534
New Orleans will use AI to answer 911 calls instead of a human (www.shreveporttimes.com)
submitted 6 days ago by to c/technology@lemmy.world
221 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 4 points 5 days ago (1 child)

No, I meant I read the article, sorry if that wasn't clear. You were bitching about people not reading it, I did.

We don't have to speculate, these same systems are being used as call agents and front line contacts for many industries now. Having dealt with these agents firsthand I can tell you they are extremely limited and pretty much garbage unless you have an issue that could easily be answered by looking at their web page.

There is no possible way to currently reduce the error rate, stop GIGO issues, audit the decision making process, stop hallucinations, etc. To even pretend these are ready for emergency situations is such a gross lack of discernment that I would consider it criminal in nature.

It is problematic to even provide the most basic service of screening of phone calls considering it could be costing previous time in an emergency. If it is a manpower issue then they should be fixing it there and not lookimg for a quick fix.

Also, you know it will not stop there. Their goal is triage so you know they are eventually moving towards far more than a simple public service emergency announcement. You will not be talking to a human in the future but a LLM that will kill people when it makes mistakes and won't be held accountable.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 0 points 5 days ago (1 child)

    No, I meant I read the article, sorry if that wasn’t clear.

    LOL. English ambiguity strikes again. Sorry for the misfire, totally my fault. Thanks for the kind reply and patience with me <3

    I do disagree that there's no way to make it work pretty safe. Instead of defaulting to holding someone in the AI loop until they talk their way out of it, you can simply set it up to stay in the loop only if it's sure it needs to continue, i.e. anything other than something it gets like "Oh, yes, that's what I was calling about", it should dump them to a human.

    And we don't know if they've set it up something like that or not. I humbly disagree that we know for sure they have not, no matter how other systems are configured.

    And I also disagree that it will necessarily turn out poorly. It literally depends on how they do it. But that said, it is analgous to taking humans out of the driver's seat in the car. At some point, it will be statistically ten times safer for automated driving to drive you than letting people drive. And when we get to that point, people will be up in arms that automated driving is killing people - when it is literally only a tenth of the previous rate of death. But we humans don't like giving up that sort of control.

    AI is new technology. I'm not totally behind them doing this, but I see the goal. And if they configure it correctly to fail-to-human rather than fail-to-AI-loop, I can see how this can save lives, not lose them. Of course if you're right and they screw it up it could be worse. I guess we'll see with time what they do and how it works.

    But I firmly believe that trying to fight against all the tech is the wrong move. It's like trying to fight to keep woodchoppers employed. The tech is there and much of what it is doing is inevitable. What we should be focused on is solving wealth inequality, solving our oligarch problem, solving the lack of social safety nets, living wages, our government working for all of us, not just bilionaires.

    So you and I probably aren't going to see eye to eye on this, but I can respect you better now that I know what you were saying better. :) Sorry to have flubbed that!

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 0 points 5 days ago* (last edited 3 days ago) (1 child)

    Pretty safe? I am sorry but if it is your life on the line or a loved one pretty safe is just not going to cut it. When the AI decides your dying grandmother isn't priority because of a reason no one can explain or audit what then.This is a huge safety issue than you hand wave away because you are apparently enamored with AI.

    I am trying to keep an open mind but replacing human thinking in emergency situations is not a reasonable application. You want a LLM model to triage your medical care with an unproven model that can and does make serious mistakes?

    https://www.foley.com/p/102ixtc/stanford-study-finds-high-percentage-of-errors-using-large-language-models-in-leg/

    This is not speculation. These companies are selling snake oil at this point and you act as their carnival barker. I do think that decades in the future these systems will be ready for these challenges, but I am not insane enough to believe this roll out is anything other than a cash grab at this point.

    I find your statement that humans don't like giving up control very disturbing. We obviously do give up control for routine low level things like GPS everyday. This is not what your are alluding to though. You are advocating for giving up a higher level of control to an algorithm that can't be held accountable.

    Maybe you are ready for a world of half baked customer agents, managers, counselors, architects, artists, doctors, teachers, judges, emergency dispatchers, etc but I am not. Ceding our control to a corporatocracy of profit driven yes men is not going to end well.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 5 days ago (1 child)

    I'll keep it simple and based on the article in question:

    Right now, people who need help are waiting longer to talk to an operator becauase operators field lots of calls about large incidents (or well-witnessed incidents).

    That's causing harm RIGHT NOW.

    The theory is that if you offer callers a way to NOT TAKE UP OPERATOR TIME for incidents that are known, then the people who need help talk to an operator faster and get help faster.

    That is what the article presents as what they're doing.

    Everything you are saying is pure speculation.

    You may well be right. We'll see.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 4 days ago (1 child)

    There are plenty of ways to mitigate this problem which do not require a AI solution. I find it insulting that you think it is the best solution considering their is no evidence of its effectiveness. In fact, I have presented why there are many reasons for concern.

    The willingness for government to invest in unproven technology rather than increase manpower or develop their own mass alert systems is dysfunctional to say the least. The fact that they often commit to these plans without consulting experts or asking for public comment is dishonest and corrupt.

    I have already had to deal with multiple people suffering from AI psychosis so I am not the biggest fan. You say real harm and AI is already doing it everyday. I see your retreat to reductionisism but I don't find it compelling in this instance. As I already alluded to, we know where this is going.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 0 points 4 days ago (1 child)

    I find it insulting that you think it is the best solution

    And I find you taking insult to be tiring.

    And where did I say it was the best solution? Do not put words in my mouth

    Go off an be unhappy in your little world of unhappiness.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •