4 ms·
I have a vague feeling that graphics and text AI are recognizing textures rather than structures. One funny example is dogs disguised as panda: https://img.huf
by est 4y ago
I have a vague feeling that graphics and text AI are recognizing textures rather than structures.
One funny example is dogs disguised as panda: https://img.huffingtonpost.com/asset/5cd73e1221000059007aca59.jpeg https://img.huffingtonpost.com/asset/5cd73e1221000059007aca5... because it's black and white. I am sure AI would gradually solve this problem but this makes me wonde, could AIs really "understand" shapes and structures, narrative logic, or even "reasoning" like humans do?
- aldousd666 4y agoConvolutional filters can be visually interrogated. You can see exactly which hits it makes at every layer of each filter. There's some cool videos on YouTube illustrating it. You can't tell as the end user, but the developers don't have to theorize about what is matching. Not with the tools we have these days.