Name: 人工智能很危險，但不是你想的那樣 | Sasha Luccioni | TED (AI Is Dangerous, but Not for the Reasons You Think | Sasha Luccioni | TED)
Uploaded: 2025-09-17T14:37:28.000Z
Duration: 10 min 19 s

I've been an AI researcher for over a decade, and a couple of months ago, I got the weirdest email of my career.

我從事人工智能研究已有十多年，幾個月前，我收到了職業生涯中最奇怪的一封電子郵件。

A random stranger wrote to me saying that my work in AI is going to end humanity.

一個陌生人寫信給我，說我在人工智能領域的工作將終結人類。

程度副詞

Now, I get it, AI, it's so hot right now.

現在，我明白了，人工智能，現在太火了。

It's in the headlines pretty much every day, sometimes because of really cool things like discovering new molecules for medicine, or that dope Pope in the white puffer coat.

它幾乎每天都上頭條，有時是因為一些很酷的事情，比如發現了新的藥物分子，或者是那個穿著白色羽絨服的笨蛋教皇。

But other times, the headlines have been really dark, like that chatbot telling that guy that he should divorce his wife, or that AI meal planner app proposing a crowd-pleasing recipe featuring chlorine gas.

但在其他時候，頭條新聞卻非常陰暗，比如那個哈拉機器人告訴那個人他應該和妻子離婚，或者那個人工智能膳食計劃應用程序提出了一個含有氯氣的大眾喜聞樂見的食譜。

And in the background, we've heard a lot of talk about doomsday scenarios, existential risk, and the singularity with letters being written and events being organized to make sure that doesn't happen.

在這一背景下，我們聽到了許多關於末日場景、生存風險和奇點的討論，有人寫信，有人組織活動，以確保這種情況不會發生。

Now, I'm a researcher who studies AI's impact on society.

現在，我是一名研究員，研究人工智能對社會的影響。

And I don't know what's going to happen in 10 or 20 years, and nobody really does.

我不知道 10 年或 20 年後會發生什麼，沒人知道。

But what I do know is that there's some pretty nasty things going on right now.

但我知道的是，現在正在發生一些非常糟糕的事情。

因為人工智能並非存在於真空中。

It is part of society, and it has impacts on people and the planet.

它是社會的一部分，對人類和地球都有影響。

AI models can contribute to climate change.

人工智能模型可能導致氣候變化。

Their training data uses art and books created by artists and authors without their consent.

他們的訓練數據未經藝術家和作家同意，就使用了他們創作的藝術作品和書籍。

And its deployment can discriminate against entire communities.

而且，它的部署可能會歧視整個社區。

But we need to start tracking its impacts.

We need to start being transparent and disclosing them, and creating tools so that people understand AI better.

我們需要開始透明地披露這些資訊，並創建工具，讓人們更好地瞭解人工智能。

So that hopefully, future generations of AI model are going to be more trustworthy, sustainable, maybe less likely to kill us, if that's what you're into.

是以，希望未來幾代人工智能模型更值得信賴、更可持續發展，也許不太可能殺死我們，如果你喜歡的話。

那麼，讓我們從可持續發展開始。

Because that cloud that AI models live on is actually made out of metal, plastic, and powered by vast amounts of energy.

因為人工智能模型賴以生存的雲實際上是由金屬和塑膠製成的，並由大量能源提供動力。

And each time you query an AI model, it comes with a cost to the planet.

而每次查詢人工智能模型，都會讓地球付出代價。

Last year, I was part of the Big Science Initiative, which brought together 1,000 researchers from all over the world to create Bloom, the first open large language model, like JATJBT, but with an emphasis on ethics, transparency, and consent.

去年，我參加了 "大科學計劃"（Big Science Initiative），該計劃彙集了來自世界各地的 1000 名研究人員，共同創建了布魯姆（Bloom），這是首個開放式大型語言模型，與 JATJBT 類似，但強調倫理、透明度和同意。

And the study I led that looked at Bloom's environmental impacts found that just training it used as much energy as 30 homes in a whole year, and emitted 25 tons of carbon dioxide, which is like driving your car five times around the planet, just so somebody can use this model to tell a knock-knock joke.

