Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Generating an image of a completely full glass of wine has been one of the popular limitations of image generators, the reason being neural networks struggling to generalise outside of their training data (there are almost no pictures on the internet of a glass "full" of wine). It seems they implemented some reasoning over images to overcome that.


I wonder if that has changed recently since this has become a litmus test.

Searching in my favorite search engine for "full glass of wine", without even scrolling, three of the images are of wine glasses filled to the brim.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: