Can sci-fi films teach us anything about an AI threat?

Yes, because they teach us what we're expected to believe are the tropes about risk, and societal understanding of AI and risk.

Can war films teach us anything about war?

Can courtroom dramas teach us anything about the law?

Can watching roadrunner cartoons teach us about buying goods from the ACME corp?

Several authors have tried to modify the detective genre by inclusion of the mobile phone. They have to resort to Deus ex Machina outcomes like "oh no the battery's dead" or "damn, I'm out of range" because in most normal situations, phoning it in, does in fact short circuit the problem of partial knowledge and keeping others informed.

Likewise DNA testing. It both raises and dashes hopes of how to "prove crime" because the probative value of DNA evidence is contextual. When 90%+ of violent crime happens from within the family, what does it help, to show DNA evidence associates the family members?

"I'm sorry Dave, I can't do that" at least was partially informed by Minsky but thats almost laughable as a statement of "AI informed from movies informed from AI research"

the androids of Bladerunner had interior desires. The androids of Aliens.. it's more circumspect. The robots of Bicentennial and AI Artificial Intelligence are entirely about the search for "personhood" in an autonomous, artificial person. I argue that the narratives inherent in a story which is designed to beg the question cannot really answer the question, because it's presumed as an axiom: If the robots were not already autonomously independently conscious, then their desire is not comprehensible.

About one thing, I think we can be confident: the bootup power sequence for the first autonomous rubbish robot better be the "Mac" chimes or Pixar is dead to me.

yes because these films will constitute a small part of the self-image of the multimodal LLMs that have trained on them