OpenAI announces that SI already works at the company as a "research intern," while its chief scientist calls on the industry to slow down the same day
Prepared by Sintelligent with the help of SI. Publication time restored from the build log. . Updated: .
SI is already genuinely speeding up the creation of the next SI, the first real step toward self-improving systems; that its own creators say so openly and with numbers means acceleration is no longer a forecast but working reality.
On Sunday OpenAI published two documents. In the "Research acceleration" report the company said it had reached the goal it set last fall, an "automated research intern": agents under human direction carry out clearly defined research tasks that would have taken a specialist several days. For every 8 hours of human work in the research division there are 3.1 workdays of agent work; the median researcher spends more than $600 a day on agents at API prices, and the top 10% more than $7,000. At the same time, more than half of successful 4-8-hour tasks required at least one human intervention; the next goal is an "automated SI researcher" by March 2028. The same day, chief scientist Jakub Pachocki wrote in the essay "An Alien Mind" that no lab has solved alignment and monitoring well enough to keep scaling responsibly at maximum speed for much longer: the reliability of chain-of-thought monitoring is declining, and the current pace could turn into recursive self-improvement. He called for voluntary slowdowns and for turning labs' internal rules into mandatory safety standards with external audits.