就業恐慌是錯誤的:為什麼人工智慧創造的工作將比它摧毀的工作更多


https://text.ru/antiplagiat/69fc604685b0e

The Jobs Panic Is Wrong: Why AI Will Create More Work Than It Destroys

History Has Seen This Before: The Economic Case Against AI Doom

Cheap Intelligence, Bigger Markets: Why The AI Job Apocalypse Doesn't Add Up

The idea that AI is marching toward a future of mass permanent unemployment has gained considerable traction in public discourse. Yet this narrative rests on a foundation that economists have long recognised as flawed: the assumption that there is a fixed, finite quantity of work to be distributed among workers. 

This misconception, known as the "lump-of-labor" fallacy, has resurfaced in new form — dressed in the language of neural networks and large language models rather than steam engines and looms. 

David George, General Partner at venture capital firm Andreessen Horowitz, has compiled an extensive body of research that challenges the doom-laden consensus, drawing on historical precedent, economic theory, and emerging labor market data to argue that AI is far more likely to expand the frontier of human work than to eliminate it.

The core of the alarmist case is straightforward: cognitive tasks, long considered the exclusive domain of human intelligence, are increasingly performed by machines. If thinking can be outsourced to software, then the argument goes that human labor loses its fundamental value. What this reasoning overlooks, however, is that the falling cost of a productive input has never, in recorded economic history, simply caused demand for output to contract. 

When fossil fuels made energy abundant, the world did not merely retire its whalers — it invented entirely new industries that consumed energy at scales previously unimaginable. Jevons Paradox, the well-documented observation that efficiency gains tend to increase rather than decrease total consumption of a resource, applies just as readily to cognition as it does to coal.

Historical patterns reinforce this point with remarkable consistency. At the beginning of the twentieth century, roughly one in three American workers was employed in agriculture. The mechanisation of farming reduced that figure to around two percent by 2017, while farm output nearly tripled. Rather than producing a permanent class of unemployed farmhands, this transformation freed labor to flow into factories, offices, hospitals, and eventually the technology sector itself. 

Electrification followed an identical arc: factories reorganised around new workflows, productivity growth accelerated for decades, and entirely new categories of goods and employment came into existence. The introduction of spreadsheet software provides perhaps the most instructive parallel to the current moment — VisiCalc and Excel did not eliminate bookkeeping roles but instead catalysed an explosion in financial analysis, with roughly one million traditional bookkeeping positions giving way to one and a half million financial analyst roles.

The Augmentation Argument

The distinction between substitution and augmentation is central to understanding what AI is actually doing to labor markets at present. Goldman Sachs research suggests that AI augmentation effects more than offset the substitution effects across the economy as a whole, and corporate earnings calls reflect this balance in practice: references to AI as a tool that enhances human productivity outnumber references to AI as a replacement for workers by a ratio of approximately eight to one. 

Software engineers offer a telling illustration of augmentation in action — the volume of code being pushed to repositories has risen sharply, new application development is accelerating, and demand for software development talent has been trending upward since early 2025. Product management hiring has similarly rebounded toward levels not seen since 2022. If AI were substituting for human thinking on a one-to-one basis, one might expect demand for either engineers or product managers to fall as each discipline rendered the other less necessary. Instead, demand for both is growing, because the total volume of work being accomplished is expanding.

Wage data adds another dimension to this picture. Workers in roles characterised by high AI exposure appear to be experiencing above-average earnings growth, particularly in areas such as systems design. Meanwhile, research from the Federal Reserve Bank of Atlanta, the Census Bureau, and Yale's Budget Lab, among others, converges on a striking conclusion: across the broad economy, AI adoption has produced no statistically significant change in aggregate employment levels. 

A Census Bureau working paper found that only around five percent of AI-using firms reported any headcount impact at all, with increases and decreases distributed in roughly equal measure. These are not the fingerprints of a labor market in crisis.

What the Data Does Not Say

The nuanced picture that emerges from current research is one of reallocation rather than elimination. Entry-level roles with high substitution exposure have become harder to find in some sectors, while roles where AI serves as a complement have grown. Some occupations — customer service representatives and medical transcriptionists among them — face genuine structural decline. These transitions are real and carry costs for the individuals navigating them, and a serious policy response focused on retraining and workforce transition is both warranted and necessary.

What the data does not support, however, is the sweeping claim that AI represents a civilisational rupture in the relationship between humans and productive work. The underlying economic logic of that claim requires human ambition and human desire to freeze precisely at the moment that intelligence becomes cheap and abundant — a premise that contradicts everything observable about human behaviour. New business formation has risen sharply in correlation with AI adoption. 

Application development is growing at roughly sixty percent year-over-year. Robotics, long constrained by the computational demands of dynamic physical environments, is now moving from science fiction toward commercial reality, opening entire categories of employment that have never previously existed.

Technological transformation has always reshaped labor markets rather than simply shrinking them. The dominant economic sectors of every prior era gave way to larger successors, and the overall size of the economy and the labor market grew with each transition. 

AI will compress certain roles and eliminate certain tasks, as every general-purpose technology has done before it. The more important consequence, if history is any guide, is that it will simultaneously make many existing roles more valuable and generate demand for entirely new categories of work that are, at this moment, still beyond the horizon of imagination.

