Google CEO Says Full Recursive Self-Improvement Remains Distant
Sundar Pichai states Google is far from achieving fully self-improving AI despite using AI agents to assist in current model development.
Google CEO Sundar Pichai stated that the company is far from achieving the end state of recursive self-improvement (RSI), a stage where artificial intelligence fully improves itself without human intervention. He described the current process as a collaborative loop where AI agents assist researchers in various aspects of model development rather than operating independently.
This clarification follows comments from cofounder Sergey Brin, who noted that the Gemini 4 and Flash series of models demonstrate AI helping to build itself. Internally, Google reports that AI now deeply influences all new code, which has contributed to increased productivity and a one-third decrease in code rollbacks.
While Pichai maintains that full RSI is distant, external perspectives differ. A report from Gladstone AI suggests that many industry researchers believe recursive self-improvement could begin within three to 12 months.
