AI may one day gain the ability to eradicate humans, but AI already has the ability to devastate us by taking our jobs and livelihoods.
At the opening ceremony of the United Nations University Artificial Intelligence Hub for Climate, Governance and Culture, I gave a speech about AI’s capacity to displace jobs. AI has already displaced a measurable portion of the workforce and may disrupt employment on an unprecedented scale. Our governmental structures are not equipped to absorb a displacement of labor that rapid or that large.
Past innovations always had significant impacts on employment. Technological innovations have replaced workers by automating discrete and definable tasks, but the making of judgments has always remained with workers. After a period of adjustment, innovations created new jobs. These principles held true through industrialization, through mechanization, and through the computerization of clerical work. However, a system capable of identifying a problem it has not previously encountered, formulating a solution, decomposing that solution into discrete steps, and resolving those steps in a logical sequence is performing the very function we had reserved to humans—making judgments. As such, AI is quite capable of replacing large numbers of professionals, clerical and administrative workers, and manufacturing workers alike. This is an entirely new category of capability, not an incremental improvement upon the old paradigm.
A number of international institutions have studied the impact of AI on employment. Studies by the International Monetary Fund[1], the World Economic Forum, the Society for Human Resource Management, and PwC[2] converge on two conclusions: that artificial intelligence will result in significant losses in clerical and administrative employment at an accelerating pace, but that the positions created will ultimately exceed those eliminated.
The World Economic Forum's Future of Jobs Report of January 2025[3] projected that 92 million positions will be displaced worldwide by 2030 based on all structural drivers, of which artificial intelligence is a significant one. In May 2025, Anthropic's chief executive stated that AI might eliminate half of entry-level white-collar employment within one to five years.[4] The studies agree that clerical workers, cashiers, bank tellers, data-entry operators and paralegals will be disproportionately affected, and that entry-level positions are already scarce. But the same report projected that 170 million new positions, many because of AI, will be created in the same period so that there will be a net increase of some 78 million jobs. The Society for Human Resource Management, in research updated this June, finds that approximately 60% of occupations contain non-technical barriers to complete automation.[5] New roles are anticipated in machine learning engineering, in data science, in cybersecurity, and in renewable energy technology.
There are, however, significant weaknesses in these conclusions. The first is methodological; the second lies in the arithmetic of the net figure. Each study was conducted against a snapshot of capability that no longer holds true and does not describe the technology as it stands in September 2026—still less as it will stand a year from now considering the high velocity of development. In May, OpenAI reported that a reasoning model had disproved a conjecture in combinatorics unresolved since 1946[6]; in August, that an internal version of Astra had produced results on 10 long-open problems in mathematics and theoretical computer science, accompanied by machine-checkable Lean 4 certificates (results not yet peer-reviewed).[7] Professor Jacob Tsimerman, a Fields Medal recipient, has stated that he is confident artificial intelligence will very shortly become robustly superhuman at what professional mathematicians presently do.[8] There is no doubt that AI has crossed a critical boundary—the capability to replace human judgment.
Blue-collar workers face the same risk. Technology companies are also training robots for domestic tasks. In India, workers are engaged to generate that training data, recording ordinary household activity on head-mounted cameras at approximately two dollars an hour.[9] Even this work, however, is temporary. In August, the New York Times reported that of some 4.7 million industrial robots deployed worldwide, the great majority operate without vision—they are blind. A slight displacement of the workpiece halts them, at a reported cost of up to $100,000 per minute in a large automotive plant. Inbolt, a French firm, has developed three-dimensional cameras that give these machines sight. Stellantis has installed them and reports favorably.[10]
It is reported that Technology companies have raised nearly $500 billion in artificial-intelligence-related debt this year.[11] Capital deployed on that scale is not committed in expectation of a plateau. To wager policy upon the assumption that present capability represents a stable ceiling would be an error. We could be facing an existential challenge within a horizon measured in years rather than decades. ChatGPT was released in November 2022; four years later, these systems are resolving problems that had stood open since 1946. We need not resolve the question of when in order to take seriously and answer the question of what if. What we cannot responsibly do is to treat 26 weeks of unemployment insurance (in the U.S.) as though it constituted a policy. Should the transformation prove less severe than I fear, so much the better—But we must be prepared should it not.
[1] Mauro Cazzaniga, Florence Jaumotte, Longji Li, Giovanni Melina, Augustus J. Panton, Carlo Pizzinelli, Emma J. Rockall & Marina Mendes Tavares, "Gen-AI: Artificial Intelligence and the Future of Work," IMF Staff Discussion Note No. SDN/2024/001 (Jan. 14, 2024), https://www.imf.org/en/Publications/Staff-Discussion-Notes/Issues/2024/01/14/Gen-AI-Artificial-Intelligence-and-the-Future-of-Work-542379
[2] "AI Reshapes Global Labour Market into Two Distinct Paths, Rewarding Human Skills: PwC 2026 Global AI Jobs Barometer," PwC (June 15, 2026), https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-jobs-barometer.html
[3] https://www.weforum.org/publications/the-future-of-jobs-report-2025/
[4]Jim VandeHei & Mike Allen, "AI Jobs Danger: Sleepwalking into a White-Collar Bloodbath," Axios (May 28, 2025), https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic
[5] "Automation, AI, and Job Displacement Risk in U.S. Employment," SHRM Research (June 18, 2026), https://www.shrm.org/topics-tools/research/automation-ai-and-job-displacement-risk-in-us-employment/2026-full-report ("60.4% of wage/salary employment has at least one nontechnical barrier to automation displacement").
[6] "An OpenAI Model Has Disproved a Central Conjecture in Discrete Geometry," OpenAI (May 2026), https://openai.com/index/model-disproves-discrete-geometry-conjecture/
[7] "Ten Advances in Mathematics and Theoretical Computer Science," OpenAI (Aug. 1, 2026), https://openai.com/index/ten-advances-in-mathematics/ .
[8] "A Winner of Math's Top Prize Says AI Will Soon Surpass Mathematicians. He Fears What Comes Next," San Francisco Chronicle, https://www.sfchronicle.com/science/article/ai-openai-math-fields-medal-22358191.php
(byline unconfirmed).
[9] "Thousands of India Workers Are Helping AI Firms Train Robots to Replace Them," Bloomberg (Aug. 12, 2026), https://www.bloomberg.com/news/features/2026-08-12/thousands-of-india-workers-are-helping-ai-firms-train-robots-to-replace-them
[10] "How to Make a Robot Better at Its Job? Give It Eyes," The New York Times (Aug. 17, 2026); https://www.nytimes.com/2026/08/17/business/robots-stellantis-inbolt.html
[11] "Global AI Debt Issuance to Top $500 Billion in 2026, Morgan Stanley Says," Reuters, via Investing.com (2026), https://www.investing.com/news/economy-news/global-ai-debt-issuance-to-top-500-billion-in-2026-morgan-stanley-says-4734497