AI 2027: month-by-month model of intelligence explosion — Scott Alexander & Daniel Kokotajlo

AI 2027 builds its rapid-progress scenario around a feedback loop: AI helps conduct AI research, producing improved systems that can accelerate more research. In the interview, scenario authors Scott Alexander and Daniel Kokotajlo explain why coding receives so much attention. Their proposed turning point is when agents contribute enough useful research work to shorten the development cycle.

The scenario lays out that chain month by month and branches into two alternative endings, so each step can be examined, including whether research assistance actually translates into faster overall progress. Its authors do not present it as a certainty: Alexander calls its pace plausible enough to prepare for, but not his own median expectation.

“most people following the field have underestimated the pace of AI progress and underestimated the pace of AI diffusion into the world”

Daniel Kokotajlo

“I think in order to have nothing ever happen, you actually need a lot to happen.”

Scott Alexander

“if you’re wondering why hasn’t the AI done this, ask yourself, did the training environment train it to do this? And often the answer is no.”

Daniel Kokotajlo

“it might be useful to think of our timelines as being like 2070, 2100. It’s just that the last 50 to 70 years of that all happened during the year 2027 to 2028”

Scott Alexander

“my P(doom) is sort of infamously high, like 70%”

Daniel Kokotajlo