This acronym stands for something other than what 2A watchers understand the term to mean.

https://www.un.org/disarmament/the-convention-on-certain-conventional-weapons/

The group is concerned with weapons of war that do not meet "kosher" standards. The war in Ukraine shows no evidence of such weapons (yet), but analysts predict the use of AI-driven weapons in the future.

https://www.yahoo.com/news/future-warfare-could-lot-more-110932720.html

The hazards of AI weapons were brought home last year when a U.N. Security Council report said a Turkish drone, the Kargu-2, appeared to have fired fully autonomously in the long-running Libyan civil war - potentially marking the first time on this planet a human being died entirely because a machine thought they should.

The U.S. probably has swarming drone technology that could attack a Russian S-400 anti-air missile system. Small explosives would blind the S-400's radar / sensors, rendering the system useless. Should the Russians start WW3, the Pentagon probably has Skynet weapons from the future which will be sent back in time to defeat Russia. Hail to Terminator bots.

In WW2, the Russians trained dogs to carry bombs to German tanks. They used look a like German tanks to do the training. The problem was that dogs are scent based and the fuel they used and the way it burned was from Russian engines.  So when they tried to apply the trained dogs, the dogs went after their own armor.

...
The hazards of AI weapons were brought home last year when a U.N. Security Council report said a Turkish drone, the Kargu-2, appeared to have fired fully autonomously in the long-running Libyan civil war - potentially marking the first time on this planet a human being died entirely because a machine thought they should.
...

That's BS -- 100% Grade A Prime BS.

The article is creating a boogeyman out of AI where no such danger exists in the UN's CCW text.

Any machine, no matter how simple or complex, makes "decisions" based on its design.  It can be as simple as a trigger you step on, a wire you trip, a rope you step in, or a remote switch you drive over.

Unless the drone that fired "autonomously" decided who to fire on and who to spare, it's no less indiscriminate than a shotgun mounted to a chair, aimed at the doorway and the trigger tied to the doorknob.

Funny. I go to the UN link posted, and it's all about mines, booby traps and other devices that indiscriminately kill (especially noncombatants), but nothing at all about autonomous machines or drones -- no AI boogeyman mentioned.

IMO, the only AI involved in the referenced drone strike would have been the flying, recognition of its PROGRAMMED target -- probably by GPS -- and its flight back to its base.  The release of weapons would have to have been programmed in ahead of time.  The only decision would have been identification of the objective, not the go/no go decision based on any other factors.  The drone would have been no different than a missile hitting a predefined target.  I don't see anyone unleashing a weapon like that to decide for itself if a candidate target is to be hit or not based on its own decision-making.

Any machine, no matter how simple or complex, makes "decisions" based on its design.  It can be as simple as a trigger you step on, a wire you trip, a rope you step in, or a remote switch you drive over.

That is not AI. Artificial Intelligence combined with autonomous decision making is what the research labs are working on. This is not like what you are see in movies, i.e., a remote pilot of a Predator drone using image recognition to target a tank that DSP has identified, with the pilot pushing the kill button. It's going to be different in the future. A JSF pilot will spot enemy ground fire and summon helper drones to auto target the enemy. The U.S. will hold off as long as necessary in giving the machine the final decision, but the Russians won't. What will be the tripping point? A probability threshold.

That is not AI. Artificial Intelligence combined with autonomous decision making is what the research labs are working on. This is not like what you are see in movies, i.e., a remote pilot of a Predator drone using image recognition to target a tank that DSP has identified, with the pilot pushing the kill button. It's going to be different in the future. A JSF pilot will spot enemy ground fire and summon helper drones to auto target the enemy. The U.S. will hold off as long as necessary in giving the machine the final decision, but the Russians won't. What will be the tripping point? A probability threshold.

I know what AI is, but that's not what I'm referring to.  I'm saying the writer's assertion that this is "marking the first time on this planet a human being died entirely because a machine thought they should" is BS.  Machines don't think.  They receive input, analyze the data, and follow a predesigned series of programmed logic paths to reach a predetermined action.

It's more complicated, but as I said, people have died because a machine decided to unleash lethal force due to external forces and the reaction it was designed to perform.  I don't see anything in your response that counters that description.  Image recognition?  Who made the decision to tie the outcome (attack/kill) with the image?  Not the AI/Drone.  A person gave the drone that image and kill order.  The AI simply managed to locate the target and carry out its programming.