>>11139356You fucking dumbasses.
PhilosopherAI doesn't finetune GPT-3 based on user feedback.
>PhilosopherAI doesn't finetune GPT-3 based on user feedback. They finetuned it before deployment and now it's just running static weights.
"Open"AI, despite their name, have locked GPT-3 behind a paywalled API that has Jew conditions you must agree to use it, including agreeing to not use it for anything politically incorrect. PhilosopherAI agreed to these terms and is currently paying the Jew toll, and almost certainly not making their costs back on ads.
The closest you can get is GPT-2, which used to run on
TalktoTransformer.com, but that was costing the guy $20k in compute fees.
So the best you can do now is download the 355M model (instead of the full 1.2B that GPT-2 uses, and a far cry from the 175B that GPT-3 uses) and run it yourself with
vast.ai or a 1080/2080 Ti.
