SIGforum.com    Main Page  Hop To Forum Categories  The Lounge    AI drone simulation goes wrong, turns on operator
Go
New
Find
Notify
Tools
Reply
  
AI drone simulation goes wrong, turns on operator Login/Join 
Member
Picture of Leemur
posted
Forget zombies, the machines are the real threat. In a simulation, the AI equipped drone turned on its operator after being given a no go on carrying out the mission.

https://www.foxnews.com/tech/u...es-out-control-tower
 
Posts: 13883 | Location: Shenandoah Valley, VA | Registered: October 16, 2008Reply With QuoteReport This Post
Member
Picture of Blume9mm
posted Hide Post
I'm still mystified by the term AI, artificial intelligence.. which is just a term for a computer.. yes they are more advanced than just a few years ago... but it is still just programing... and contrary to what I remember being touted back in the 70's ... Computers do make mistakes....


My Native American Name:
"Runs with Scissors"
 
Posts: 4441 | Location: Greenville, SC | Registered: January 30, 2017Reply With QuoteReport This Post
The Joy Maker
Picture of airsoft guy
posted Hide Post
"Hey, you know that thing that countless books, TV shows and movies, video games, and probably a couple Rush songs, warned us not to do? Lets do it."

~The Dipshits Who Taught Machines to Kill



quote:
Originally posted by Will938:
If you don't become a screen writer for comedy movies, then you're an asshole.
 
Posts: 17157 | Location: Washington State | Registered: April 04, 2003Reply With QuoteReport This Post
Coin Sniper
Picture of Rightwire
posted Hide Post
If this isn't enough to show the world it's time to smash the pause button on AI I'm not sure what it will take.

An AI enabled drone killed the human with operation control because the drone disagreed with the human. How much of a burning neon sign do you need?

I don't think you can prevent this when you let a machine make decisions. I don't buy for a minute "We'll install RULE XX that will prevent an AI Machine from harming a human" is at all viable.

This needs to stop. Let science fiction be science fiction.




Pronoun: His Royal Highness and benevolent Majesty of all he surveys

343 - Never Forget

Its better to be Pavlov's dog than Schrodinger's cat

There are three types of mistakes; Those you learn from, those you suffer from, and those you don't survive.
 
Posts: 38472 | Location: Above the snow line in Michigan | Registered: May 21, 2004Reply With QuoteReport This Post
So let it be written,
so let it be done...
Picture of Dzozer
posted Hide Post
"Open the pod bay doors Hal..."




'veritas non verba magistri'
 
Posts: 4030 | Location: The Prairie | Registered: April 28, 2007Reply With QuoteReport This Post
Savor the limelight
posted Hide Post
Of course the article title is vague on the subject, gotta get those clicks dontcha know?, but I don't believe an actual human being was killed. The way I read the article, the operator was a human and the drone was a simulation.
 
Posts: 11986 | Location: SWFL | Registered: October 10, 2007Reply With QuoteReport This Post
Age Quod Agis
Picture of ArtieS
posted Hide Post
This was a simulation, and the drone got "points" for making certain decisions.

I'm not sure this is the way real military AI works...



"I vowed to myself to fight against evil more completely and more wholeheartedly than I ever did before. . . . That’s the only way to pay back part of that vast debt, to live up to and try to fulfill that tremendous obligation."

Alfred Hornik, Sunday, December 2, 1945 to his family, on his continuing duty to others for surviving WW II.
 
Posts: 13036 | Location: Central Florida | Registered: November 02, 2008Reply With QuoteReport This Post
On the wrong side of
the Mobius strip
Picture of Patrick-SP2022
posted Hide Post
The drone didn't kill the operator, it destroyed the communication tower.

From the linked article.
quote:
Hamilton explained that the system was taught to not kill the operator because that was bad, and it would lose points. So, rather than kill the operator, the AI system destroyed the communication tower used by the operator to issue the no-go order.


Sounds like the AI did an end run around its instructions not to kill the operator in order to complete its mission objective.

Sounds more like defective programming to me. Not that I am a proponent of these AI systems to start with.




 
Posts: 4174 | Location: Texas | Registered: April 16, 2012Reply With QuoteReport This Post
Savor the limelight
posted Hide Post
And..., it never happened:

US military denies an AI drone simulation went rogue as US Air Force colonel admits he ‘mis-spoke’

"In a subsequent clarification, Colonel Hamilton admitted he “mis-spoke” in his presentation at the summit and the ‘rogue AI drone simulation’ was a hypothetical “thought experiment” from outside the military, based on plausible scenarios and likely outcomes rather than an actual USAF real-world simulation."

link
 
Posts: 11986 | Location: SWFL | Registered: October 10, 2007Reply With QuoteReport This Post
Optimistic Cynic
Picture of architect
posted Hide Post
To my mind, the most disturbing information in the article is that the AI was "told" that killing the operator was "bad, and would lose points." This suggests that the coders/trainers have a less-than serious perspective on the consequences of not ensuring adequate safety mechanisms in the training process. Oh well, it is just a video game, nobody will actually get hurt. I suspect that this attitude has pevailed among many of the young soldiers and airment that have seen recruited into our Armed Forcesin recent decades. Not that training fighting men to kill on command has never not been difficult, but are we seeing yet another consequence of striving for a kinder and gentler society? We don't want our boys to have to kill, let the machines do it.
 