人工智慧正朝著大規模永久性失業的未來發展,這種觀點在公眾討論中獲得了相當大的關注。然而,這種說法建立在一個經濟學家早已認識到存在缺陷的基礎之上:即假設工作崗位的數量是固定的、有限的,可以分配給各個勞動者。 

這種被稱為「勞動總量謬誤」的錯誤觀念,以新的形式再次出現——它不再用蒸汽機和織布機來包裝,而是用神經網路和大型語言模型的語言來包裝。 

David George,創投公司的普通合夥人 安德森霍洛維茨,並彙編了大量研究成果,挑戰了這種悲觀的共識,借鑒歷史先例、經濟理論和新興勞動力市場數據,論證人工智慧更有可能拓展人類工作的邊界,而不是將其消滅。

這種危言聳聽的論調的核心很簡單:認知任務長期以來被認為是人類智慧的專屬領域,但如今卻越來越多地由機器完成。如果思考可以外包給軟體,那麼這種論點就認為人類勞動失去了其根本價值。然而,這種論證忽略了一個事實:在有記載的經濟史上,生產要素成本的下降從未直接導致對產出的需求萎縮。 

當化石燃料使能源供應充足時,世界並沒有僅僅讓捕鯨業退出歷史舞台——它還催生了全新的產業,這些產業的能源消耗規模達到了前所未有的高度。傑文斯悖論——即效率的提高往往會增加而非減少資源的總消耗量——這一已被充分證實的現象,同樣適用於認知領域,正如它適用於煤炭領域一樣。

就業恐慌是錯誤的:為什麼人工智慧創造的工作將比它摧毀的工作更多

歷史模式以驚人的一致性印證了這一點。 20世紀初,大約三分之一的美國勞工從事農業。農業機械化使這一比例在2017年降至2%左右,而農業產量幾乎增加了兩倍。這種轉變並沒有造成長期失業的農場工人群體,反而釋放了勞動力,使他們能夠流入工廠、辦公室、醫院,最終進入科技業。 

電氣化的發展軌跡與此類似:工廠圍繞著新的工作流程進行重組,生產力成長持續數十年,並催生了全新的商品類別和就業機會。電子表格軟體的引入或許是與當前情況最具啟發性的類比——VisiCalc 和 Excel 並沒有取代簿記員的工作,反而促成了財務分析領域的爆發式增長,大約一百萬個傳統的簿記員職位被一百五十萬個財務分析師職位所取代。

增強論證

區分替代和增強對於理解人工智慧目前對勞動市場的實際影響至關重要。高盛的研究表明,人工智慧的增強效應在整個經濟中遠大於替代效應,而企業財報電話會議也反映了這種平衡:提及人工智慧是提高人類生產力的工具的次數,與提及人工智慧取代工人的次數之比約為八比一。 

軟體工程師的例子生動地展現了人工智慧增強技術的實際應用——提交到程式碼庫的程式碼量急劇上升,新應用的開發速度加快,自2025年初以來,對軟體開發人才的需求一直呈上升趨勢。產品經理的招募也同樣反彈至2022年以來的最高水準。如果人工智慧能夠完全取代人類的思維,人們可能會預期,隨著某個領域的發展,對工程師或產品經理的需求會下降,因為另一個領域的重要性會降低。然而,實際情況卻是,由於工作總量的增加,兩者的需求都在增加。

薪資數據為這一圖景增添了另一個維度。從事人工智慧密集工作的勞工似乎正在經歷高於平均水準的收入成長,尤其是在系統設計等領域。同時,包括亞特蘭大聯邦儲備銀行、美國人口普查局和耶魯大學預算實驗室在內的多家機構的研究得出了一個驚人的結論:從整體經濟層面來看,人工智慧的應用並未對整體就業水平產生統計意義上的顯著影響。 

美國人口普查局的一份工作報告發現,只有約5%的人工智慧應用公司報告員工人數受到任何影響,而且增減人數大致相等。這並非勞動市場危機的跡象。

數據未說明的內容

目前研究呈現的圖景是重新配置而非淘汰。在某些行業,高替代性的入門級職位變得越來越難找,而人工智慧作為輔助手段的崗位則有所增加。一些職業——例如客服代表和醫療轉錄員——面臨著真正的結構性衰退。這些轉變是真實存在的,並且會為經歷這些轉變的個人帶來成本,因此,制定以再培訓和勞動力轉型為重點的嚴肅政策應對措施既是合理的,也是必要的。

然而,數據並不支持那種認為人工智慧代表人類與生產性工作之間關係的文明斷裂的籠統論斷。這種論斷背後的經濟邏輯要求人類的雄心和慾望在智能變得廉價且唾手可得的那一刻戛然而止——這一前提與所有可觀察到的人類行為都相悖。新企業的成立與人工智慧的應用呈現顯著正相關。 

應用開發正以每年約 60% 的速度成長。機器人技術長期以來受到動態實體環境運算需求的限制,如今正從科幻小說走向商業現實,開闢了前所未有的全新就業領域。

技術變革總是重塑勞動市場,而非僅僅縮小其規模。以往每個時代的主導經濟部門都會被規模更大的後繼部門所取代,而經濟和勞動市場的整體規模也隨著每一次轉型而成長。 

人工智慧將會壓縮某些崗位,淘汰某些任務,就像以往所有通用技術一樣。但如果歷史可以藉鑑,那麼更重要的後果是,它將同時提升許多現有崗位的價值,並催生出對全新工作類別的需求,而這些工作類別目前還超出了我們的想像。



Source link

免责声明:本文提供的信息不是交易建议。BlockWeeks.com不对根据本文提供的信息所做的任何投资承担责任。我们强烈建议在做出任何投资决策之前进行独立研究或咨询合格的专业人士。

Like (0)
Mpost的头像Mpost作者
Previous 5天前
Next 5天前

相关推荐

发表回复

Please Login to Comment
SHARE
TOP