I played a vanity game with @huggingface transformer demo with my name as a prompt. I first used my full name, but because of my last name, all hypotheses were very “Indian” & didn’t appear diverse. So I had to backoff to my first name. Here are some crazy cherrypicked examples.
-
Show this thread
-
This was definitely written by an HN user.
pic.twitter.com/6OfSFw8wfj
1 reply 0 retweets 4 likesShow this thread -
-
1 reply 0 retweets 6 likesShow this thread -
Huggingface knows about my transformer plans!
pic.twitter.com/MTGIIRqnx1
2 replies 0 retweets 6 likesShow this thread -
1 reply 0 retweets 1 likeShow this thread -
Almost all hypotheses were positive, which makes me think they’re doing some posthoc filtering for anything toxic (jigsaw?). That seems have made this demo safe, but has the unintended consequence of taking the edge off the hypotheses which makes generated text interesting.
2 replies 0 retweets 8 likesShow this thread
Replying to @deliprao
It might also be that the Delips of this world create overwhelmingly positive responses. Btw, this was my first experiment:pic.twitter.com/OPMLjX0hEU
10:46 PM - 13 Jun 2019
from Palo Alto, CA
1 reply
0 retweets
3 likes
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.


