4 ms·
This seems like a case where some prompt engineering (starting out with three good pickup lines) could improve the output a lot, or so I'd predict. Remember th
by Eliezer 6y ago
This seems like a case where some prompt engineering (starting out with three good pickup lines) could improve the output a lot, or so I'd predict.
Remember that tests like these do not upper-bound GPT-3's competence; they sometimes just show that you didn't ask the question the right way.
- wodenokoto 6y ago> Remember that tests like these do not upper-bound GPT-3's competence; The upper bound of a random character generator is the complete works of Shakespeare ... so I am not sure we are interested in the upper bounds.