Google Engineer Says AI Research Shows No Breakthrough Speedup, Self-Improving AI Still Distant

Google engineer Haydn Belfield has stated that self-improving artificial intelligence remains far from being realized. According to The Decoder, Belfield highlighted that research at Deepmind has not experienced any measurable acceleration driven by AI itself.

The Decoder also reports that while AI now generates more than 75 percent of production code, 88 percent of code changes still require human supervision. Deepmind’s internal General Capability Index indicates steady progress without any sudden increases in development speed.

This cautious outlook on AI progress is relevant for Japanese markets, where investors are closely monitoring AI’s impact on technology sectors, including software development and automation trends.