Eine Schule in Florida wurde abgeriegelt, nachdem AI eine Klarinette als Waffe gekennzeichnet hatte
https://www.techspot.com/news/110591-florida-school-went-lockdown-after-ai-flagged-clarinet.html
Eine Schule in Florida wurde abgeriegelt, nachdem AI eine Klarinette als Waffe gekennzeichnet hatte
https://www.techspot.com/news/110591-florida-school-went-lockdown-after-ai-flagged-clarinet.html
22 Kommentare
Similar situation in Baltimore earlier this year.
How does the school have the ability to drop $250K on that shit?
First doritos now clarinet. Are we sure its artificial intelligence and not artificial stupidity?
It’s amazing how much states like Florida will spend on stuff like this and how little time they’ll spend passing common sense gun control laws. This is putting all of the onus on school children to manage the consequences of adult political inaction.
“Out of an abundance of caution”
9/11 paving the way for the broad acceptance of the 1% doctrine really is long-game destabilization.
Well, it can be, on the right hands.
Before this type of AI security things are put. they have to be tested maybe? ….
This is the issue with treating AI like it’s some flawless means to improve security, instead of a tool that needs context and human judgment. These systems don’t understand reality, they understand shapes and probabilities, and modes they have been trained with…
. A long black object = gun, apparently. ???
If anything, this just proves we’re rolling this tech out way too fast in high-stakes environments. Schools aren’t beta testing labs. Add a human in the loop ….
the intent is good, but maaan….
This is just as much an issue with people as with AI. An interesting thing about AI (and computers in general) is that when a computer spits out a number, humans immediately accept it as valid and correct.
Take the global nemesis of every data person in IT – the one-off Excel spreadsheet created by a pompous moron. The business has accepted those numbers for years, yet when a report uses verified and validated SQL to get a result, it is different from the Excel report and middle management refuses to accept the results because for years they have been telling upper management the wrong numbers.
It’s the same for these AI systems. If an AI system reports an 100% match, it is immediately accepted. There is no doubt or even double-checking.
It also is ignorance not just by the general public, but especially by moronic, technically inept middle and upper management who don’t know the underpinning of how AI works and how easily AI can be mislead by bad training.
It’s great how the principal shifts the blame to the student for holding her clarinet like a gun (which is just so American) and passes off the effort to change things onto parents who are now told they must speak with their kids on the importance of not doing this. It’s not the child’s fault that they live in a surveillance state.
Image recognition is shit, it needs a clear image of the whole object. Ever been in a school DT classroom and seen how awful a lot of the work is? Now imagine a 2 year old trying to identify what each item is from 100ft away…
Hot Dog, Not Hotdog
Just two months ago, some kid was cuffed by the cops because AI flagged his pack of chips as a gun.
[https://www.bbc.com/news/articles/cgjdlx92lylo](https://www.bbc.com/news/articles/cgjdlx92lylo)
Remember to never vote on any official who supports these tools, and to make sure every candidate is aware of voters‘ preferences.
Is there human review anywhere in the process? I’d assume the AI would flag suspicious images, alert a human supervisor, and wait for them to confirm or reject things…
If I went to that school I would be trying to trick the Ai with everyday items to close down the school. I last time it was a bag of chips
better safe than sorry. thats just more training data for the ai anyways. we could also just ban guns but that makes too much sense.
So the AI initially flagged it, a human reviewed it and decided yes this is a gun & contacted police…? Jfc
Clarinet can be dangerous in the right hands.
A Florida school spent roughly $250k on an AI system that locked down a campus over a clarinet.
Meanwhile in Australia, similar AI age-estimation systems are already producing false positives and false negatives: locking out adults by classifying them as under 16 or 18, including cases where valid government ID is presented and still rejected, while also misclassifying under-16s as 16–17 or, in some cases, over 18.
I mean, I suppose this is technology news, but this seems like a non-issue to me.
The entire point of detection systems like this is to identify things that *might* be dangerous. No system is going to be 100% accurate – even humans are not. If the system is inaccurate, you want it to be overly sensitive because a false positive on a gun is way less dangerous than a false negative. So the AI system was basically like „hey, I can’t identify this thing, might be a weapon.“ The school did exactly what it should do in an uncertain situation involving a possible weapon: they prioritized keeping the kids safe, then they identified that the item was not a weapon and resumed normal operations.
I know we love to hate on AI but this is also a country where school shootings are so routine that kids do active shooter drills in school. Why would the school not err on the side of caution versus potentially having another Uvalde?
I mean, you can say that the money would be better spent doing something else, or have an issue with the constant surveillance of humans, and those are legitimate issues but I feel like are slightly separate from the question of whether the AI is 100% accurate. it won’t be. Nothing will ever be.
Like any AI, a human needs to review it first. Every time. That should be what these AI surveillance systems do. Spot something suspicious to show a human, who then should be the one to decide what to do. The school should send their emergency response bill to the surveillance company.
This one time ,
In band camp…
I’ve personally witnessed clarinets being used as sonic weapons. But that wasn’t really an intentional use.