To even judge "George Carlin: I'm Glad I'm Dead," an hour-long YouTube video of bad AI-generated images, narrated and purportedly "written" and "performed" by a generative artificial intelligence model as an imitation of a dead famous human man, as a "comedy special" would be to lend it legitimacy. "I'm Glad I'm Dead" is a weak, watered down, boring approximation of the artistic performances a human person honed over the course of a lifetime by combining his experiences of living with Carlin's specific interpretation of the world and his incredible amount of work. And while his performances were insightful, upsetting, offensive, and cathartic, above all they were funny, and this shit is not funny at all.

But expressing an opinion like that comes with the implication that this computer's attempt at fake art was bad, and not that any computer's attempt at fake art is and will be bad. The very concept of a computer's attempt at fake art is bad. To quote a particular dead famous human man, it's all bullshit and it's bad for you.

I'd be worried about that if I thought "I'm Glad I'm Dead" was actually "written" by AI, but I don't. What we're talking about here is a combination of technologies that can be impressive--the Fake George Carlin of the video sounds a fair amount like George Carlin sounded and if you don't listen to the words or clock that "I'm Glad I'm Dead" is pretty radically devoid of actual humor, you might be briefly fooled. But the Large Language Models (LMM) we're talking about when we talk about "AI" "writing" something are statistical programs that use a lot of data to analyze patterns and then guess at making other patterns. An LLM cannot create anything new; it can only rearrange what it already has. And having watched "I'm Glad I'm Dead," I'm pretty sure this thing is more Mechanical Turk than Mr. Data.