File: windows-xp-desktop-backgr(...).jpg (57 KB, 800x600)

57 KB JPG

The real issue with AI as a tool for artists is that prompting what, for all intents and purposes, is a man with the taste of Kinkade in a black box simply doesn't cut it in terms of bringing to life a specific creative vision. AI video and art creators will only get across the most general idea of a prompt, and you will only get the AI's vision because it works by direct verbal command. This is fine if you have very general ideas, like "Peter Griffin vs. Goku epic fight," in which the specifics don't matter much as long as the surface level aesthetics and vibes are generally right. But if you are artistically minded, or really seeking any use case beyond "toy," then there is no reliable, practical way to guide or finesse it toward what you want. One can only roll the dice over and over, wasting time and electricity, and then iterate upon the closest outputs one can get. This then leads to degeneration of the generated material. Hypothetically (and, admittedly, ignoring ethical concerns regarding training data for a moment), if generative AI somehow provided mechanisms of very fine control that allowed the user to sculpt out exactly what he envisions, then that would satisfy the conditions for being a viable way to create art, and could potentially result in the "democratization of art" that defenders of current-day models insist is coming to pass, assuming a dextrous mastery of the tool is simpler for one to acquire than mastery of traditional mediums. But as of now, no matter how much these models improve the fidelity of their outputs, they will never be truly revolutionary in practice unless there are other, finer means of controlling it besides prompting and/or feeding it an image to move. In its present form, it can only produce waste or sloth.