How smarter AI will change creativity

更智能的人工智能將如何改變創(chuàng)造力

Foundation models

基礎(chǔ)模型(大模型)

The promise and perils of a breakthrough in machine intelligence

機(jī)器智能突破的前景和風(fēng)險

Picture a computer that could finish your sentences, using a better turn of phrase; or use a snatch of melody to compose music that sounds as if you wrote it (though you never would have); or solve a problem by creating hundreds of lines of computer code—leaving you to focus on something even harder. In a sense, that computer is merely the descendant of the power looms and steam engines that hastened the Industrial Revolution. But it also belongs to a new class of machine, because it grasps the symbols in language, music and programming and uses them in ways that seem creative. A bit like a human.

【1】a turn of phrase 措詞

想象一下，有這樣一臺電腦，它可以用更好的措辭完成你的句子；或者可以用一段旋律來作曲，就好像是你寫的一樣(雖然你從來沒有寫過)；或者可以通過寫數(shù)百行代碼來解決一個問題，從而讓你專注于更困難的事情。從某種意義上說，計算機(jī)只是加速工業(yè)革命的動力織機(jī)和蒸汽機(jī)的后代。但它也屬于一類新的機(jī)器，因?yàn)樗莆照Z言、音樂和編程中的符號，并能以看似有創(chuàng)意的方式使用它們。有點(diǎn)像人類。

The “foundation models” that can do these things represent a breakthrough in artificial intelligence, or AI. They, too, promise a Revolution, but this one will affect the high-status brainwork that the Industrial Revolution never touched. There are no guarantees about what lies ahead—after all, AI has stumbled in the past. But it is time to look at the promise and perils of the next big thing in machine intelligence.

能夠完成這些事情的“基礎(chǔ)模型”代表著人工智能的突破。它們也承諾會帶來一場革命，但這場革命將影響工業(yè)革命從未觸及的高級腦力勞動。未來的情況無法保證，畢竟，人工智能在過去也曾遭遇過挫折。但是時候看看機(jī)器智能領(lǐng)域下一個重大事件的前景和風(fēng)險了。

Foundation models are the latest twist on “deep learning” (dl), a technique that rose to prominence ten years ago and now dominates the field of AI. Loosely based on the networked structure of neurons in the human brain, dl systems are “trained” using millions or billions of examples of texts, images or sound clips. In recent years the ballooning cost, in time and money, of training ever-larger dl systems had prompted worries that the technique was reaching its limits. Some fretted about an “AI winter”. But foundation models show that building ever-larger and more complex dl does indeed continue to unlock ever more impressive new capabilities. Nobody knows where the limit lies.

基礎(chǔ)模型是“深度學(xué)習(xí)的最新發(fā)展，該技術(shù)在10年前嶄露頭角，如今在人工智能領(lǐng)域占據(jù)主導(dǎo)地位。深度學(xué)習(xí)系統(tǒng)大致基于人類大腦神經(jīng)元的網(wǎng)絡(luò)結(jié)構(gòu)，可以通過使用數(shù)百萬或數(shù)十億個文本、圖像或聲音片段樣本進(jìn)行“訓(xùn)練”。近年來，訓(xùn)練規(guī)模越來越大的深度學(xué)習(xí)系統(tǒng)的時間和金錢成本不斷膨脹，這讓人們擔(dān)心這項(xiàng)技術(shù)正在達(dá)到極限。一些人擔(dān)心“人工智能的冬天”。但基礎(chǔ)模型顯示，構(gòu)建更大、更復(fù)雜的深度學(xué)習(xí)模型確實(shí)會繼續(xù)“解鎖”更重要的新功能。沒有人知道極限在哪里。

The resulting models are a new form of creative, non-human intelligence. The systems are sophisticated enough both to possess a grasp of language and also to break the rules coherently. A dog cannot laugh at a joke in the New Yorker, but an ai can explain why it is funny—a feat that is, frankly, sometimes beyond readers of the New Yorker. When we asked one of these models to create a collage using the title of this leader and nothing more, it came up with the cover art for our American and Asian editions, pictured (we tried to distract our anxious human designers with a different cover in our European editions).

