Finance&Economics

February 2nd 2023|947words

How powerful new technologies transform economies

IT CAN TAKE a little imagination to see how some innovations might change an economy. Not so with the latest AI tools. It is easy—from a writer’s perspective, uncomfortably so—to think of contexts in which something like ChatGPT, a clever chatbot which has taken the web by storm since its release in November, could either dramatically boost a human worker’s productivity or replace them outright. The GPT in its name stands for “generative pre-trained transformer”, which is a particular kind of language model. It might well stand for general-purpose technology: an earth-shaking sort of innovation which stands to boost productivity across a wide-range of industries and occupations, in the manner of steam engines, electricity and computing. The economic revolutions powered by those earlier GPTs can give us some idea how powerful AI might transform economies in the years ahead.

ears ahead.

想要看到一些創(chuàng)新如何改變經(jīng)濟(jì)，可能需要一些想象力。但最新的人工智能工具卻并非如此。從作家的角度來看，想象一個(gè)像ChatGPT這樣的智能聊天機(jī)器人（自去年11月發(fā)布以來就席卷網(wǎng)絡(luò)）可以顯著提高人類工作者的工作效率，或完全取代人類工作者的情景，是很容易的，但這卻令人不安。其名稱中的GPT表示“生成式預(yù)訓(xùn)練轉(zhuǎn)換器”，這是一種特殊的語言模型。它很可能代表通用技術(shù)的：一種翻天覆地的創(chuàng)新，這將提高廣泛行業(yè)和職業(yè)的生產(chǎn)力，就像蒸汽機(jī)、電力和計(jì)算機(jī)一樣。由這些早期“GPTS”推動(dòng)的經(jīng)濟(jì)革命，可以讓我們了解未來幾年人工智能將如何強(qiáng)大地改變經(jīng)濟(jì)。

In a paper published in 1995, Timothy Bresnahan of Stanford University and Manuel Trajtenberg of Tel Aviv University set out what they saw as the characteristics of a general-purpose technology. It must be used in many industries, have an inherent potential for continued improvement and give rise to “innovational complementarities”—that is, induce knock-on innovation in the industries which use it. AI is being adopted widely, seems to get better by the day and is being deployed in ever more R&D contexts. So when does the economic revolution begin?

在1995年發(fā)表的一篇論文中，斯坦福大學(xué)的Timothy Bresnahan和特拉維夫大學(xué)的Manuel Trajtenberg，闡述了他們所認(rèn)為的通用技術(shù)的特征。這種技術(shù)必須在許多行業(yè)中使用，具有持續(xù)改進(jìn)的內(nèi)在潛力，并產(chǎn)生“創(chuàng)新互補(bǔ)性”——也就是說，在使用這種技術(shù)的行業(yè)中誘導(dǎo)連鎖創(chuàng)新。我們現(xiàn)在正廣泛采用人工智能，似乎一天比一天好，并把這種技術(shù)部署在越來越多的研發(fā)環(huán)境中。那么經(jīng)濟(jì)革命何時(shí)開始呢？

The first lesson from history is that even the most powerful new tech takes time to change an economy. James Watt patented his steam engine in 1769, but steam power did not overtake water as a source of industrial horsepower until the 1830s in Britain and 1860s in America. In Britain the contribution of steam to productivity growth peaked post-1850, nearly a century after Watt’s patent, according to Nicholas Crafts of the University of Sussex. In the case of electrification, the key technical advances had all been accomplished before 1880, but American productivity growth actually slowed from 1888 to 1907. Nearly three decades after the first silicon integrated circuits Robert Solow, a Nobel-prizewinning economist, was still observing that the computer age could be seen everywhere but in the productivity statistics. It was not until the mid-1990s that a computer-powered productivity boom eventually emerged in America.

