2 ms·

Wouldn't this kind of priming be brittle and unreliable? Has anyone successfully primed GPT-3 to solve complex problems consistently?