4 ms·
So it was rewarded/reinforced every time it brought things back to a positive message at the end?
by logicallee 4y ago
So it was rewarded/reinforced every time it brought things back to a positive message at the end?
- skilled 4y agoNot necessarily trained only to be positive but yes your assumption is right. I don’t know the exact details but I think there have been some OpenAI employees talking about this on their personal blogs. If I can find the links over the weekend I will make another reply.
- logicallee 4y agoThanks, I haven't seen any of their blogs and would love a link.