

the country that is able to most rapidly and effectively

那能最快、最有效

integrate new technology into warfighting wins.

在战争中用上新技术的国家 就会取得胜利。

There's no reason to believe this will be any different with AI.

AI 也绝对没有可能会成为例外。

But in the AI arms race, we're already behind.

但在 AI 的军备竞赛中， 我们已经落后了。

From a technological perspective,

从技术角度来看，

China is already ahead of the United States

中国在计算机视觉 AI 上 已经领先美国。

in computer vision AI.

And in large language models, like ChatGPT,

在大语言模型领域， 如 ChatGPT 上，

they are fast followers.

中国紧随我们之后。

In terms of military implementations,

就军事部署而言，

they're outspending us:

中国的支出超过了美国：

adjusted for total military budget,

从军事总预算来看，

China's spending ten times more than the United States.

中国的支出要超过美国十倍。

Why are we so far behind?

为什么我们会落后这么多？

The answer is twofold.

答案有两方面。

First, data supremacy.

一方面，“数据至上”。

Despite having the largest fleet of military hardware in the world,

虽然美国有世界上 最庞大的军事硬件储备，

most of the data from this fleet is thrown away or inaccessible,

这些硬件输出的大部分数据 都会被丢弃或无法访问，

hidden away on hard drives that never see the light of day.

藏于硬盘，不见天日。

This is our Achilles heel.

这就是我们的阿喀琉斯之踵。

In an AI war, everything boils down to data.

在 AI 战争中， 一切的精髓就是数据。

For defense AI,

就防御型 AI 而言，

data from the internet is not enough.

来自互联网的数据是不够的。

Most of the data needs to come from our military assets,

大多数数据得源于我们的军用设备、

sensors and collaborations with tech companies.

传感器和与科技公司的合作。

Military commanders need to know how to use data as a military asset.

军事指挥官必须知晓 如何将数据当作军事资产使用。

I've heard this first-hand many times,

我好几次直接听人说到这个观点，

from my conversations with military personnel,

来自我和军事人员的对话，

including most recently

包括最近一次，

from Lieutenant General Richard R. Coffman,

与美国陆军未来司令部副司令， 理查德·R·考夫曼

deputy commanding general for United States Army Futures Command.

（Richard R. Coffman）中将的对话。

Second, despite being home to the leading technology companies

第二，虽然美国是站在人工智能前线的

at the forefront of artificial intelligence,

主导科技公司的大本营，

the US tech industry has largely shied away

美国的科技行业大多会选择

from taking on government contracts.

避免承接政府合约。

Somewhere along the line,

久而久之，

tech leaders decided that working with the government was taboo.

科技领导者们认为 与政府合作是个禁忌。

As a technologist,

作为一名技术专家，

I'm often asked how I'm bettering this world.

我总是会被人问到 我为这个世界做出了什么贡献。

This is how I'm improving the future of our world:

我是这么改善世界的未来的：

by helping my country succeed

通过帮助我国走向成功，

and providing the best tools and technology

提供最好的工具和技术，

to ensure that the United States government can defend its citizens,

让美国政府有能力保卫国民、

allies and partners.

同盟和合作伙伴。

(Applause)

（掌声）

The Ukraine war has demonstrated that the nature of war has changed.

俄乌战争证明了 战争的本质已经发生了变化。

Through AI overmatch, Ukraine is challenging an adversary

在这场 AI 实力悬殊的战争中， 乌克兰挑战的对手

with far superior numbers of troops and weapons.

有着远超己方的军队规模和武器。

Before the Ukraine war,

在俄乌战争前，

Russia had spent an estimated 65 billion US dollars

俄罗斯在军费支出上 花费了大约 650 亿美元，

on its military expenditures,

whereas Ukraine only spent about six billion dollars.

而乌克兰只支出了 大约 60 亿美元。

It's estimated that Russia had over 900,000 military troops

据估计，俄罗斯拥有 超过 90 万名军人、

and 1,300 aircraft,

1300 架飞机，

whereas Ukraine only had 200,000 military troops and 130 aircraft.

