Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, and Romeo Dean predict that the impact of superhuman AI over the next decade will be enormous, exceeding that of the Industrial Revolution. They have written a scenario, AI 2027, that represents their best guess about what that might look like. The scenario is informed by trend extrapolations, wargames, expert feedback, experience at OpenAI, and previous forecasting successes. The CEOs of OpenAI, Google DeepMind, and Anthropic have all predicted that AGI will arrive within the next 5 years. Sam Altman has said OpenAI is setting its sights on “superintelligence in the true sense of the word” and the “glorious future.” AI 2027 explores two possible endings: a “slowdown” and a “race” ending. While not a recommendation or exhortation, the goal of AI 2027 is predictive accuracy. The authors encourage debate and alternative scenarios, aiming to spark a broad conversation about the future of AI and how to steer toward positive outcomes. The scenario was written iteratively, informed by tabletop exercises and feedback from over 100 individuals, including experts in AI governance and technical work. By making concrete predictions and inviting public disagreements, the authors seek to evaluate the accuracy of their forecasts in the coming years. Despite the challenges of predicting the trajectory of superhuman AI, the authors believe it is valuable to attempt in order to identify important questions and potential risks. One of the authors previously wrote a lower-effort AI scenario in August 2021, which was surprisingly successful in predicting certain trends in AI development. AI 2027 aims to provide a detailed and quantitative exploration of how AI could transform the world by 2027, inviting readers to consider the implications and uncertainties of a future powered by superhuman intelligence.
