牛津大学ai超越人类编年史英文版.docxVIP

When Will Al Exceed Human Performance ? Evidenee from Al Experts Katja Grace1 2, John Salvatier2, Allan Dafoe1,3, Baobao Zhang, and Owain Evans1 1 Future of Humamty Insbtute. Oxford University 2Al Impacts ^Department of Political Sdenoe. Yate University Abstract Advances in arti?dal intelligence (Al) will transform modern life by reshaping transportabon. health, aoence. ?nance. and the military [1. 2. 3]. To adapt pubhc poiicy. we need to better anbeipete these stfvnoea [4. 5], Here、ve report the results from a large survey ot machine learning researchers on their belefs about progress in Al. Researchers predict Al will outperform humarvs in many activities in the next ten years, such as translabng languages (by 2024). writing high-school eseays (by 2026). driving a truck (by 2027), working in retail (by 2031), writing a besteelling book (by 2049), and working as a surgeon (by 2053). Researchers believe there i8 a 50% c^snoe of Al outperforming humans in all task8 in 45 years and of automating all human jobs m 120 years, with Astan respondents expecting these dates much sooner than North Amencan?. These resUU will inform discussion amongst researchers and policymakers about antiapabng and managing trends in Al. Introductio n Advances in arti?cial inteDigence (Al) will have massive social consequences. Sett-driving technology might replace millions of driving jobs over the coming decade. In addition to possible unemployment, the transition will bring new challenges, such as rebuilding infrastructure, pro tecting vehicle cyber-security, and adapting laws and regulation8 (5). New challenges, both for Al developers and policy-makers. will aldo arise from applications in law enforcement, military teeb- noJogy. and marketing [6]. To prepare for these challenges, accurate forecasting of transformative Al would be invaluable. Several sources provide objective evidence about future Al advances: trends in computing hardware (7J. task perforfxwnce [8]. snd the automation of