Anthropic warns that it’s very close to RSI capability, aka self-building AI

Anthropic researchers have identified that current AI development is nearing the threshold of Recursive Self-Improvement, a state where models become proficient at enhancing their own training processes. This capability could lead to a rapid, autonomous acceleration in system performance, fundamentally changing the pace of technological advancement. By reaching this stage, the industry faces new technical and safety challenges regarding the control and predictability of future models.