Sam Altman announces internal goal: to develop an AI capable of independently conducting AI research by March 2028

Chainwire news: At a company forum roundtable, OpenAI CEO Sam Altman confirmed that OpenAI has set March 2028 as the official target for achieving the “automated researcher”—an AI system that can independently conduct AI research. He said that once achieved, the impact will be twofold: the system itself will prove that AI can handle advanced cognitive work, and it will also accelerate the pace of subsequent AI research—“progress in the following year will almost certainly be faster than our current speed.” Altman compared the present moment to the pre-COVID night in early 2020. He recalled that at the end of January 2020, OpenAI researchers realized earlier than the outside world that the pandemic would change everything; they began working from home in advance and disinfected door handles by貼ing copper strips, only to be mocked by the media. One night, he walked alone on the streets of the Mission District in San Francisco, looking through the windows at people in bars breathing face to face: “Something crazy is happening; change has already happened, but society hasn’t digested it yet. We tried to tell the world that it’s coming, but it’s hard to be understood. And now I have exactly the same feeling again.” The roundtable also revealed a detail: over the past few months, many OpenAI researchers have gone through a shift from writing code themselves to having AI write most of the code. Altman said this “capability overhang” is being underestimated—ordinary users try the AI again only about once every six months, while the AI’s capabilities are leaping every few weeks: “they stay on the default chat model and don’t even know that reasoning models have already arrived.” The roundtable was held on April 6, the same day OpenAI released its white paper “Industrial Policy for the Intelligent Age.”