Do You Think We'll See AI 911 Operators Soon?

[Wes]

If you've ever called Fedex, you'd know that the phone is answered by a robot who says, something like, "Thank you for calling Fedex, how can I help you today?" and if things work out right, you can schedule a pickup without ever talking to a person. This morning I was reading a story in the Atlanta Journal-Constitution about how there's callers left waiting in Atlanta after calling 911. The industry standard is call should be answered in 20 seconds, but many were not. The article mentioned long training times for employees, a shortage of personnel, and other factors. I thought: I wonder if someone is going to try to make an AI 911 Operator?

In theory, this is something an AI could be good at.

• It's limited in scope
• It can respond immediately. Unlike human operators who can handle only one call at a time, AI systems can respond to multiple calls simultaneously, potentially reducing wait times significantly.
• It's always available: AI doesn't tire, ensuring consistent service availability around the clock without the need for shifts or breaks, which could be crucial during disasters when call volumes spike.
• It could can communicate in multiple languages and adapt to various communication needs, including text-based interactions for the hearing impaired, enhancing accessibility for a broader segment of the population.
• It could also do a sort of management of calls to dispatch units more efficiently (maybe)
• It's probably better at remembering details than people.
• Saves money

But there's also some big issues:

• Lack of Human Empathy: AI lacks the human touch, which can be crucial during distressing situations. The empathy, reassurance, and immediate understanding offered by human operators can be vital for callers in crisis.
• AI might struggle with understanding and appropriately responding to complex, nuanced situations that require human judgment, intuition, and experience.
• Privacy, especially if the AI is outsourced to some business, which it probably would be. The integration of AI into emergency services raises significant privacy concerns, especially regarding data collection, storage, and usage. Ensuring the protection of sensitive information would be paramount.
• Technical Failures and Limitations: Dependence on technology introduces risks of system failures, bugs, or limitations in the AI's programming that could lead to inadequate responses or failure to understand the severity of a situation. Basically if this was a thing it would have be done really well and pass the call to a human if it ran into issues
• Public Trust and Acceptance: Gaining public trust in an AI system to handle life-and-death situations could be challenging. People might be skeptical about an AI's ability to understand and appropriately respond to emergencies.
• Could eliminate a person's job

Do you think we'll see some local governments trying to actually do this in the next 10-20 years?

 

[Sunny D]

Trying? Sure. Probably be a while before its anything more than a trial though.

Main issue I see is that in genuine 911 emergencies, pretty much every call has an outlier. The caller might be panicked, in a threatening situation, there might be cues in the background the AI system hasn't been trained to recognise. I think a good dispatcher will always be better than an AI.

 

[demibear]

They already are attempting it since government employees in charge aren't the most technical individuals (buzzwords, buzzwords, buzzzzzz). There was a story of NYC rolling out an "AI" to help navigate people through the bureaucratic nightmare they have. But it has a 10% chance of giving completely wrong information like "You are permitting to throw trash in the wrong colored trash bag." and they decided it was good enough (they already spend money on God knows what, so it wasn't money behind that decision). The devil in me says they built in a guarantee way to collect fines and fees from people since I doubt a judge there will let the excuse "well the AI you made and said can be treated as legal advice said this was okay" work.

Maybe when AI achieve animal "general intelligence" we might actually have functioning virtual assistants that could do this kind of thing. But for now, they are toys/tools at best and enablers for lazy people unable to realize they are being fed something wrong on the worst end.

 

[Charmaylarg Dufrain]

My big fear at current is something like AI 911 response being a thing. As it stands a lot of politicians and the like don't understand a lot of technology and so it only gets regulated when something like tik-tok or the like comes to their attention.

So when greed comes into play and people can save money and also make more, then you get things like major car companies putting AI-salesmen who gave out free trucks, chatbots in airports that send air marshals after you even if you don't use it because they hit some random checkbox on you, being used by lawyers and using made up information, and such scenarios.

A program would likely be made first for a texting-911 feature. And would be really shady. One where you actually call and a robot-operator tried to handle the call? Absolute disaster. People sounding hysetical or with thick regional accents will confuse it 9/10 times. Static or music in the background, it making assumptions by some checklist.

Im pretty sure we can expect it. But it's going to be bad.

 

[raz]

Real 911 operators get a lot wrong. Once deep learning models can be programmed and debugged for accuracy by the humans smart enough to do so, then I look forward to the day some robot can send proper help with greater efficiency than most human 911 operators do.

 

[Sunny D]

Real 911 operators get a lot wrong. Once deep learning models can be programmed and debugged for accuracy by the humans smart enough to do so, then I look forward to the day some robot can send proper help with greater efficiency than most human 911 operators do.

I think we're still ages away from DLMs being sophisticated enough to deal directly with reality.