Allowing police officers to submit LLM-written reports reveals a remarkable misunderstanding of what LLMs do, a profound indifference to the notion of integrity in the communications of law enforcement with the justice system, or both.
Given how readily subject to suggestion human witnesses—including police officers—are known to be, this is a disaster.
Yes, police reports aren't always the most accurate, but introducing an additional layer of non-accountability is bad.
@ct_bergstrom As a defense lawyer, I would demand access to the source code (and lots of related discovery) for any LLM used to generate a police report. They will not like that at all. They won’t give it up, of course. But that then puts their case in serious danger of being dismissed. There are tons of other issues, e.g., who is the witness? The cop or the LLM or the LLM engineers?
The real problem is finding out that they used an LLM in the first place. Terrifying.
@D_J_Nathanson @ct_bergstrom You'd also want access to the raw training data including sources & amounts paid for copyrighted material
They won't like that at all
This will be to prove there's no bias in generated reports
E.G. if the training data doesn't include instances of officer error, brutality, law breaking etc...
Then the LLM can't generate a report which allows any of those interpretations
The officer will always be depicted as behaving in an exemplary manner even if they didn't
@D_J_Nathanson @ct_bergstrom grand juries should also demand answers to these questions, and not admit any LLM produced police report without these explanations.