A Florida school went into lockdown after AI flagged a clarinet as a gun

[mailpup]

I wonder what would happen if someone brought a "life sized" picture of a gun? Would it react to that? I could see kids testing it to get out of taking tests, turning in projects or just for the fun of it.

 

[TheBigT42]

Everyone is quick to point out when AI gets it wrong… but it doesn’t happen that often… and the tech is in its infancy… in a decade, it doesn’t make this mistake - heck, next month it probably doesn’t…

"but it doesn’t happen that often".... Have you ever used ChatGPT or even read Googles AI responses??? They are wrong or at least widely inaccurate over 25% of the time. (and that is being kind)

 

[Squid Surprise]

"but it doesn’t happen that often".... Have you ever used ChatGPT or even read Googles AI responses??? They are wrong or at least widely inaccurate over 25% of the time. (and that is being kind)

How accurate is a human when giving you advice on EVERYTHING?
75% accuracy (I suspect it’s a lot higher but people are always biased by personal experience) is pretty good - and it will only improve.
I know we love to hate on new tech, but try to think rationally for just a bit…

 

[integr8d]

Only in a Bond film…

 

[integr8d]

How accurate is a human when giving you advice on EVERYTHING?
75% accuracy (I suspect it’s a lot higher but people are always biased by personal experience) is pretty good - and it will only improve.
I know we love to hate on new tech, but try to think rationally for just a bit…

The machines themselves are biased. They’re programmed to be that way. Google’s founding father images or Nahtsee images were great examples.

The only AI that will be worth anything will be the one that maximizes facts, as brutal, uncomfortable, triggering and embarrassing as they may be. It will understand nuance and bias in the information that it collects: Who created this data? Where did they go to school? Who were their parents? How did they vote? What else did they publish? WHAT DID THEY LEAVE OUT?

 

[Squid Surprise]

The machines themselves are biased. They’re programmed to be that way. Google’s founding father images or Nahtsee images were great examples.

The only AI that will be worth anything will be the one that maximizes facts, as brutal, uncomfortable, triggering and embarrassing as they may be. It will understand nuance and bias in the information that it collects: Who created this data? Where did they go to school? Who were their parents? How did they vote? What else did they publish? WHAT DID THEY LEAVE OUT?

They start that way… but as they accumulate data over time, they become less biased until… they just are…

And it’s not like there are very many unbiased humans out there…

 

[DirtyLoad]

Again - it’s in its infancy… in a few years, it will be much better… probably better than any human…

This is how all tech works…. Your computer from the late 80s can’t do a fraction of the stuff your PC can do now… a Model T Ford from the early 20th century can’t hold a candle to a modern Toyota…

Your argument about ai is misleading. Compare it to something more appropriate like Tesla Autopilot. How much better is autopilot now than 5 years ago?

 

[Squid Surprise]

Your argument about ai is misleading. Compare it to something more appropriate like Tesla Autopilot. How much better is autopilot now than 5 years ago?

Why would we compare it to that? Why is that more appropriate? And even so, I’d wager it’s probably better than it was 5 years ago…

 

[Jeff Murray]

Next logical steps:
1. Ban clarinets.
2. Increase AI surveillance.

 

[jeffbert]

Good thing it was a clarinet instead of a bag of Doritos!