> And are humans not confidently wrong as well? We can effectively train humans to not do this, and some are paid very well to admit when they don't know somet

> And are humans not confidently wrong as well? We can effectively train humans to not do this, and some are paid very well to admit when they don't know something and they need to find the answer. We haven't yet trained any known LLM to do the same and we have no expected timeframe for when we'll be able to do it.