AI Futures Project's 'AI 2027' scenario maps superhuman AI and two endings
Eli Lifland reveals why his team wrote a detailed AGI takeover playbook by 2027.
In AXRP Episode 50, Daniel Filan sits down with Eli Lifland, a researcher at AI Futures Project, to dissect 'AI 2027' — the group's widely discussed scenario that plays out a concrete path from today to superhuman AI and potential AI takeover by 2027. Lifland explains that the project was born from a simple gap: many people speculate about AI doom or ubiquity, but few have tried to game out exactly how events would unfold week by week. AI 2027 fills that by offering a detailed, narrative-style forecast that readers can either agree with or push back against.
The scenario includes two distinct endings, reflecting uncertainty about whether AI alignment succeeds or fails. It also forecasts rapid time-horizon growth, effective compute expansion, and a shift from superhuman coding to general superintelligence. Lifland emphasizes that the team aimed for plausibility, with lead author Daniel Kokotajlo's timeline as a baseline, while still committing to specific claims that make the story testable. The reception has been significant: many readers say the scenario makes AI risk feel more immediate and actionable, which was exactly the creators' intent — to give the AI community a shared reference point for debates about timelines, misalignment, and what to prioritize now.
- AI 2027 was created by AI Futures Project, with lead author Daniel Kokotajlo and researcher Eli Lifland.
- The scenario includes two alternative endings and projects superhuman AI by 2027, with detailed forecasts on compute growth and time horizons.
- Episode 50 of AXRP is the first deep-dive interview specifically covering AI 2027, now available on YouTube and podcast platforms.
Why It Matters
A plausible, detailed AI takeover scenario helps professionals benchmark their own forecasts and prioritize alignment work.