Anthropic and China race toward recursive self-improvement in AI

Anthropic claims its Mythos model is approaching recursive self-improvement (RSI), a concept first coined by British mathematician Jack Good 61 years ago as an 'intelligence explosion.' RSI is considered AI's holy grail, where a machine enters a runaway cycle of self-improvement, leaving human intelligence behind. The company has called for a global pause on AI development due to risks of losing control once RSI is achieved, though critics dismiss this as marketing hype. Meanwhile, Chinese AI companies and researchers are also pursuing self-evolving AI systems. A Chinese developer stated that an implementable path to AI model self-evolution is already visible, speaking to a room of major policymakers and investors. The race between the US and China to achieve RSI could cement an unassailable lead for the first to succeed.