FULL STORY

AI 2027 team moves up intelligence explosion timeline

The AI 2027 team sharply moved up its forecast for an intelligence explosion to 2027, sparking fierce debate and a wave of updated predictions clustering around 2027-2028.

2026-11-11 ~ 2026-11-12 · 4 episodes · 45 posts

Episode 1 · AI 2027 Team Moves Up Timeline: Intelligence Explosion May Come in 2027 (2026-11-11, 35 posts)

AI Futures Project (the team behind the AI 2027 scenario, including Daniel Kokotajlo, Eli Lifland, and Brendan Halstead) has released a Q3.5 timeline update. The core conclusion: the intelligence explosion is now projected to occur in 2027, with overall progress running about 5 months ahead of their original "AI 2027" scenario. The update has drawn wide attention, as it marks a notable acceleration in the team's expected arrival time of AGI.

Confirmed

  • The team released the Q3.5 timeline update, judging that the intelligence explosion will still happen in 2027, roughly 5 months ahead of the original AI 2027 scenario (m2, m5).
  • All median predictions for a superhuman AI researcher (SAR) now fall between May and November 2027 (m4).
  • The accompanying chart shows that the five forecasters (Eli, Thomas, Daniel, Romeo, Brendan) mostly converge on September 2028 for the arrival of "top expert-level AI," with estimates consistently revised downward since April 2025 (m3).
  • The update is primarily grounded in METR's latest estimates and data: the team fit recent METR data to progress driven by autonomous coding capability (approaching "automated coder AC" level), concluding that Anthropic's current pace of AI progress exceeds prior expectations (m1, m4, m5).

Not Yet Confirmed

  • The team itself cautions that public evidence supporting the earlier timeline is limited, and judgments beyond the METR data carry substantial uncertainty (m4).

Why It Matters

  • AI 2027 is one of the most widely circulated AGI scenario exercises in the field; the authors' decision to pull the timeline forward will directly shape expectations in policy circles and the safety community about when an intelligence explosion might arrive.
  • @DKokotajlo's post framed it as an "urgent update," signaling the team believes the new data are significant enough to change the pace assumptions of the original scenario rather than serving as a routine revision.

15 more related posts →

Episode 2 · Forecaster Moves Top-Expert AI Arrival Estimate Up to 2028 (2026-11-11, 2 posts)

Forecaster Eli Lifland has moved his estimate for top-expert-level AI from 2034 up to 2028, and an AI Futures report assigns a 35-65% probability to AI capable of automating all white-collar work arriving before 2028.

Episode 3 · AI Risk Community Split Over Whether an Intelligence Explosion Hits Next Year (2026-11-11, 6 posts)

The AI risk community has erupted in fierce debate over whether an intelligence explosion will arrive next year. Daniel Kokotajlo cited a new METR report saying an intelligence explosion is highly likely in 2027—about 5 months earlier than the AI-2027 scenario exercise he co-authored—and said he feels scared, sad, and angry. Tamay Besiroglu publicly pushed back: he believes the risk circle is converging on a consensus that an explosion is very likely next year and sees this as positive progress, but he personally judges that in a few years we'll find it didn't happen, and those who bet on it will lose substantial credibility.

Confirmed

  • Based on the new METR report, Kokotajlo moved his intelligence explosion expectation up to around 2027, about 5 months earlier than the AI-2027 scenario
  • Besiroglu acknowledges a consensus tendency in the risk circle toward "explosion very likely next year," but predicts that judgment will be falsified and bettors' credibility damaged
  • Hammond (see quoted discussion) puts a low probability on an intelligence explosion within 1–3 years
  • Oxford philosopher Toby Ord offers his personal estimate: with no mitigations, the probability of an intelligence explosion next year is about 1/8—well below the roughly 50% some claim—and he personally predicts an explosion is unlikely next year
  • Ord argues that even if an explosion may not happen, slowing AI development remains very important

Why it matters

  • The debate is effectively a test of the credibility of AI-2027-style warning frameworks: Ord notes that explosion-forecasters like Kokotajlo are only predicting "superhuman programmers" by 2027, not dramatic real-world impact that same year—not far from skeptics' view of "capability gains with limited external impact"; the two sides' actual disagreement may be amplified by public discourse
  • Ord stresses that the logic of AI-2027-style arguments is "the probability is non-trivial absent mitigation," so even a low probability can justify deceleration policy—providing a defense for risk decisions that doesn't hinge on any single prediction

Episode 4 · Leaked AI Timeline Forecasts Cluster Around 2027-2028 (2026-11-11, 2 posts)

A compiled forecast table shows six AI timeline forecasters' median predictions clustering around 2027-2028. Oxford philosopher Toby Ord cautioned the table omits key forecasters and slowdown scenarios, limiting its reference value.