而乌克兰只有 20 万名军人、130 架飞机。

Technologies such as drones,

像无人机、

AI-based targeting and image intelligence

基于 AI 的目标定位和图像情报、

and Javelin missiles

标枪反坦克导弹这样的技术，

have enabled a shocking defense of Ukraine.

让乌克兰具备了惊人的防御能力。

AI is proving invaluable for defending Ukrainian cities and infrastructure

AI 在保卫乌克兰城市和基础设施

against missile and drone bombardment.

免受导弹和无人机轰炸上功不可没。

At Scale, we’re using our novel machine learning algorithm

我们在 Scale 公司 通过新型机器学习算法

for battle damage assessment in key areas affected by the war.

对战争造成影响的关键领域 进行战争损害评估。

We've rapidly analyzed over 2,000 square kilometers

我们快速分析了 超过 2000 平方公里，

and have identified over 370,000 structures,

识别出了超过 37 万座建筑物，

including thousands not previously identified

包括上千座之前没有 被其他开源数据集识别的建筑物。

by other open source data sets.

We focused on Kyiv, Kharkiv and Dnipro

我们关注基辅、哈尔科夫和第聂伯罗，

and provided our data directly

以公开、可获取数据集的形式

in a publicly accessible data set to the broader AI community.

向受众更广的 AI 社区 直接提供数据。

One of the key problems we're solving

我们要解决的一个关键问题是

is using AI to analyze massive amounts of imagery and detect objects

利用 AI 分析大量图像， 识别物体，

because humans just can't keep up.

因为人类根本没有这个能力。

We've received an overwhelming response from our free AI-ready data set

我们免费的 AI 可用数据集 收到了热烈的反馈，

and have provided it directly to the United States and NATO allies.

并将其直接提供给了 美国及北约盟国。

And it's been downloaded over 2,000 times by AI companies, researchers,

下载量已超过 2000 次， 来自 AI 公司、研究者、

developers and GIS practitioners.

开发人员和地理信息系统 （GIS）从业者。

AI can also be used for change detection.

AI 也可以用于变化检测。

Simply put, algorithms can constantly monitor imagery

简而言之，算法可以持续监测影像，

and notify a human to investigate further if there's a change or a movement.

如果发生了变化或移动， 就会通知人类进一步调查。

It's clear that AI is increasingly powering warfare.

很显然，AI 越来越能 为战争添砖加瓦。

And based on the rate of progress in the AI field,

根据 AI 领域发展的速度，

I predict that in ten years, it will be the dominant force.

我预计它会在十年内成为中坚力量。

Disinformation and misinformation are already huge problems.

虚假信息和误传误报 已经是个大问题了。

And this technology is only going to make it worse.

这项技术只会加重这个问题。

Tools like ChatGPT have enabled AI to generate imagery,

像 ChatGPT 这样的工具 已经可以让 AI 生成图像、

text, audio, video, code and even reason.

文本、音频、视频、 代码，甚至是推理。

These tools can generate realistic-looking and realistic-sounding content,

这些工具可以生成似是而非、 以假乱真的内容，

which on top of bot-run social media accounts

再加上由机器人运营的 社交媒体账号，

will make it nearly impossible to identify disinformation

让人几乎不可能识别出 线上的虚假信息和以讹传讹。

and misinformation online.

Bad actors can use these tools to supercharge misinformation

图谋不轨的人利用这些工具 让虚假信息变本加厉，

and propagate falsehoods.

大肆传播谣言。

China already uses disinformation campaigns

中国在台湾进行了大量 的虚假信息宣传

and social media manipulations heavily in Taiwan,

和社交媒体操纵，

particularly during elections.

尤其是在大选期间。 （注：为原文立场，不代表译者意见）

Or take Russia's propaganda machine,

或者举俄罗斯的宣传机器为例，

which in the wake of Russia's invasion of Ukraine

在俄罗斯入侵乌克兰之后，

created a deepfake of Ukrainian President Zelensky

制作了乌克兰总统 泽连斯基（Zelensky）的深度伪造，

calling for Ukrainian troops to surrender.

