A bunch of news articles made significant errors in reporting the change in AI Futures’ timelines between when they wrote AI 2027 and now.

One example being comparing and contrasting their modal estimates published last April (AI 2027) with medians published in December (their recent update).

This has the effect of making the change in their expectations for when different thresholds of powerful AI will be crossed look drastically larger than it really is.

Worth reading if you felt confused.