歷史給我們的第一個(gè)經(jīng)驗(yàn)是，就算是最強(qiáng)大的新技術(shù)，也需要時(shí)間來改變經(jīng)濟(jì)。詹姆斯·瓦特在1769年為他的蒸汽機(jī)申請了專利，但是直到19世紀(jì)30年代的英國和19世紀(jì)60年代的美國，蒸汽動(dòng)力才取代了水成為工業(yè)動(dòng)力的來源。據(jù)蘇塞克斯大學(xué)的尼古拉斯·克拉夫特稱，在英國，蒸汽機(jī)對生產(chǎn)力增長的貢獻(xiàn)，在1850年后達(dá)到頂峰，也就是瓦特獲得專利的近一個(gè)世紀(jì)之后。以電氣化為例，關(guān)鍵的技術(shù)進(jìn)步都在1880年之前完成，但實(shí)際上在1888年到1907年，美國的生產(chǎn)率增長有所放緩。在第一個(gè)硅集成電路問世近30年后，諾貝爾經(jīng)濟(jì)學(xué)獎(jiǎng)得主羅伯特?索洛仍在觀察，計(jì)算機(jī)時(shí)代這些集成電路無處不在，但在生產(chǎn)率統(tǒng)計(jì)數(shù)據(jù)中卻看不到。直到20世紀(jì)90年代中期，計(jì)算機(jī)驅(qū)動(dòng)的生產(chǎn)力繁榮才最終在美國出現(xiàn)。

The gap between innovation and economic impact is in part because of fine-tuning. Early steam engines were wildly inefficient and consumed prohibitively expensive piles of coal. Similarly, the stunning performance of recent AI tools represents a big improvement over those which sparked a boomlet of AI enthusiasm roughly a decade ago. (Siri, Apple’s virtual assistant, was released in 2011, for example.) Capital constraints can also slow deployment. Robert Allen of New York University Abu Dhabi argues that the languid rise in productivity growth in industrialising Britain reflected a lack of capital to build plants and machines, which was gradually overcome as capitalists reinvested their fat profits.

創(chuàng)新和經(jīng)濟(jì)影響之間的差距部分是由于微調(diào)。早期的蒸汽機(jī)效率極低，并且消耗大量昂貴的煤。同樣，最近人工智能工具的驚人表現(xiàn)，與大約10年前引發(fā)人工智能熱潮的工具相比，已經(jīng)有了很大的進(jìn)步。（比如，蘋果的虛擬助手Siri于2011年發(fā)布）。資本的限制也會減緩技術(shù)部署速度。紐約大學(xué)阿布扎比分校的羅伯特·艾倫認(rèn)為，英國工業(yè)化過程中生產(chǎn)率增長的緩慢，反映了建造工廠和機(jī)器所需資本的缺乏，而這一問題隨著資本家將他們豐厚的利潤進(jìn)行再投資而逐漸被克服。

More recent work emphasises the time required to accumulate what is known as intangible capital, or the basic know-how needed to make effective use of new tech. Indeed, Erik Brynjolfsson of Stanford University, Daniel Rock of the Massachusetts Institute of Technology and Chad Syverson of the University of Chicago suggest a disruptive new technology may be associated with a “productivity J-curve”. Measured productivity growth may actually decline in the years or decades after a new technology appears, as firms and workers divert time and resources to studying the tech and designing business processes around it. Only later as these investments bear fruit does the J surge upward. The authors reckon that AI-related investments in intangible capital may already be depressing productivity growth, albeit not yet by very much.

最近的研究強(qiáng)調(diào)了，積累所謂的無形資本或有效利用新技術(shù)所需的基本知識所需的時(shí)間。的確，斯坦福大學(xué)的Erik Brynjolfsson、麻省理工學(xué)院的Daniel Rock和芝加哥大學(xué)的Chad Syverson認(rèn)為，顛覆性的新技術(shù)可能與“生產(chǎn)力J形曲線”有關(guān)。在一項(xiàng)新技術(shù)出現(xiàn)后的幾年或幾十年里，由于企業(yè)和員工將時(shí)間和資源用于研究這項(xiàng)技術(shù)，并圍繞該新技術(shù)設(shè)計(jì)業(yè)務(wù)流程，衡量的生產(chǎn)率增長實(shí)際上可能會下降。只有在這些投資取得成果之后，J曲線上才會大幅上升。作者認(rèn)為，與人工智能相關(guān)的無形資本投資可能已經(jīng)在抑制生產(chǎn)率增長，盡管還不是很大。