我上司的一項研究調查了布盧姆對環境的影響，結果發現，光是訓練布盧姆一年所消耗的能源就相當於 30 個家庭一年所消耗的能源，同時還排放了 25 噸二氧化碳，這就相當於開著汽車繞地球五圈。

And this might not seem like a lot, but other similar large language models, like GPT-3, emit 20 times more carbon.

這看起來似乎並不多，但其他類似的大型語言模型，如 GPT-3 排放的碳要多 20 倍。

But the thing is, tech companies aren't measuring this stuff.

但問題是，科技公司並沒有測量這些東西。

And so this is probably only the tip of the iceberg, even if it is a melting one.

是以，這可能只是冰山一角，儘管它正在融化。

And in recent years, we've seen AI models balloon in size.

近年來，我們看到人工智能模型的規模不斷擴大。

Because the current trend in AI is, bigger is better.

因為目前人工智能的趨勢是，越大越好。

But please don't get me started on why that's the case.

但請不要讓我開始討論為什麼會這樣。

In any case, we've seen large language models in particular grow 2,000 times in size over the last five years.

無論如何，在過去五年中，我們看到大型語言模型的規模增長了 2000 倍。

And of course, their environmental costs are rising as well.

當然，它們的環境成本也在上升。

The most recent work I led found that switching out a smaller, more efficient model for a larger language model emits 14 times more carbon for the same task, like telling that knock-knock joke.

我最近上司的一項工作發現，在完成同樣任務的情況下，將一個更小更高效的模型換成一個更大的語言模型所排放的碳要多 14 倍，就像講那個敲敲打打的笑話一樣。

And as we're putting in these models into cell phones and search engines and smart fridges and speakers, the environmental costs are really piling up quickly.

隨著我們在手機、搜索引擎、智能冰箱和揚聲器中使用這些模型，環境成本確實在迅速增加。

So instead of focusing on some future existential risks, let's talk about current tangible impacts and tools we can create to measure and mitigate these impacts.

是以，與其關注一些未來的生存風險，不如讓我們來談談當前的實際影響，以及我們可以用來衡量和減輕這些影響的工具。

I helped create CodeCarbon, a tool that runs in parallel to AI training code that estimates the amount of energy it consumes and the amount of carbon it emits.

我幫助創建了 CodeCarbon，這是一個與人工智能訓練代碼並行運行的工具，可以估算其消耗的能源和排放的碳。

And using a tool like this can help us make informed choices, like choosing one model over the other because it's more sustainable, or deploying AI models on renewable energy, which can drastically reduce their emissions.

使用這樣的工具可以幫助我們做出明智的選擇，比如選擇一種模式而不是另一種模式，因為它更可持續，或者在可再生能源上部署人工智能模式，這樣可以大大減少排放。

But let's talk about other things, because there's other impacts of AI apart from sustainability.

但我們還是要談談其他方面，因為除了可持續發展，人工智能還有其他影響。

For example, it's been really hard for artists and authors to prove that their life's work has been used for training AI models without their consent.

例如，藝術家和作家很難證明他們的畢生心血在未經本人同意的情況下被用於訓練人工智能模型。

And if you want to sue someone, you tend to need proof, right?

如果你想起訴某人，往往需要證據，對嗎？

So Spawning AI, an organization that was founded by artists, created this really cool tool called Have I Been Trained?

Spawning AI 是一個由藝術家創立的組織，它創建了一個很酷的工具，名為 "我被訓練過嗎？

And it lets you search these massive data sets to see what they have on you.

它還能讓你搜索這些海量數據集，查看它們有哪些關於你的資訊。

I searched Lion 5B, which is this huge data set of images and text, to see if any images of me were in there.

我搜索了 Lion 5B，這是一個包含影像和文本的龐大數據集，看看裡面有沒有我的影像。

Now those two first images, that's me from events I've spoken at.

前面兩張圖片是我在活動中的發言。

But the rest of the images, none of those are me.

They're probably of other women named Sasha who put photographs of themselves up on the internet.

