AI 2027 · AI Futures Project · April 2025

Scrub the timeline. Flip the assumptions. See which ending you get.

A concrete what-if: what happens if AI progress keeps compounding through the end of 2027? Each generation of AI accelerates AI research itself, so progress speeds up rather than staying linear. This explorer walks the scenario's shared trunk — mid-2025 to October 2027 — then hands you the fork: race or slowdown. The toggles show which assumptions flip the outcome.

1 · The shared trunk — mid-2025 to October 2027

Drag the scrubber. The story uses a fictional leading US lab (OpenBrain) and a fictional consolidated Chinese effort (DeepCent) to avoid naming real companies.

Jul 2025Mid 2026Oct 2027 · the fork

2 · What flips the ending?

Six assumption toggles. The oversight vote is the master fork; the others shift how likely the pause is to hold.

A note on dating

AI 2027 is a 2025 scenario to explore, not the authors' current forecast. The authors later revised their views — a mid-2025 note pushed the median timeline back roughly a year and a half, and by late 2025 / early 2026 the lead author's median AGI estimate had moved to the early 2030s. A 2026 follow-up sketches a more optimistic path. Read this as a what-if with a fork, not a prediction.

Also from the authors: the slowdown ending is explicitly NOT their policy recommendation. It depicts what they thought would more likely preserve human control — not a roadmap they endorse.

Share this scenario

Your month, your toggle choices, and the ending they produce — with attribution to the authors.