03900nam a2200277 i 4500 47548 ES-BaOER 20260618135415.0 cr ||||||||||| 260617s2025 xx |||||s||||o0|| ||eng|d ES-BaOER cat ES-BaOER AI 2027 [Recurs electrònic] / Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean [S.l.] : AI Futures Project, 2025 71 p. : digital, fitxer PDF (8,45 Mb) Es tracta d'un informe prospectiu sobre l’evolució de la intel·ligència artificial en els propers anys. S’hi explora la possibilitat que els sistemes d’IA arribin a accelerar el seu propi desenvolupament, fet que podria conduir a l’aparició d’una superintel·ligència artificial (AIG). I examina les implicacions tecnològiques, econòmiques, socials i geopolítiques d’aquest escenari, així com els reptes relacionats amb la seguretat, el control i l’alineament dels sistemes avançats d’IA. El seu objectiu és fomentar la reflexió sobre els possibles impactes d’aquesta tecnologia i la necessitat de preparar-se per a canvis profunds en la societat. (Font: El Centre) Mid 2025: stumbling agents -- Late 2025: the world's most expensive AI -- Early 2026: coding automation -- Mid 2026: China waker up -- Late 2026: AI takes some jobs -- January 2027: agent-2 never finishes learning -- Feabruary 2027: China steals Agent-2 -- March 2027: algorithmic breakthroughs -- April 2027: Alignment for Agent-3 -- May 2027: National security -- June 2027: Self-improving AI -- July 2027: the cheap remote worker -- August 2027: the geopolitics of superintelligence -- September 2027: Agent-4, the superhumanAI researcher -- October 2027: goverment oversight -- Race ending: November 2027: superhuman politicking ; December 2027: The Agent-5 collective ; 2028: The AI Economy ; 2029: The Deal ; 2030: Takeover -- Slowdown ending: November 2027: tempted by power ; December 2027: A US-China Deal? ; January 2028: A safer strategy ; February 2028: superhuman capabilities, superhuman advice ; March 2028: election prep ; April 2028: Safer-4 ; May 2028: superhuman AI released ; June 2028: AI alignment in China ; July 2028: The deal ; August 2028: Treaty verification ; September 2028: Who controls the AIs ; October 2028: the AI economy ; November 2028: election ; 2029: Transformation ; 2030: peaceful protest -- Appendices: A. Training process and LLM psychology: why we keep saying “hopefully” ; B. The AI R&D progress multiplier: what do we mean by 50% faster algorithmic progress? ; C. Why our uncertainty increases substantially beyond 2026 ; D. The theft of Agent-2 model weights ; E. Neuralese recurrence and memory ; F. Iterated distillation and amplification (IDA) ; G. Why we forecast a superhuman coder in early 2027 ; H. The alignment plan ; I. Managing a corporation of AIs ; J. How we’re forecasting the capability progression beyond superhuman coders ; K. Alignment over time ; L. Our uncertainty continues to increase ; M. The slowdown ending is not a recommendation ; N. Superintelligent mechanistic interpretability ; O. Supepersuasion ; P. Superintelligence-enabled coordination technology ; Q. Robot economy doubling times ; R. Power grabs ; S. Verification mechanisms for an international agreement ; T. OpenBrain’s new alignment strategy ; U. Robot economy doubling times ; V. So who rules the future? ; W. Reminder that this scenario is a forecast, not a recommendation Kokotajlo, Daniel 18965 Alexander, Scott 18966 Larsen, Thomas 18967 Lifland, Eli 18968 Dean, Romeo 18969 AI Futures Project 18970 E-Link https://share.google/rlSTUYbGxfKPy9qZV + Informació de l'obra https://ai-2027.com EL 47548 47548 0 0 udc 0 0 G CERC CERC RD 2026-06-17 D 0 E260016 E260016 2026-06-17 16:10:04 https://cercles.diba.cat/cgi-bin/koha/opac-retrieve-file.pl?id=4cda4d26ac0e8726945a1453504aaf51 2026-06-17 EL