Shortly after writing this answer, it occurred to me that you could dialog with GPT-3, rather than another human (or yourself). I’m not sure how accessible this is now, but there are certainly many alternatives. If even a silent rubber duck is useful, perhaps even an ELIZA-level chatbot is more useful, but I think options much better than that are freely available now.
As I mentioned in my other reply, silent ducks are not helpful for me, and similarly neither are chatbots. I indeed have used AI Dungeon for this purpose before and found it to be helpful for fictional stuff (as it’s intended), but less effective as a philosophical conversation partner. I have no idea how to access the real, full GPT-3, but I greatly would like to, for this exact reason.
https://www.eleuther.ai/ released GPT-J, which is supposed to have similar or superior performance when compared to GPT-3′s medium-sized models, although it can’t match the largest Davinci model on some tasks. Might be worth a try, but I’m not sure what kind of hardware you need to run it.
DeepMind’s Chinchilla suggests that GPT-3 was undertrained, so it’s possible to get better performance with fewer parameters. This space has been rapidly evolving, so I’m not sure what the current best free options are.
Shortly after writing this answer, it occurred to me that you could dialog with GPT-3, rather than another human (or yourself). I’m not sure how accessible this is now, but there are certainly many alternatives. If even a silent rubber duck is useful, perhaps even an ELIZA-level chatbot is more useful, but I think options much better than that are freely available now.
As I mentioned in my other reply, silent ducks are not helpful for me, and similarly neither are chatbots. I indeed have used AI Dungeon for this purpose before and found it to be helpful for fictional stuff (as it’s intended), but less effective as a philosophical conversation partner. I have no idea how to access the real, full GPT-3, but I greatly would like to, for this exact reason.
https://www.eleuther.ai/ released GPT-J, which is supposed to have similar or superior performance when compared to GPT-3′s medium-sized models, although it can’t match the largest Davinci model on some tasks. Might be worth a try, but I’m not sure what kind of hardware you need to run it.
DeepMind’s Chinchilla suggests that GPT-3 was undertrained, so it’s possible to get better performance with fewer parameters. This space has been rapidly evolving, so I’m not sure what the current best free options are.