由此產(chǎn)生的模型是一種創(chuàng)造性的、非人類智能的新形式。這些系統(tǒng)足夠復(fù)雜，既能掌握語言，又能連貫一致地打破規(guī)則。小狗不會因?yàn)椤都~約客》上的笑話而發(fā)笑，但人工智能可以解釋為什么它有趣，坦率地說，這一“壯舉”有時超出了《紐約客》的讀者的能力。當(dāng)我們讓其中一個模型用這篇文章的標(biāo)題創(chuàng)作一幅拼貼畫時，它創(chuàng)造了我們美國版和亞洲版的封面圖，如圖所示(我們試圖在歐洲版本，使用不同封面來分散我們焦慮的人類設(shè)計師的注意力)。

Foundation models have some surprising and useful properties. The eeriest of these is their “emergent” behaviour—that is, skills (such as the ability to get a joke or match a situation and a proverb) which arise from the size and depth of the models, rather than being the result of deliberate design. Just as a rapid succession of still photographs gives the sensation of movement, so trillions of binary computational decisions fuse into a simulacrum of fluid human comprehension and creativity that, whatever the philosophers may say, looks a lot like the real thing. Even the creators of these systems are surprised at their power.

【1】eerie 怪異恐怖的

基礎(chǔ)模型有一些令人驚訝和有用的特性。其中最可怕的是它們的“突發(fā)”行為——也就是說，技能(比如理解笑話或匹配情景和諺語的能力)是通過大規(guī)模和深層模型獲得的，而不是經(jīng)過深思熟慮設(shè)計的結(jié)果。就像快速連續(xù)的靜態(tài)照片給人一種“移動”的感覺一樣，數(shù)萬億的二元計算決策融合成一個有著易變的人類理解和創(chuàng)造力的模擬物，不管哲學(xué)家怎么說，這看起來很像真實(shí)的東西。即使是這些系統(tǒng)的創(chuàng)造者也對它們的威力感到驚訝。

This intelligence is broad and adaptable. True, foundation models are capable of behaving like an idiot, but then humans are, too. If you ask one who won the Nobel prize for physics in 1625, it may suggest Galileo, Bacon or Kepler, not understanding that the first prize was awarded in 1901. However, they are also adaptable in ways that earlier ais were not, perhaps because at some level there is a similarity between the rules for manipulating symbols in disciplines as different as drawing, creative writing and computer programming. This breadth means that foundation models could be used in lots of applications, from helping find new drugs using predictions about how proteins fold in three dimensions, to selecting interesting charts from datasets and dealing with open-ended questions by trawling huge databases to formulate answers that open up new areas of inquiry.

這種智能具有廣泛性和適應(yīng)性。確實(shí)，基礎(chǔ)模型有可能表現(xiàn)得像傻瓜一樣，但人類也一樣。如果你問誰獲得了1625年的諾貝爾物理學(xué)獎，答案可能是伽利略、培根或開普勒，他們不知道諾貝爾物理學(xué)獎是1901年頒發(fā)的。然而，它們也具有早期人工智能系統(tǒng)所不具備的適應(yīng)性，也許是因?yàn)樵谀撤N程度上，繪畫、創(chuàng)意寫作和計算機(jī)編程等不同學(xué)科的符號操作規(guī)則之間存在相似之處。這種廣度意味著基礎(chǔ)模型可以應(yīng)用于很多領(lǐng)域，從通過預(yù)測蛋白質(zhì)在三維空間中的折疊方式以發(fā)現(xiàn)新藥，到從數(shù)據(jù)集中選擇有趣的圖表，以及通過搜索龐大的數(shù)據(jù)庫來制定答案，以處理開放式問題，這開辟了新查詢領(lǐng)域。

That is exciting, and promises to bring great benefits, most of which still have to be imagined. But it also stirs up worries. Inevitably, people fear that ais creative enough to surprise their creators could become malign. In fact, foundation models are light-years from the sentient killer-robots beloved by Hollywood. Terminators tend to be focused, obsessive and blind to the broader consequences of their actions. Foundational ai, by contrast, is fuzzy. Similarly, people are anxious about the prodigious amounts of power training these models consume and the emissions they produce. However, ais are becoming more efficient, and their insights may well be essential in developing the technology that accelerates a shift to renewable energy.

