It is exactly that. One may have a backtest. You'll see that real calls will be falsely detected.

What "modern AI" even means, you are arguing about taste of oysters with the people who have eaten them, unlike you.

Oh, of course you can't have a backtest with a proprietary centrally hosted model until it's deployed. Shit-shit-shit.

They have job openings that people simply aren’t applying for, it’s not a question of funding.

It is. Double the pay, see how many people there are. If still too few, double it again.

I'm certain it'll be less money wasted that on this bot done the lawful way, with proper compensations to victims and their families. We are not considering the situation where it's not.

They’re getting too many phone calls to handle and many of those phone calls should not be going to them in the first place. What should they do?

Hire more operators.

Contact centers have not been invented yesterday, it's just plain bullshit this simple task is somehow hard today, when it has already been simplified far beyond what people in year 1977 dreamed about.

It's the actual job of the government, BTW, and not playing Caesar with real armies or playing Master of Orion with real systems.

And next you lose them

Companies are already going away from such ideas..

One thing left unclear is how the determination is made about emergency versus non emergency.

If it's a separate number, ok, seems clear cut enough.

If it's human always answers and if it's some bullshit they just click a button to punt to AI instead of just hanging up, ok.

If they are saying the AI answers and does the triage and hands off immediately to a human when "emergency detected", then I could see how that promise could fail.

It's called capitalism. If you can't hire anyone, maybe the pay sucks.

And what examples of ai makes you think that it will be 100% accurate because that's what is needed in this situation. Not a single one of those calls can be lost in the shuffle.

The important thing is that they can tune this to attempt to hold false negatives constant while decreasing false positive rate.

A lot depends on the implementation rather than the idea itself. I've read plenty of stories of people stuck on hold with 9-1-1 - including deaths - as well as cases where they've been hung up on by shitty operators.

An AI system might be able to do some basic triage to prioritize calls for the human operators and actually result in faster access/response and saved lives. It might also be able to do things like transcribing information such as addresses or location for responders.
If the AI is planned to be a replacement for humans rather than an augmentation though, lives will likely be lost

I know, and maybe it will, my faith is just very low.

The opposite is extremely common too. People get on the phone and instantly go into raw panic mode and yell about 500 words at you before you've even had a chance to read your greeting. After putting down some choice words to control them a bit, you find out they found a bag of weed in their teenager's bag or their neighbour's playing music too loud.

I'm a dispatcher (not in the USA) and our managers start flipping out and running round like their heads are on fire if the wait time reaches 30 seconds. If there's more than 3 calls in the emergency queue then they sit down and take them themselves (If you've ever worked in any call centre at all, emergency or not, you'll know shit has to really hit the fan before management will consider doing this!)

Usually high queue time/numbers are just multiple calls for the same incident (think large RTC's or very public assaults/stabbings right in the middle of a heavily trafficked city centre) so we can get that queue down very quickly, especially as 99% of the time any call after the initial one will simply be "we're already aware and we've got crews en route, bye".

When the operator identifies the call as a non-emergency (which takes an absolute maximum of 2 minutes, even for very complicated calls), they simply say "please call the non-emergency line on XXX, thanks, bye". Why is the AI required?

I agree that people shouldn't be calling the emergency line with rubbish, but unfortunately they do, because the non-emergency line isn't as well publicized and even if they do know about it people think that "non-emergency" means "we can't be bothered dealing with it" and so calling the emergency line somehow means their issue will be taken more seriously.

Where I work we take both non and emergency calls, and have a separate number for each. The phone system we use will make sure the emergency calls come through first, so it's not uncommon to have zero emergency calls queueing while the non-emergency queue sits at 10-20 minutes (just like any other call centre, we have the boards up on the wall showing the stats).

It seems like this AI thing is trying to solve the problem of people calling the emergency number for a call that doesn't need an emergency response, which is super common. Either people don't know about the non-emergency line, or they think the non-emergency line is for other people and calling the emergency line will get their issue sorted faster. The first kind are usually very apologetic when you ask them to call the non-emergency, the second kind will argue with you and we're instructed to just hang up on them after repeating the instruction to keep the emergency queue free.

The thing is, anyone with half a brain can identify a non-emergency call within max 2 minutes. It's probably the easiest part of the whole job. But it definitely requires a human, because people will call up shouting and screaming like they're mid-way through getting stabbed, when really they're just a grumpy old fuck who's neighbours are playing rap music. And on the flip side, plenty of people are able to make a full-on emergency call in an almost spookily calm tone, and even more so if they're not directly involved (Common example is a teacher or social worker calling something in a child's disclosed to them about their parents). So being able to read between the lines in a way humans are very good at, but robots are not, is obviously super important.

no

maybe to have in cases where they are under to much load, such as a massive emergency where they get way more calls than they can handle.

as a backup only.

but even then it'll encourage them to have less personnel.

never had a conversation that didn't hallucinate every now and then

you're too concerned about those "consequences" but have you considered that they get to fire people as well and save money?

did you think of all the taxes they'll cut from the rich? no, you only think about yourself and what will happen to you in an emergency

Why is the AI required?

Because they don't have enough human operators to field all of the calls they're getting. If they did then they wouldn't be having to look into using AI to screen them.

This is in the article.

Without reducing headcount, right? Right?

That... doesn't answer my question at all. Why is the AI specifically required? How is it an improvement over making the job more attractive to humans and getting more of them to do the job instead?

They do want to hire more humans, there are job openings they've posted that are not being fulfilled. Since they're not being fulfilled and they don't have the money to increase their salaries to draw in more, they're having to look for ways to make the resources they do have stretch farther. Hence, AI screening to shunt the non-emergency calls away from their existing human emergency dispatchers.

If someone calls 911 how on earth do you know its a non-emergency before speaking with someone?