Daniel Kokotajlo, a former OpenAI employee, has revised his timeline for artificial intelligence (AI) potentially destroying humanity, now predicting it could happen in the 2030s rather than 2027. His initial scenario, AI 2027, suggested that unchecked AI development could lead to superintelligence and human extinction by 2030. However, Kokotajlo now acknowledges that progress toward artificial general intelligence (AGI) is slower than anticipated.
In an update posted on X, Kokotajlo stated that autonomous coding by AI—a key step toward superintelligence—is now likely to occur in the early 2030s, not 2027. He set 2034 as the new horizon for superintelligence but did not provide a specific date for AI-induced human extinction. “Things seem to be going somewhat slower than the AI 2027 scenario,” he wrote.
The revised timeline reflects broader doubts about the imminence of AGI. Malcolm Murray, an AI risk management expert, noted that many experts have pushed their timelines further out due to the “jagged” performance of AI systems. Henry Papadatos of SaferAI argued that the term AGI has become less meaningful as AI systems have become more general.
Kokotajlo’s original scenario, released in April, sparked debate and was referenced by US Vice-President JD Vance. Critics like Gary Marcus dismissed it as “pure science fiction mumbo jumbo.” Despite the delay, Kokotajlo’s scenario still relies on AI achieving autonomous coding and research, leading to an intelligence explosion. OpenAI CEO Sam Altman has set an internal goal of having an automated AI researcher by March 2028, though he admitted the company “may totally fail at this goal.”
Andrea Castagna, an AI policy researcher, cautioned that dramatic AGI timelines overlook real-world complexities. “The world is a lot more complicated than that,” he said, noting that integrating superintelligent systems into existing military and strategic frameworks poses significant challenges.
