News for hackers. Stuff that matters.

A comment on DALL·E: Creating Images from Text.

Re:DALL·E: Creating Images from Text by captainmuon on Tuesday January 05, 2021 @04:22PM (#25651252)
Maybe I'm cynical, but I'm really skeptical. What if this is just some poor image generation code and a few hundred minimum wage workers manually assembling the examples? Unless I can feed it arbitrary sentences we can never know.

I would be disappointed, but not surprized if OpenAI turns out to be the Theranos of AI...

[ Parent | on HN ]
Re:DALL·E: Creating Images from Text by apatap on Tuesday January 05, 2021 @05:20PM (#25651991)
At least GPT-3 can generate texts much faster than a worker would need to create them manually.
[ Parent | on HN ]
Re:DALL·E: Creating Images from Text by captainmuon on Tuesday January 05, 2021 @05:54PM (#25652363)
Right, but I bet the images shown here were preselected.
[ Parent | on HN ]
Re:DALL·E: Creating Images from Text by sanxiyn on Tuesday January 05, 2021 @06:19PM (#25652630)
They swear they aren't preselected. I am willing to bet you, but I am not sure how to settle the bet.
[ Parent | on HN ]