U.S. Democratic Lawmakers Urge Congress to Legislate Ban on AI Recursive Self-Improvement This Year
U.S. Rep. Alex Bores (D) urges Congress to ban recursive self-improvement (RSI) in AI this year, citing a growing consensus backed by Sanders, Bannon, and Bengio. He warns that AI errors and alignment issues will intensify with each generation. OpenAI aims to achieve this by March 2028, while Anthropic employees suggest it could be reached as early as 2027. Referring to a METR report, Bores contends these projections are credible, making an immediate ban urgent.
U.S. Democratic Congressman Alex Bores has called for Congress to legislate a ban this year on recursive self-improvement (RSI) in AI—the process of allowing AI to automatically build the next generation of AI. He states that this has become a consensus supported by multiple parties, including Sanders, Bannon, and Bengio. He argues that any errors or alignment deviations in current AI will be magnified with each subsequent generation.
OpenAI has stated its plan to achieve this by March 2028, while employees at Anthropic claim it could be reached as early as 2027. Bores believes that a METR report indicates these predictions are likely accurate, making an immediate ban imperative.
Alex Bores (@AlexBores): https://x.com/AlexBores/status/2118494668769878037