The temperature scale, I think. You divide the logit output by the temperature before feeding it to the softmax function. Larger (resp. smaller) temperature results in a higher (resp. lower) entropy distribution.

come up with new and unexpected things that never existed before

I’m not sure this is possible if the tech is still primarily built by learning from data, which by definition, has existed.

only in the podcasts I listen to

Yes definitely. Many of my fellow NLP researchers would disagree with those researchers and philosophers (not sure why we should care about the latter’s opinions on LLMs).

it’s using tokens, which are more like concepts than words

You’re clearly not an expert so please stop spreading misinformation like this.

You seem very certain on this approach, but you gave no sources so far. Can you back this up with actual research or is this just based on your personal experience with chatgpt4?

Can you provide the source of a few of these completely different LLMs?

add even a small amount of change into an LLM […] radically alter the output

You mean perturbing the parameters of the LLM? That’s hardly surprising IMO. And I’m not sure it’s convincing enough to show independence, unless you have a source for this?

two totally independent LLMs

How do you propose to get these independent LLMs? If both are trained using similar objectives e.g., masked token prediction, then they won’t be independent.

Also, assuming independent LLMs could be obtained, how do you propose to compute this hallucination probability? Without knowing this probability, you can’t know how many verification LLMs are sufficient for your application, can you?