Oh, how long until an AI ends up swatting people. There will probably even be different categories:
Legit swat scenario, wrong address (or otherwise incorrect details make it turn tragic)
Caller deliberately trying to swat innocent victim, AI goes along with it and adds credibility (not that humans seem to take away much credibility, so maybe this one shouldn’t really be a black mark for AIs if it does happen)
AI just misinterprets a caller’s description, deploys swat team
AI doesn’t trust the caller’s description, deploys swat team (anyone ever seen an agent start suspecting they were in an adversarial situation where something was injecting things into the context? I have)
AI uses dispatching tool incorrectly, accidentally deploys swat team
Oh, how long until an AI ends up swatting people. There will probably even be different categories:
Or just as likely never dispatches in the case of a real emergency. I have a hard time thinking of what could go right.