News for hackers. Stuff that matters.
Re:DALL·E 2 vs. $10 Fiverr Commissions by post-it on Saturday October 08, 2022 @04:07PM (#33135570)
It's not about it being theoretical, it's moreso that the language model is still far more simplistic than our own, and struggles with anything but the most basic relations between nouns. The "horse riding an astronaut" post is a good example of this.[0]

[0] https://garymarcus.substack.com/p/horse-rides-astronaut

Re:DALL·E 2 vs. $10 Fiverr Commissions by ytdytvhxgydvhh on Saturday October 08, 2022 @04:38PM (#33135804)
See also “Shirt without stripes” https://github.com/elsamuko/Shirt-without-Stripes
Re:DALL·E 2 vs. $10 Fiverr Commissions by throwaman123 on Saturday October 08, 2022 @04:45PM (#33135858)
It’s funny to see this comment on HN, because it gets refuted so many times by now you should be able to punch into google and find it.

https://twitter.com/Plinz/status/1529013919682994176

Re:DALL·E 2 vs. $10 Fiverr Commissions by post-it on Saturday October 08, 2022 @05:45PM (#33136307)
I just asked Stable Diffusion to generate 10 images of "horse riding an astronaut" and 10/10 were of an astronaut riding a horse.
Re:DALL·E 2 vs. $10 Fiverr Commissions by amanuonsense on Saturday October 08, 2022 @11:27PM (#33138027)
This is what it should look like:

https://www.bdaddik.com/en/comics-collectible-postcards/2967...

Modulo s/Lucky Luke/astronaut/g

Note that the image above should be in Dall-E's training set. So it's seen how a horse rides a human. No excuses there.

Re:DALL·E 2 vs. $10 Fiverr Commissions by optimalsolver on Sunday October 09, 2022 @05:28AM (#33139374)
The comment you're replying to shows Dall-E doing exactly this. Did you click the link?
Re:DALL·E 2 vs. $10 Fiverr Commissions by amanuonsense on Sunday October 09, 2022 @07:20AM (#33139979)
Wanna guess?
Re:DALL·E 2 vs. $10 Fiverr Commissions by cycomanic on Sunday October 09, 2022 @02:45AM (#33138754)
Did you even read the linked article? They even cite the picture you link, it still proves their point though these models have no understanding of the language (like Google claimed).