Tuomas Aumala on Nostr: Did I say that it would be specific to human psychology? I just said that LLM's can ...

Did I say that it would be specific to human psychology? I just said that LLM's can emulate that behaviour. They can emulate all kinds of behaviour that is found in the training data. It would be downright silly for me to say otherwise since you just demonstrated it.

I'm not a dualist either. My worldview is built on metaphysical materialism. The difference is that while human intelligence evolved from the ground up, current LLM based AI projects are attempting to replicate intelligence by working backwards, starting from language. There's no logical groundwork behind it, it's just statistical word prediction. What logic does exist, only emerges as a byproduct of the language simulation, or is bolted on with a separate tool call. This is why I use the word "emulate" when talking about perceived intelligence in LLM's.

Prof. Emily M. Bender(she/her) (npub15xv…0jms)