Increasingly coming to terms that the #GenerativeArt I've been interested in & practicing is somewhat incompatible w/ the generativeness of #AI image generators. One is about process, the other about results. Fundamentally they share the same roots, but overall v. different topics of inquiry... The latter (AI) is treated by almost everyone as a pseudo-magical black box & creativity is about finding/exploiting quirks in the human<>model interface (e.g. at the prompt level and/or providing custom learning/sampling materials/sources). The former, is about shaping & crafting a generative process itself, often the more simple, the better
#GenerativeArt #AI Maybe this is a false dichotomy, but it's something I can't stop thinking about of late. That, and also, that at some point soon we'll need new language/terms to differentiate between these (IMHO not just) nuances...
@toxi I resonate with this a lot. I find the black box prompt based generative stuff incredibly tedious and uninteresting, beyond other issues with it, and I've been thinking about that. I got into computing because I get excited about making beautiful little systems that I can understand and that can still surprise me and others. I want to build little watches, and prompt-based AI feels like I am mashing mud together...
@nasser @toxi to be fair, the right way to work with AI is to actually build it yourself. None of that stuff is magic or a black box when you actually dig into it, but the way it is currently being hyped, as a consumer-facing prompt interface to a pre-trained model, is pretty boring indeed. Building and training your own models, especially if you write things from scratch is an explorative and intentional process no different from any other kind of creative coding.
