

I think there’s a really important distinction between “getting the same result” when that outcome is guaranteed and when it isn’t. Using a brick instead of a hammer to squash something will get you the same result every time. But with an LLM there’s no guarantee you’re going to get any specific outcome - will it hallucinate this time or not? - and so even if it gives you what you wanted this time, you have to account for the probability that existed it would not have done (and that you wouldn’t have known it let you down)
Hi-viz, you say?