I don't have a problem with hallucinations, rather the confabulations. We want AI to make predictions under uncertainty that could be wrong. What we don't need

I don't have a problem with hallucinations, rather the confabulations. We want AI to make predictions under uncertainty that could be wrong. What we don't need is getting know facts wrong. One of the best responses I got from ChatGPT was when it said "I don't know", and on questioning it - it responded that it was an unsolved problem and it couldn't objectively take a side iun the argument.