The AI industry’s most ambitious goal is for artificial intelligence to improve itself without human oversight. However, a recent study suggests that this may take longer than anticipated. Researchers found that current AI agents are unable to conduct open-ended research, which requires judgment and creativity needed for genuine breakthroughs.

The key question now is whether open-ended research is crucial for recursive self-improvement or if AI systems can achieve it by improving on narrower tasks alone. The results of this study may temper claims about the imminent arrival of recursive self-improvement in AI systems.

This is a summary. Read the original at technologyreview.com.