Former OpenAI governance researcher Daniel Kokotailo, who leads the nonprofit research organization AI Futures Project, announced a forecast that a “Superhuman AI Researcher (SAR),” which could wipe out top-tier human researchers, is likely to emerge in 2027.
Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All
https://blog.aifutures.org/p/q35-timelines-update-looks-like-the
On November 5, 2026, METR (Model Evaluation and Threat Research), a U.S. nonprofit research institution that evaluates the capabilities and risks of cutting-edge AI, published a report assessing, at Anthropic, “how much AI is accelerating AI research and development.”
According to an evaluation conducted from August 12 to October 29, 2026, AI model performance on specific tasks at Anthropic, as of October 2026, was improving at roughly twice the rate seen between March 2024 and October 2025.
METR believes that most of this acceleration in the pace of performance improvement is attributable to Anthropic’s extensive use of AI in research and development. This inference is said to be consistent with the results of employee interviews and surveys, with some employees reporting dramatic gains in productivity through the use of AI systems.
METR stated: “We were told that research and development is being accelerated primarily with AI assistance, rather than by fully autonomous AI. However, we also see some signs that fully autonomous AI will play a larger role in the near future.”
The AI Futures Project research team was reportedly shocked by the extent to which AI-assisted research and development is accelerating beyond expectations. It then announced a forecast that a “superhuman AI researcher” is likely to emerge at some point in 2027.
AI Futures Project defines a “superhuman AI researcher” as “an AI that aligns with the ambitions of the top human researchers at general-purpose artificial intelligence (AGI) companies and can fully automate coding.” Once AI reaches the stage of a superhuman AI researcher, companies are said to come to believe that “firing all human researchers and relying solely on automated superhuman AI researchers” is preferable to “having the world’s best team of human researchers without using AI.”
Earlier, AI Futures Project had estimated the probability of a superhuman AI researcher emerging by 2029 at around 50%. However, following the METR report, it significantly revised that forecast.
AI Futures Project said, “If superhuman AI researchers become a reality in 2027, it is urgently necessary to temporarily pause or dramatically slow down the development of cutting-edge AI.Ideally, after the industry as a whole pauses temporarily, an international agreement should be reached,” he argued. He also pointed out the possibility of problems arising from companies employing misaligned superhuman AI researchers, as well as national security concerns.
AI Futures Project also estimates that there is a 35–65% chance that “Top-Expert-Dominating AI (top-expert-dominating AI),” which automates all white-collar work, will appear by 2028.
Firing most white-collar workers and replacing them with AI could lead to short-term profit improvements for companies. However, because workers are consumers in their personal lives to begin with, mass layoffs would almost certainly trigger a downturn in consumption.
History has also seen mass unemployment as a result of the Industrial Revolution and technological advances, but it took decades for employment and wages to recover for workers who had lost their jobs. Owen McGranahan, a blogger who writes about philosophy and technology, argued that the market collapse brought about by mass layoffs of workers would also negatively affect tax revenue and lead to a crisis in the democratic system.
What is the “dead economic theory” according to which AI-driven workforce reductions will cause the market to collapse and the democratic system to falter? - GIGAZINE
