Study: AI models that consider user's feeling are more likely to make errors

Overtuning can cause models to "prioritize user satisfaction over truthfulness.”

Read full article from source