4 ms·
Trained on 5 billion images to create a 6 GB model. Every image contributed no more than a few bits of information.
by habibur 4y ago
Trained on 5 billion images to create a 6 GB model. Every image contributed no more than a few bits of information.
- numpad0 4y agoBillion and Giga is both 10^9, so about 6/5 or 1.2 byte or 10 bits, and somehow enough to regurgitate examples from training set[0]. Oh how convincing “AI is just learning as humans do” arguments are. 0: https://twitter.com/kortizart/status/1588915427018559490 https://twitter.com/kortizart/status/1588915427018559490
- jjcon 4y agoWell I TIL humans are incapable of copying another piece of art even when asked to do so
- numpad0 4y agoIt's never too late to learn to draw!
- jjcon 4y agoI could learn to mine pigments and create paints too but we have better options now
- numpad0 4y agoPick up a pencil, start from drawing a circle, a real circular circle... Caveman in front of a VT100 is still a caveman. Caveman with a charred willow branch in hand, is something a bit more than that.
- Guillaume86 4y agoMaybe this incredibly famous photo from a National Geographic cover is overrepresented in the training set?