Current Events > Computer Generated Swatting Service Is Causing Havoc Across America

(+) Yeah! (+)

https://www.vice.com/en/article/k7z8be/torswats-computer-generated-ai-voice-swatting

Well fuck

People sure do suck

MC_BatCommander posted...

People sure do suck

Seriously. Who the fuck thinks up this kind of shit.

tag

Tbh, this seems like an issue that should've been addressed a decade ago, but we refused to place resources where we need them.

This is literally the kind of shit that caused OpenAI to deem GPT-3 'too dangerous' in the first place and put it behind a vault and censor/filter it to hell. This was years ago.

The idea was that the model was too good at generating genuine looking messages/text that it was basically perfect for scammers and other malicious intent. So, good job to whoever decided to be a maroon in a similiar fashion.

The real problems here are:

1. That it takes so little evidence to get a bunch of cops to burst into your house with guns.
2. That our phone system is so broken with call spoofing that shouldn't be possible in the first place.

Luckily the FCC is doing something about the call spoofing problem.

https://www.fcc.gov/call-authentication

Purple_Cheetah posted...

This is literally the kind of shit that caused OpenAI to deem GPT-3 'too dangerous' in the first place and put it behind a vault and censor/filter it to hell. This was years ago.

The idea was that the model was too good at generating genuine looking messages/text that it was basically perfect for scammers and other malicious intent. So, good job to whoever decided to be a maroon in a similiar fashion.

The article makes it sound like all the actual text/script/thinking is being done by a real person and the AI part is just voice synthesis. Though they could easily be wrong about this since it's not like they're gonna go ask him obviously

I saw this. Its being used a lot by people online against their enemies more and more.

There are a lot of unhinged people out there that are more than willing to go to lengths like this to get rid of people they dont like. Its kinda scary.

ChocoboMogALT posted...

Tbh, this seems like an issue that should've been addressed a decade ago, but we refused to place resources where we need them.

I mean we have senators so old and out of touch that they still struggle to extract pdfs from an email.

kirbymuncher posted...

The article makes it sound like all the actual text/script/thinking is being done by a real person and the AI part is just voice synthesis. Though they could easily be wrong about this since it's not like they're gonna go ask him obviously

This is a bit late, but I wasn't on earlier.

Back in about 2019/2020 OpenAI open/crowd sourced GPT-3, the 175 billion parameter text model. It was originally supposed to be more publicly available since it was more of a collab. However, they deemed it too dangerous because the sentence and word structure was far too realistic/proper that it'd allow anyone to come up with genuine looking scam messages/text. Not to mention fuel disinformation or other illegal material. So they sealed it away and only allowed it to be used for 'wholesome' content like ad slogans or something very very watered down. Not to mention it cost a lot.

The absolute closest the general population ever got to experience it was through AIDungeon's original dragon model back from 2020-early 2021. It was very impressive, unfortunately due to the nature of the model and some quirks it was literally generating hate speech, racism, genocide, violent/sexual content with minors/animals, and other very questionable material on its own. You didn't have to trigger it, despite their claims that it was the users causing it since that's not how it worked.

Doesn't take long to figure out why they yanked that away after much controversy. Similiar material is no doubt deep within ChatGPT, but it's heavily lobotomized/censored.

This is what I was refering to when I said OpenAI deemed even text generation too dangerous years ago, as AI can easily be used for malicious intent especially when it starts to grow so rapidly.

*
What's even worse is that apparently GPT-4 is even more impressive, making that already highly advanced GPT-3 apparently seem dumb. I can't even imagine, considering how capable it was already. This is also ignoring the voice/image aspect, which is a whole new can of worms. Though it's similiar in that it can be abused.

This happened to my kids school. Called in a shots fired.

DarthDemented posted...

This happened to my kids school. Called in a shots fired.

That's awful.