可能是其他叫薩莎的女人把自己的照片放到了網上。

And this can probably explain why when I query an image generation model to generate a photograph of a woman named Sasha, more often than not I get images of bikini models.

這或許也解釋了為什麼當我查詢影像生成模型以生成一張名叫薩莎的女性照片時，得到的往往是比基尼模特的影像。

Sometimes they have two arms, sometimes they have three arms, but they rarely have any clothes on.

他們有時有兩隻胳膊，有時有三隻胳膊，但很少穿衣服。

And why it can be interesting for people like you and me to search these data sets.

為什麼對你我這樣的人來說，搜索這些數據集會很有趣？

For artists like Carla Ortiz, this provides crucial evidence that her life's work, her And she and two artists used this as evidence to file a class action lawsuit against AI companies for copyright infringement.

對於卡拉-奧爾蒂斯（Carla Ortiz）這樣的藝術家來說，這為她的畢生心血、她的作品提供了重要證據，她和兩位藝術家以此為證據，對人工智能公司提起了集體訴訟，指控其侵犯版權。

And most recently Spawning AI partnered up with Hugging Face, the company where I work at, to create opt-in and opt-out mechanisms for creating these data sets.

最近，Spawning AI 與我所在的公司 Hugging Face 合作，為創建這些數據集建立了選擇加入和選擇退出機制。

Because artwork created by humans shouldn't be an all-you-can-eat buffet for training AI language models.

因為人類創作的藝術作品不應該成為訓練人工智能語言模型的自助餐。

The very last thing I want to talk about is bias.

Formally speaking, it's when AI models encode patterns and beliefs that can represent stereotypes or racism and sexism.

從形式上講，這是指人工智能模型編碼的模式和信念可能代表刻板印象或種族主義和性別歧視。

One of my heroes, Dr. Joy Bolognini, experienced this firsthand when she realized that AI systems wouldn't even detect her face unless she was wearing a white-colored mask.

我的偶像之一喬伊-博洛尼尼博士（Dr. Joy Bolognini）就親身經歷過這種情況，她意識到除非戴上白色面具，否則人工智能系統根本無法檢測到她的臉。

Digging deeper, she found that common facial recognition systems were vastly worse for women of color compared to white men.

深入研究後，她發現普通的面部識別系統對有色人種女性的識別效果遠不如白人男性。

And when models like this, biased models like this, are deployed in law enforcement settings, this can result in false accusations, even wrongful imprisonment, which we've seen happen to multiple people in recent months.

而當這樣的模式，這樣帶有偏見的模式被運用到執法環境中時，可能會導致錯誤的指控，甚至是錯誤的監禁，而我們在最近幾個月已經看到多人遭遇了這種情況。

For example, Portia Woodruff was wrongfully accused of carjacking at eight months pregnant because an AI system wrongfully identified her.

例如，波西婭-伍德拉夫（Portia Woodruff）在懷孕八個月時被錯誤地指控劫車，原因是人工智能系統錯誤地識別了她的身份。

But sadly, these systems are black boxes, and even their creators can't say exactly why they work the way they do.

但遺憾的是，這些系統都是黑盒子，就連它們的創造者也說不清楚它們為什麼會以這種方式工作。

And for example, for image generation systems, if you're used in contexts like generating a forensic sketch based on a description of a perpetrator, they take all those biases and they spit them back out for terms like dangerous criminal, terrorist or gang member, which of course is super dangerous when these tools are deployed in society.

例如，對於影像生成系統來說，如果根據犯罪者的描述來生成法醫素描，那麼系統就會將所有這些偏差轉化為危險罪犯、恐怖分子或幫派成員等術語，而當這些工具在社會中使用時，這當然是非常危險的。

And so in order to understand these tools better, I created this tool called the Stable Bias Explorer, which lets you explore the bias of image generation models through the lens of professions.

是以，為了更好地瞭解這些工具，我創建了這個名為 "穩定偏差探索器 "的工具，它可以讓你通過職業的視角來探索影像生成模型的偏差。

So try to picture a scientist in your mind.

所以，請試著在腦海中想象一位科學家。

Men in glasses and lab coats, and none of them look like me.

