I didn’t expect artificial visual art to be such a low hanging fruit.

What about AI music? Is it as good but more discreet or is there something with music that is more complex?

One other thing that I didn’t expect is the asymmetry of ressources between training and inferencing. It seems to be like 5 or 6 orders of magnitude, AI has always been anthropomorphised with the same entity seemingly both « learning » and « acting ».

That makes current AI both extremely centralized for training and relatively decentralized to actually do something, I don’t know if it will change anything, but I don’t think it has been much thought about

For example AI could be soon stolen/copied and run locally like any software…