Post #2409910

I find fascinating the process of anthropomorphizing large language model AI from a human behavior standpoint and, having both watched people interact with it and interacting with it myself, I would even say it's kind of the right approach as long as you remember that it is fundamentally not human. What it is, fundamentally, is a very complicated mathematical formula recursively operating on its own output given a seed input and feedback on whether its output is correct as per a human observer. As that observer, you do not have the capacity to understand every step of the formula, even if you could see it, which you can't, because it is a black box. We have, as humans, a mechanism in our psychology to handle systems that act that way. It's called theory of mind. And we use it when interacting with other humans because we have the same constraints on predicting their behavior. So people who deal with AI by pretending it's a human are using the right tool for the job, as long as they remember that they're using a tool, they're not interacting with a person.