GPT-3 is only applicable in a narrow field too, but that field is "speaking English" which activates the Eliza effect.
Yes, but that idea + scale + SFT + RL made a much more general problem solver that can solve difficult verifiable tasks irrespective of human observers or the Eliza effect.
Yes, but that idea + scale + SFT + RL made a much more general problem solver that can solve difficult verifiable tasks irrespective of human observers or the Eliza effect.