這是令人興奮的，并有望帶來巨大的好處，其中大部分好處仍有待想象。但它也引發(fā)了擔(dān)憂。人們不可避免地?fù)?dān)心，人工智能系統(tǒng)的創(chuàng)造力足以讓其創(chuàng)造者感到驚訝，這可能會變得有害。事實(shí)上，基礎(chǔ)模型與好萊塢鐘愛的有感知能力的殺人機(jī)器人相差極大。終結(jié)者傾向于專注、偏執(zhí)，對自己行為的更廣泛后果視而不見。相比之下，基礎(chǔ)人工智能是模糊的。同樣，人們對訓(xùn)練這些模型所消耗的巨大電力及其排放感到焦慮。然而，人工智能系統(tǒng)正變得越來越高效，它們的見解對于開發(fā)加速向可再生能源轉(zhuǎn)變的技術(shù)方面很可能至關(guān)重要。

A more penetrating worry is over who controls foundation models. Training a really large system such as Google’s palm costs more than $10m a go and requires access to huge amounts of data—the more computing power and the more data the better. This raises the spectre of a technology concentrated in the hands of a small number of tech companies or governments.

一個更尖銳的擔(dān)憂是關(guān)于誰控制著基礎(chǔ)模型。訓(xùn)練一個像谷歌的PaLM這樣的大型系統(tǒng)一次花費(fèi)超過1000萬美元，并且需要訪問大量的數(shù)據(jù)——計算能力越強(qiáng)，數(shù)據(jù)越多越好。這引發(fā)了技術(shù)集中在少數(shù)科技公司或政府手中的擔(dān)憂。

If so, the training data could further entrench the world’s biases—and in a particularly stifling and unpleasant way. Would you trust a ten-year-old whose entire sense of reality had been formed by surfing the internet? Might X- and American-trained ais be recruited to an ideological struggle to bend minds? What will happen to cultures that are poorly represented online?

如果是這樣的話，訓(xùn)練數(shù)據(jù)可能會以一種特別令人窒息和不愉快的方式進(jìn)一步加深世界的偏見。你會相信一個十歲的孩子嗎? 他對現(xiàn)實(shí)的全部感知都是通過網(wǎng)上沖浪形成的。X和美國訓(xùn)練的人工智能系統(tǒng)會不會被吸納到一場意識形態(tài)斗爭中來“扭曲”人們的思想？對于那些在網(wǎng)絡(luò)上表現(xiàn)不佳的文化，會發(fā)生什么？

And then there is the question of access. For the moment, the biggest models are restricted, to prevent them from being used for nefarious purposes such as generating fake news stories. Openai, a startup, has designed its model, called dall-e 2, in an attempt to stop it producing violent or pornographic images. Firms are right to fear abuse, but the more powerful these models are, the more limiting access to them creates a new elite. Self-regulation is unlikely to resolve the dilemma.

然后是訪問的問題。目前，最大的模型應(yīng)用受限，以防止它們被用于邪惡的目的，如制造假新聞。OpenAI是一家創(chuàng)業(yè)公司，它設(shè)計了一種名為DALL- E 2的模型，試圖阻止它產(chǎn)生暴力或色情圖片。公司擔(dān)心模型濫用是對的，但這些模型越強(qiáng)大，對它們的使用以創(chuàng)造新的“精英”就越受限。自我監(jiān)管不太可能解決這一困境。

Bring on the RevolutionFor years it has been said that ai-powered automation poses a threat to people in repetitive, routine jobs, and that artists, writers and programmers were safer. Foundation models challenge that assumption. But they also show how ai can be used as a software sidekick to enhance productivity. This machine intelligence does not resemble the human kind, but offers something entirely different. Handled well, it is more likely to complement humanity than usurp it.?

多年來，有人說人工智能驅(qū)動的自動化對從事重復(fù)性、例行工作的人構(gòu)成了威脅，對藝術(shù)家、作家和程序員更安全?；A(chǔ)模型對這種假設(shè)提出了挑戰(zhàn)。但它們也展示了人工智能如何作為軟件助手來提高生產(chǎn)力。這種機(jī)器智能與人類不同，但提供了一些完全不同的東西。如果處理得好，它更有可能與人類“互補(bǔ)”，而不是“取代”人類。