A @OpenAI GPT-2 prompt taken from @GaryMarcus's The Algebraic Mind, showing what neural nets couldn't do. Sometimes a quantitative change in model/data complexity can give rise to qualitatively different behavior.pic.twitter.com/FGKg9W9UnU
-
I think it's interpreting the prompt as the start to a poem or some other literary device. If you add numbering it gets the hint:pic.twitter.com/VtqLPyaSO1
now that is interesting. do you get essentially the same result every time?
-
-
Those were first attempts; not actually sure if their decoding scheme is stochastic or not. Would be nice to see someone do a more systematic probing of the model, as I'm confident its ungrounded nature does entail some major limitations.
- 2 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.