Great article, and a nice change from the doomsday "AI is taking over--human art is obsolete" posts I've been reading for years now. I've been using Midjourney to create AI art for my own amusement, to see if I can create images that exceed my very limited artistic ability. I've been amazed at how good it is, and how well it can adapt so many artists and styles into something fairly convincing. But the problem is that is can only base things on exciting models, and it doesn't really understand these models, much less the styles and aesthetics behind them. So the result can be visually striking, but like your opening sentence, grammatically (or syntactically) off. It also can't understand basic things like what human hands are or are used for--it can only look at how hands are portrayed in previous work. So hands often have too many fingers, or are too long, or melt along a given object. In short, it doesn't know who we are, why we create art, or what it's supposed to do to its audience: it can only trace what it sees with great accuracy. It would take true consciousness to bridge that gap, and I'm not sure that will happen in our lifetimes, if ever (maybe when we finally meet the first aliens from a distant galaxy??)