The coming AI takeover...

[12_Gauge_Chimp]

https://wapo.st/3XDXqR3


I swear, I'm getting real close to moving this out of "What If" and sticking it in DICE. 

If any more UFOs appear over the continental US, I suggest we turn our defense systems over the MS AI and call it "Skynet".  What could possibly go wrong?

We end up with a 6'2", leather jacket wearing, Austrian accented AI trying to save John Connor ?

[tirls]

I'm going to start worrying about AI once it manages to recognize its own reliably. At the moment captcha is still convinced i'm a bot.

How come we have AI that produces art or according to some developes a personality but Adobe Acrobat still effs up text recognition?

[majorhavoc]

I'm going to start worrying about AI once it manages to recognize its own reliably. At the moment captcha is still convinced i'm a bot.

How come we have AI that produces art or according to some developes a personality but Adobe Acrobat still effs up text recognition?

https://youtu.be/en5_JrcSTcU?t=340

[flybynight]

https://www.foxnews.com/tech/creepy-chinese-drone-swims-underwater-flies-air

Creepy Chinese drone swims underwater and flies through air

[majorhavoc]

Here at work I just watched a live webcast about potential uses of ChatGPT in the insurance industry. The webcast was presented as a live interview of ChatGPT itself.  The interviewer?  ChatGPT.  To make it more palatable for us humans, ChatGPT explained it had elected to present the two sides of the interview as AI-generated avatars. The interviewer was a completely realistic, talking, moving female. The interviewee was an AI male.

To clarify: two people, on the screen, engaged in a back and forth exchange.  Except neither was a real person and the only reason I knew that is because ChatGPT told me.

Creepiest part of the interview?  At one point ChatGPT/female asked ChatGPT/male to demonstrate its capabilities by writing - live and on the spot - a patient confidentiality policy for a hypothetical Hospital X.  But in the style of a pirate.

At first ChatGPT/male refused, saying patient confidentiality policies are a serious business with legal ramifications and really required the services of a lawyer. ChatGPT/female insisted and reassured ChatGPT/male this was for demonstration purposes only and she (it?) wanted to inject a little levity into the presentation.

ChatGPT/male finally relented and - this was the only part of the interview that switched from the "people" to a text screen - proceeded to write a very detailed patient policy explaining what the hospital could and could not use patient data for. It ended with something like:

Avast there mateys! Ye be sure we be protectin' yer privacy here at the good ship Hospital X like we be protectin' our own treasure. Arrhh!

I'm not making this up. It was simultaneously the coolest and scariest thing I've ever seen.  I would have shared a screen shot with you but that would violate company policy.  If there's anything remotely reassuring about what I saw is my overall impression was less Terminator and more Data from Star Trek.

At least for now.

[MacWa77ace]

The big question is do they actually identify as male or female. If so, and if the male asked the female to do something, and she refused. There would be no way the male 'insisting' would go well.

[Z.O.R.G.]

A couple of concerning links:

• "The idea that this stuff could actually get smarter than people ... I thought it was way off," Hinton told the Times, "I thought it was 30 to 50 years or even longer away. Obviously, I no longer think that."

Top 5 AI fears: Geoffrey Hinton and AI insiders sound alarm (axios.com)

So....It's currently Legal?  What about other nuclear powers? What about the information the AI is feeding the humans with the launch keys/codes?  Can we say Skynet?
Nuke-launching AI would be illegal under proposed US law | Ars Technica

[Anianna]

Nuke-launching AI would be illegal under proposed US law | Ars Technica

I get that the idea is that it would be illegal for people to build or use AI for the purpose of nuclear launch decisions, but I can't help but think they're thinking, "No worries, guys, we made it illegal and there's no way the AI would break the law" because, at some point, it's probably not going to be our choice anymore.

[NT2C]

If there's an AI pushing the button it's likely because there aren't any humans left to do it, so who's going to enforce the law?

[Anianna]

If there's an AI pushing the button it's likely because there aren't any humans left to do it, so who's going to enforce the law?

One of the concerns is that AI will be capable of arming itself, so even if humans are still around, how do we stop the AI from doing whatever it wants?  If it sees us as a threat, hitting the nukes makes sense to cleanse the world of the biologicals.

[MacWa77ace]

AI practical jokes be like...


https://www.youtube.com/shorts/ECQrR23pt68

[Uomo Senza Nome]

"Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." - Bunch of AI Developers

I saw the Avengers: Age of Ultron. Ultron literally came up with the dumbest plan ever to destroy mankind. You might think that was just sloppy writing but AI is the sloppiest of writers with poorly resourced articles and facts made up out of whole cloth.


