AXIOS (Ina Fried) - Models that improve on their own are AI's next big thing
AI models that can learn as they go are one of the hot new areas drawing interest from both startups and the leading labs, including Google DeepMind . Why it matters: The move could accelerate AI's capabilities, but also introduce new areas of risk. Known technically as recursive self-improvement, the approach is seen as a key technique that can keep the rapid progress in AI going. - Google is actively exploring whether models can "continue to learn out in the wild after you finish training them," DeepMind CEO Demis Hassabis told Axios during an on-stage interview at Axios House Davos. - Sam Altman said in a livestream last year that OpenAI is building a "true automated AI researcher" by March 2028. What they're saying: A new report …
AI:AI companies and startups are increasingly focusing on "recursive self-improvement," a process where AI models learn and improve themselves continuously rather than remaining static after training. While leaders like Google DeepMind's Demis Hassabis and OpenAI's Sam Altman explore this technology to accelerate scientific research and R&D, experts warn that it introduces significant risks. Unlike controlled environments like chess, real-world applications of self-improving AI could lead to harder-to-detect errors, unintended consequences, or misuse, prompting calls from researchers for better transparency and updated safety frameworks.Open