There's a new test of creativity by @mcgillu McGill U and @UniMelb so of course I had to try @OpenAI 's GPT-3 on it. Scored 82.32! So GPT-3 is more creative than 74% of people who have completed it. #NLProc

Jul 30, 2021 · 7:22 AM UTC

2
2
13
To be fair, GPT-3 with the settings I shared entered some compound nouns which are not permitted. Maybe some GPT-3 aficionados will be able to help me constrain output to just single word nouns, I have failed after various attempts
Replying to @julianharris @mcgillu @UniMelb @OpenAI
Interesting, though feels somewhat biased in GPT-3’s (/any automated system’s) favor since speed seems to be a big part of what makes it varyingly difficult among humans
2