"They seem to very naturally understand the divide between artificial intelligence and humans and the notion that they both belong to the second realm."

I don't think they "understand" anything in the human sense. I specified in the prompt that both speakers are AIs, so they act like such. I also mentioned humans in the prompt as out-group beings, meaning the speakers aren't humans.

"Are those just regurgitated and recombined word and concept sequences...?"

Yes, I think so. If you look closely, most of the questions they ask are The Great Questions about AI and humanity. Easy to find on the internet. Tbh, I wanted something more sophisticated from them, but I got what I expected.

When reading these types of articles displaying examples of GPT-3's language abilities, we tend to let our imagination and hopes get in the way of our rational analyses.

Your last point is extremely interesting. I think we humans actually behave in a similar way. We also regurgitate what we've read or heard. We have mental states, which differentiates us from AI, but most of what we write and say are recombinations of things that were there already.

Very little creative production is truly novel - if we get philosophical, I'd argue nothing is 100% novel, ever. If we go deep enough, we'll always find the origins of creative work in something that already existed - regardless of the amount of transformations that took place to go from one to the other. New concepts are always born from already existing concepts.

Writing at the intersection of AI, philosophy, and the cognitive sciences | alber.romgar@gmail.com | Get my articles for free: https://mindsoftomorrow.ck.page

Writing at the intersection of AI, philosophy, and the cognitive sciences | alber.romgar@gmail.com | Get my articles for free: https://mindsoftomorrow.ck.page