呼吁乌军投降。

This deepfake was easy to spot,

很容易看出这个深度伪造，

but the next one may not be.

但下一个就未必了。

This also takes place within our borders,

这也发生在我国国内，

from social media algorithm manipulation

包括社交媒体算法操纵、

to advertising microtargeting and geofencing,

广告精准投放、地理围栏、

to deepfakes of politicians and bot-run social media accounts.

政治人物的深度伪造和 由机器人运营的社交媒体账号。

The United States is not excused

美国也不能幸免于持续恶化的 虚假信息和误传误报。

from exacerbating disinformation and misinformation.

These tools are universally accessible at low or no cost,

这些工具可以被大众 以低价或免费轻易获取，

meaning they can be employed by anyone anywhere

也就是说他们可以 在任意地点被任意人士雇佣

to undermine the sanctity of democracy globally.

在全球范围内破坏 民主的神圣不可侵犯性。

However, all hope is not lost.

但是，还有希望。

If we properly invest

如果我们适当地向 数据基础设施和数据准备投资，

into data infrastructure and data preparation,

all this can be avoided.

这些都可以被避免。

Deterrence is nothing new to military thinking.

威慑在军事思想中 不是什么新鲜事。

As we saw in World War II with the atomic bomb,

就像第二次世界大战中的原子弹，

it was a primary factor in deterring foreign adversaries

它成为了 60 多年以来 威慑外国敌对势力

from going to nuclear war for more than six decades.

避免进入核战争的主要因素。

Because the stakes of going to war with such a technology

用这项技术打仗的代价

were simply too high.

太高了。

We're likely to see a new calculus emerge with AI.

我们估计会看到 随 AI 而来的新算式。

It's uncharted territory, nobody knows what it will look like

这是一个无人之境， 没有人知道会是什么样的

or the toll it will take.

或会付出什么代价。

How do we know if our AI is better than our adversaries'?

我们该如何得知我们的 AI 会不会比敌对势力更厉害呢？

We won't.

我们无法得知。

But one thing is clear:

但有一点很明确：

AI can only be as powerful as the underlying data

AI 的厉害之处来自

that is used to fuel its algorithms.

它输入算法的底层数据。

Data will be a new kind of ammunition in the era of AI warfare.

数据会成为 AI 战时代的新型弹药。

In the tech industry, we often talk about missions.

身处科技行业的我们 总是会谈到“使命”。

They're often frivolous.

通常都是随便说说的。

Do they really change the world or save lives?

我们真的会改变世界或拯救生命吗？

This mission, on the other hand, really matters.

但是这个使命真的很重要。

The AI war will define the future of our world

AI 战争会定义世界的未来，

and has the potential to shift the balance of diplomatic power.

有可能会改变外交能力的平衡。

It's clear that digital warfare

很显然数字战争

is not some dystopian reality, tucked away in a faraway future.

不是什么反乌托邦的事实， 远在天边。

It is taking place in the here and now.

它正在此时此刻此地上演。

We cannot sit by the sidelines

我们不能袖手旁观，

and watch the rise of an authoritarian regime.

眼看着威权政体逐渐壮大。

It is in moments like this

在这样的时刻，

that technologists can either rise to the challenge or stand idle.

技术专家要么挺身而出迎接挑战， 要么袖手旁观。

I encourage my fellow technologists

我号召科技同僚们，

to understand the austerity and severity of our times

理解眼前的艰难困苦，

and commit themselves to supporting national security.

投身于支持国家安全。

While I find it shocking that most American AI companies

虽然我很意外大多数的 美国 AI 公司

have chosen not to support national security,

都会选择不去支持国家安全，

I do hope others join us.

我还是希望其他人可以加入我们。

We must fight for the world we want to live in.

我们必须争取 我们愿意身处其中的世界。

It's never mattered more.

此诚危急存亡之秋也。

Thank you.

谢谢。

(Applause)

（掌声）