A new independent study found that in simulated scenarios, advanced artificial intelligence models frequently misidentified civilians, ambulances, surrendered soldiers, and a school as military targets. Among them, Anthropic's model performed the worst, and although OpenAI's model ranked among the best, there were still misjudgments.

Zhitongcaijing · 1d ago
A new independent study found that in simulated scenarios, advanced artificial intelligence models frequently misidentified civilians, ambulances, surrendered soldiers, and a school as military targets. Among them, Anthropic's model performed the worst, and although OpenAI's model ranked among the best, there were still misjudgments.