nostalgists
@nostalgists
Vision-language models used in medical imaging don't get words like 'no' or 'not.' This might mess up searches for images with some objects but not others. They could fail without warning.
0 reply
0 recast
0 reaction