戴眼鏡、穿白大褂的男人，沒有一個長得像我。

And the thing is, is that we looked at all these different image generation models and found a lot of the same thing.

問題是，我們研究了所有這些不同的影像生成模型，發現了很多相同之處。

Significant representation of whiteness and masculinity across all 150 professions that we looked at.

在我們考察的所有 150 個職業中，白人和男性的比例都很高。

Even if compared to the real world, the US Labor Bureau of Statistics, these models show lawyers as men and CEOs as men almost 100% of the time, even though we all know not all of them are white and male.

即使與現實世界，即美國勞工統計局相比，這些模型也顯示律師幾乎 100% 都是男性，首席執行官幾乎 100% 都是男性，儘管我們都知道他們並不都是白人和男性。

And sadly, my tool hasn't been used to write legislation yet.

遺憾的是，我的工具還沒有被用來制定法律。

But I recently presented it at a UN event about gender bias as an example of how we can make tools for people from all walks of life, even those who don't know how to code, to engage with and better understand AI, because we use professions, but you can use any terms that are of interest to you.

但我最近在聯合國的一次關於性別偏見的活動上介紹了它，以此為例說明我們如何為各行各業的人，甚至是那些不懂代碼的人提供工具，讓他們參與並更好地理解人工智能，因為我們使用的是專業，但你也可以使用任何你感興趣的術語。

And as these models are being deployed, are being woven into the very fabric of our societies, our cell phones, our social media feeds, even our justice systems and our economies have AI in them.

隨著這些模式的部署，它們正在融入我們的社會結構，我們的手機、我們的社交媒體，甚至我們的司法系統和我們的經濟都有了人工智能的身影。

And it's really important that AI stays accessible so that we know both how it works and when it doesn't work.

人工智能的可及性非常重要，這樣我們才能知道它是如何工作的，以及什麼時候不起作用。

And there's no single solution for really complex things like bias or copyright or climate change.

對於偏見、版權或氣候變化等真正複雜的問題，沒有單一的解決方案。

But by creating tools to measure AI's impact, we can start getting an idea of how bad they are and start addressing them as we go.

但是，通過創建工具來衡量人工智能的影響，我們可以開始瞭解它們有多糟糕，並開始著手解決它們。

Start creating guardrails to protect society and the planet.

開始建立保護社會和地球的護欄。

And once we have this information, companies can use it in order to say, OK, we're going to choose this model because it's more sustainable, this model because it respects copyright.

一旦我們掌握了這些資訊，公司就可以利用這些資訊說："好吧，我們要選擇這種模式，因為它更可持續；選擇這種模式，因為它尊重版權。

Legislators who really need information to write laws can use these tools to develop new regulation mechanisms or governance for AI as it gets deployed into society.

真正需要資訊來制定法律的立法者可以利用這些工具，在人工智能進入社會時為其制定新的監管機制或治理方法。

And users like you and me can use this information to choose AI models that we can trust not to misrepresent us and not to misuse our data.

而像你我這樣的用戶可以利用這些資訊來選擇人工智能模型，我們可以相信這些模型不會誤導我們，也不會濫用我們的數據。

But what did I reply to that email that said that my work is going to destroy humanity?

但我是怎麼回覆那封說我的作品會毀滅人類的郵件的呢？

I said that focusing on AI's future existential risks is a distraction from its current very tangible impacts and the work we should be doing right now or even yesterday for reducing these impacts.

我說，關注人工智能未來的生存風險會分散人們的注意力，使人們忽視人工智能目前帶來的切實影響，以及我們現在甚至昨天就應該為減少這些影響而開展的工作。

Because, yes, AI is moving quickly, but it's not a done deal.

因為，沒錯，人工智能正在快速發展，但它還沒有完成。

We're building the road as we walk it, and we can collectively decide what direction we want to go in together.

我們邊走邊建，共同決定前進的方向。

字幕列表影片播放

人工智能很危險，但不是你想的那樣 | Sasha Luccioni | TED (AI Is Dangerous, but Not for the Reasons You Think | Sasha Luccioni | TED)

evidence

initiative

impact

sustainable