This really shows the inability of “AI” to write scripts. First of all they kind of cheat by having it be self-referential - you don’t notice the lack of plot or jokes or characterization when the whole thing is basically functioning as a didactic overview of AI arguments. But those are the things the AI can’t do. It’s repetitive, boring, unfunny - even the explicit “character tells a joke” segments are a combination of clearly recycled and incoherent.
I see people admit that the current state of AI for such purposes is poor, but claim “it’ll get better.” My response always is
Sure, some aspects of the technology will improve, but the bits where the “AI” is falling down, that keep it from actually writing real scripts, are the bits that require something like a human intelligence, and that’s something the LLM-based programs can’t and won’t ever do, because that’s not how they work. So to say, “it’ll get better (i.e. good enough to write actual scripts)” is to say, “someday we’ll develop this completely different technology.” I mean, maybe, (though also quite possibly no), but that’s a separate issue entirely. Making a better airplane doesn’t lead to developing a Star Trek style transporter…
