4 ms·
I was curious about this as well. Here is what I got: https://i.imgur.com/Ey0i3My.png https://i.imgur.com/Ey0i3My.png
by BudaDude 4y ago
I was curious about this as well. Here is what I got: https://i.imgur.com/Ey0i3My.png https://i.imgur.com/Ey0i3My.png
- lancesells 4y agoDall-E knows what a giraffe is IMO
- nextaccountic 4y agoI found it incredible too The text model of dall-e at very least can map "Long-necked spotted mammal that eats leaves from trees" near the same representation of "giraffe"
- Marazan 4y agoVery similar to my images for the prompt: https://imgur.com/8dw2brf https://imgur.com/8dw2brf Intriguing that it's gone for a headshot for all of them. I suspect it says something about the source text
- coel 4y agoI was curious about this too and wondered if it could then relate to my description of a giraffe so put in "legs of a yellow mammal with a long neck and spots" and it still nailed it.
- mechagodzilla 4y agoThe pictures are incredibly similar to what one gets when you type that same string in to google image search - it makes me wonder if you could do reverse-image search on some of the pictures it produces and find whole training pictures embedded in it, the way one can find blocks of intact training text stored inside language models.