Comment by jwells89
3 years ago
ML generalization and human abstraction are very different beasts.
For example, a human artist would have an understanding of how line weight factors into stylization and why it looks the way it does and be able to accurately apply these concepts to drawings of things they’ve never seen in that style (or even seen at all, if it’s of something imaginary).
The best an ML model can do is mimic examples of line art in the given style within its training data, the product of which will contain errors due to not understanding the underlying principles, especially if you ask it to draw something it hasn’t seen in the style you’re asking for. This is why generative AI needs such vast volumes of data to work well; it’s going to falter in cases not well covered by the data. It’s not learning concepts, only statistical probabilities.
I know what you're saying, and for sure existing models can be difficult to force into the really weird corners of the distributions (or go outside the distributions). The text interfaces are partially to blame for this though, you can take the images into Gimp and do some crude naive modifications and bring them back and the model will usually happily complete the "out-of-distribution" ideas. The Stable Diffusion toolboxes have evolved far away from the original simple text2image interfaces that midjourney and dalle use.
The models will generalize (because that's the most efficient way of storing concepts) and you can make an argument that that means they understand a concept. Claiming "it's not learning concepts, only statistical probabilities" trivialises what a modern neural network with billions of parameters and dozens of layers is capable of doing. If a model learns how to put a concept like line width 5, 10 and 15 pixels into a continuous internal latent property, you can probably go outside this at inference at least partially.
I would argue that improving this is at this point more about engineering and less about some underlying unreconcilable differences. At the very least we learn a lot about what exactly generalization and learning means.