A smart bullshit artist who loves the sound of their own voice is going to generate more hallucinations than a less smart, more humble, more self aware person.

A smart bullshit artist who loves the sound of their own voice is going to generate more hallucinations than a less smart, more humble, more self aware person. Making LLMs more knowledgeable is great (more data, bigger models, yay!), but there are other avenues of plausible attack as well. Enabling LLMs to know when they have veered off distribution might work. That is, the LLM doesn't have to know more of the world, it just has to know what it knows and stay there. A person who says "I don't know" is a lot more valuable than an overzealous one who spouts nonsense confidently. Encouraging an LLM to say that there is a disagreement about a topic rather than picking one lane is also a valuable way forward.