From AI 2027 to Uncertainty
Artificial intelligence researcher Daniel Kokotajlo, a former OpenAI employee, has tempered his earlier predictions about the speed of AI’s march toward superintelligence. Kokotajlo previously sparked debate with his scenario AI 2027, which envisioned unchecked AI development leading to autonomous coding systems, rapid self‑improvement, and ultimately the destruction of humanity.
Debate Over AI Timelines
Released in April, AI 2027 quickly attracted both admirers and critics. US Vice President JD Vance referenced the scenario in a discussion about America’s AI arms race with China, while Gary Marcus, emeritus professor of neuroscience at NYU, dismissed it as “pure science fiction mumbo jumbo.”
Timelines for transformative AI—often referred to as artificial general intelligence (AGI), or systems capable of replacing humans at most cognitive tasks—have become a fixture in AI safety circles. The launch of ChatGPT in 2022 accelerated many forecasts, with some experts predicting AGI within decades or even years.
Rolling Back Predictions
Kokotajlo and his team initially identified 2027 as the “most likely” year for AI to achieve fully autonomous coding, though they acknowledged longer timelines were possible. Now, Kokotajlo has signaled doubts about the imminence of AGI, suggesting that the path to superintelligence may take longer than expected.