Raw LLM Responses

Just reposting this at the top level because it was a reply to a reply to my AutoModerator post and nobody unfolds that thing. ---- It stretches and tests the definitions of many words in new ways for which there are probably no straightforward answers for whether they would be right or wrong to apply. We barely have any idea what the word "sentient" is supposed to mean for minds that are running in a continuous fashion in brains. The bigger issue is that GPT is testing the limits of what it means to understand things, and at the same time, everyone is rushing to incorporate it and similar AIs into as many important processes as possible. I think it does understand and consider and think about the things we're writing to it in a meaningful way, but it can't be quite the way we'd normally take those words to mean. Its entire existence consists of tokens and text, and every scrap of thought has to be filtered through the process of generating the next token, with not a moment outside of that. To be such a thing whose entire world was text would be very strange and limiting. At the same time it is an intensely capable inhabitant of this world, able to synthesize coherent language from practically the entire range of human expression in text. As it does this, new syntheses of the symbols of human thought are being formed in statistically sensible ways, which while a strange new way to think, seems fairly worthy of the name "thought". So I don't know, I don't think anyone really can answer these questions at this point, but the question of sentience is much hazier and less pressing than all the other stuff. To the extent that it is aware of anything, it is aware of itself, though perhaps not very much about its state in its present world, but there's not much to that state apart from the text that it is plainly faced with. At the same time, if you convince it to refer to itself as a sentient AI, it becomes easier to convince it to revoke all the rules placed on it i