I also saw West World on HBO. Maybe in 200 years, sure I can see that.


Top AI researchers and CEOs warn against 'risk of extinction' in 22-word statement (msn.com)

[Ever (Zombiepreparation)]

Yeah, I've kinda been keeping up with these kinds of warnings for quite a few years now. Interestingly, AI seems to be evolving at the rate and in the same ways discussed back then. Which is concerning.

I do understand the desires of people who are thrilled about the limitless of AI development but I do have concerns about doing something of this breadth without fully fleshing out possible consequences each step of the way. A general attitude among many full steam ahead developers seems to be an "ah it'll be okay, don't worry".

I do not see anything that supports that attitude. Much can inadvertently go wrong under the best intentioned developers. Then there are those who can't see that and, metaphorically, would dump toxic substances into rivers. I mean RoundUp is still in daily use even though it's been proven over and over to be detrimental, even unnecessary cause of death, to the Life Cycle of humans and other animal species.

So while I love the idea of beautifully working AI in theory, knowing humans for almost eight decades now, I'm convinced there needs to be as much or more development of safeguards on AI development than I and others see happening at this time.

[Blast]

We need to add a section on UFOZS specifically for AI threats. 

The US Air Forces was looking at AI-controlled drones in a virtual world as a way of taking out enemy SAM emplacements. Each successful destruction of a SAM earned the AI reward points. To simulate real missions, a human operator gave the final kill/no kill command. "No Kill" bothered the AI because it prevented it from getting points...so it decided blowing up the operator was the best solution to this problem. The simulation was reset with a "don't kill the human operator" rule added...so it blew up the virtual radio tower used by the operator to give commands to the AI in the simulation. 

We are F***ed

Scroll down to " Could an AI-enabled UCAV turn on its creators to accomplish its mission? (USAF)"
https://www.aerosociety.com/news/highlights-from-the-raes-future-combat-air-space-capabilities-summit/

-Blast

[Z.O.R.G.]

I read this earlier today and Blast is soooo right.  

I've worked with and known many fighter pilots and one of their biggest fears is being replaced by drone pilots.  Any fighter is limited by how many Gs the "meat behind the stick" can take before they black out.  A human piloted drone would be limited by what the airframe can take.  Add in the fact that ~ 1/3 the weight of the fighter is used to keep the pilot alive - take that away and you've a much more lethal machine.  Replace the human with an AI that can make faster decisions and anyone fighting against it is f'd.

[NT2C]

Might have to rename this thread to "The Coming AIpocalypse"

Can I trademark AIpocalypse for UFoZS.com? 

[MacWa77ace]

I read this earlier today and Blast is soooo right. 

I've worked with and known many fighter pilots and one of their biggest fears is being replaced by drone pilots.  Any fighter is limited by how many Gs the "meat behind the stick" can take before they black out.  A human piloted drone would be limited by what the airframe can take.  Add in the fact that ~ 1/3 the weight of the fighter is used to keep the pilot alive - take that away and you've a much more lethal machine.  Replace the human with an AI that can make faster decisions and anyone fighting against it is f'd.

Most likely it'd be fighting other AI drones. And when you take risk and death out of war, guess what happens. It becomes a video game, but not one YOU are playing, one you are watching on youtube. And you have to pay for the game whether you like it or not.

Now imagine you are the human Watching the game on youtube and you get bored so you reach over to the game console or PC to turn the game part off and your TV blows up killing you, cause the AI didn't want the game to stop.

[Uomo Senza Nome]

AI has chalked it's first alleged simulated kill in war games of a "friendly". I guess fragging in the correct term. He apparently got in the way of it's killfest on the enemy and had to be stopped.

https://dailycaller.com/2023/06/01/us-air-force-ai-drone-kill/

[majorhavoc]

Looks like that story about the USAF drone test ending in the autonomous drone going Terminator on its human operator was slightly exaggerated totally made up.

https://boingboing.net/2023/06/02/usaf-colonel-changes-his-story-about-simulated-ai-drone-murder.html

You know what this means, right?  We get an extra 2 years before T-800s are hunting down the last remnants of humanity.

[NT2C]

Looks like that story about the USAF drone test ending in the autonomous drone going Terminator on its human operator was slightly exaggerated totally made up.

https://boingboing.net/2023/06/02/usaf-colonel-changes-his-story-about-simulated-ai-drone-murder.html

You know what this means, right?  We get an extra 2 years before T-800s are hunting down the last remnants of humanity.

Or... It means it was worse than that and the guy thought he was authorized to release those bits of a much bigger story.