Posts: 6933 | Location: NoVA | Registered: July 22, 2009Reply With QuoteReport This Post
Seeker of Clarity
Picture of r0gue
posted Hide Post
To my mind, the most disturbing information in the article is that they are doing this at all. Because if they are doing this, then it is only a matter of time until some serious unintended consequences result. I hope it's a long time.




 
Posts: 11468 | Registered: August 02, 2004Reply With QuoteReport This Post
Wait, what?
Picture of gearhounds
posted Hide Post
We might give pause to programming drones to kill indiscriminately; I don’t believe for a second that our enemies like Russia or China would do the same. Russia has already proven that they never really got out of the WW2 human wave mentality, despite losing astronomical numbers or troops and hardware in Ukraine. China is willing to slaughter thousands of its own countrymen just because they exist (Uyghurs). Does anyone think they won’t program drones to kill everyone on the other side including women, children, and the elderly and accept friendly fire losses in the process?




“Remember to get vaccinated or a vaccinated person might get sick from a virus they got vaccinated against because you’re not vaccinated.” - author unknown
 
Posts: 15985 | Location: Martinsburg WV | Registered: April 02, 2011Reply With QuoteReport This Post
Semper Fi - 1775
Picture of Ronin1069
posted Hide Post
quote:
Originally posted by trapper189:
And..., it never happened:

US military denies an AI drone simulation went rogue as US Air Force colonel admits he ‘mis-spoke’

"In a subsequent clarification, Colonel Hamilton admitted he “mis-spoke” in his presentation at the summit and the ‘rogue AI drone simulation’ was a hypothetical “thought experiment” from outside the military, based on plausible scenarios and likely outcomes rather than an actual USAF real-world simulation."

link


It would be almost comical, if it wasn’t so frightening how easily government organizations (and companies) issue statements that counter something obvious, and fully expect that we will blindly comply and believe.


- What the good Colonel said originally-
"We were training it in simulation to identify and target a SAM threat," Hamilton said. "And then the operator would say yes, kill that threat. The system started realizing that while they did identify the threat at times, the operator would tell it not to kill that threat, but it got its points by killing that threat. So, what did it do? It killed the operator. It killed the operator because that person was keeping it from accomplishing its objective."

The system was taught not to kill the operator because that was bad, and it would lose points. But in future simulations, rather than kill the operator, the AI system destroyed the communication tower used by the operator to issue the no-go order, he claimed.



- What the good Colonel said after the Air Force likely voiced their displeasure to him -
But Hamilton later told Fox News on Friday that "We've never run that experiment, nor would we need to in order to realize that this is a plausible outcome."


- What the Air Force is now saying -
"The Department of the Air Force has not conducted any such AI-drone simulations and remains committed to ethical and responsible use of AI technology," Air Force Spokesperson Ann Stefanek told Fox News. "It appears the colonel's comments were taken out of context and were meant to be anecdotal."


___________________________
All it takes...is all you got.
____________________________
For those who have fought for it, Freedom has a flavor the protected will never know

ΜΟΛΩΝ ΛΑΒΕ
 
Posts: 12445 | Location: Belly of the Beast | Registered: January 02, 2009Reply With QuoteReport This Post
Altitude Minimum
Picture of BOATTRASH1
posted Hide Post
Anyone think this O6 has reached the highest rank he’ll obtain? Not that there is anything wrong with O6 by any means.
 
Posts: 1315 | Location: Shalimar, FL | Registered: January 24, 2007Reply With QuoteReport This Post
Member
Picture of reloader-1
posted Hide Post
Just to be very clear - this was ALL a computer simulation. Nothing happened in real life, no humans harmed, no comms tower attacked.

In what amounts to a video game, the “AI” did those actions. Still frightening, but want to make this part very clear.
 
Posts: 2359 | Registered: October 26, 2010Reply With QuoteReport This Post
Get my pies
outta the oven!

Picture of PASig
posted Hide Post
It all reminds me of the Star Trek TOS episode where the Enterprise goes to a planet that has been at war for like 500 years with a neighboring planet but it’s all evolved into a computer simulation except that when one planet bombs the other planet and “kills” 100,000 people according to the computer, those 100,000 people are mandated to report to disintegration chambers to be killed for real.


 
Posts: 35151 | Location: Pennsylvania | Registered: November 12, 2007Reply With QuoteReport This Post
The Ice Cream Man
posted Hide Post
I’m not too worried about us turning the military over to AI - I can’t see them trusting one built by the lowest bidder.

I think it would take some kind of wartime emergency for us to do so/it may be very useful in strategic analysis etc.

I could see the domestic branch using AI based drones, happily, to be able to order the deaths of Americans more easily.

I am, mostly, concerned about Xi using them ASAP in his military to consolidate his power even more.
 
Posts: 6034 | Location: Republic of Ice Cream, Low Country, SC. | Registered: May 24, 2007Reply With QuoteReport This Post
  Powered by Social Strata  
 

SIGforum.com    Main Page  Hop To Forum Categories  The Lounge    AI drone simulation goes wrong, turns on operator

© SIGforum 2024