Of course for many people, questions about the effects of AI on growth take a back seat to concerns about consequences for workers. Here, history’s messages are mixed. There is good news: despite epochal technological and economic change, fears of mass technological unemployment have never before been realised. Tech can and does take a toll on individual occupations, however, in ways that can prove socially disruptive. Early in the Industrial Revolution, mechanisation dramatically increased demand for relatively unskilled workers, but crushed the earnings of craftsmen who had done much of the work before, which is why some chose to join machine-smashing Luddite movements. And in the 1980s and 1990s, automation of routine work on factory floors and in offices displaced many workers of modest means, while boosting employment for both high- and low-skilled workers.

當(dāng)然，對許多人來說，人工智能對經(jīng)濟(jì)增長的影響問題，被置于對工人影響的擔(dān)憂之后。在這里，歷史給我們的信息是好壞參半的。好消息是：盡管發(fā)生了劃時(shí)代的技術(shù)和經(jīng)濟(jì)變革，但對大規(guī)模技術(shù)性失業(yè)的擔(dān)憂從未成為現(xiàn)實(shí)。然而，科技可以也確實(shí)會對個(gè)人職業(yè)造成影響，其影響方式可能會對社會造成破壞。在工業(yè)革命早期，機(jī)械化極大地增加了對相對無技能工人的需求，但減少了以前做過大部分工作的工匠的收入，這就是為什么一些人選擇加入粉碎機(jī)器的勒德運(yùn)動(dòng)。在20世紀(jì)80年代和90年代，工廠和辦公室日常工作的自動(dòng)化，取代了許多中等收入的工人，同時(shí)促進(jìn)了高技能和低技能工人的就業(yè)。

Gee, Pretty Terrific

AI might well augment the productivity of workers of all different skill levels, even writers. Yet what that means for an occupation as a whole depends on whether improved productivity and lower costs lead to a big jump in demand or only a minor one. When the assembly line—a process innovation with GPT-like characteristics—allowed Henry Ford to cut the cost of making cars, demand surged and workers benefited. If AI boosts productivity and lowers costs in medicine, for example, that might lead to much higher demand for medical services and professionals.

人工智能很可能提高所有不同技能水平的工人的生產(chǎn)力，甚至是作家。然而，這對一個(gè)職業(yè)整體來說意味著什么，這取決于生產(chǎn)率的提高和成本的降低，是否會導(dǎo)致需求的大幅增長，還是說只是小幅增長。當(dāng)流水線——一種具有g(shù)pt特征的工藝創(chuàng)新——可以讓亨利·福特削減汽車制造成本時(shí)，需求激增，工人受益。例如，如果人工智能提高了生產(chǎn)力并降低了醫(yī)療成本，這可能會導(dǎo)致對醫(yī)療服務(wù)和專業(yè)人員的需求大幅增加。

AI might well augment the productivity of workers of all different skill levels, even writers. Yet what that means for an occupation as a whole depends on whether improved productivity and lower costs lead to a big jump in demand or only a minor one. When the assembly line—a process innovation with GPT-like characteristics—allowed Henry Ford to cut the cost of making cars, demand surged and workers benefited. If AI boosts productivity and lowers costs in medicine, for example, that might lead to much higher demand for medical services and professionals.

強(qiáng)大的人工智能有可能打破歷史模式。一項(xiàng)幾乎可以處理我們普通人能做的任何任務(wù)的技術(shù)，將把人類帶入未知的經(jīng)濟(jì)領(lǐng)域。然而，即使在這種情況下，過去也有一些經(jīng)驗(yàn)可循。伴隨著蒸汽革命而來的持續(xù)經(jīng)濟(jì)增長，以及伴隨著電氣化和其他后來的創(chuàng)新而來的進(jìn)一步加速增長，本身就是前所未有的。這些創(chuàng)新促使人們爭相發(fā)明新思想和新制度，以確保重大的經(jīng)濟(jì)變革轉(zhuǎn)化為廣泛的繁榮，而不是混亂。也許很快就會是再次變革的時(shí)候。