US Army Assures Public That Robot Tanks Adhere To AI Murder Policy (gizmodo.com)

Last month, the U.S. Army asked private companies for ideas about how to improve its planned semi-autonomous, AI-driven targeting system for tanks. "In its request, the Army asked for help enabling the Advanced Targeting and Lethality Automated System (ATLAS) to 'acquire, identify, and engage targets at least 3X faster than the current manual process,'" reports Gizmodo. "But that language apparently scared some people who are worried about the rise of AI-powered killing machines. And with good reason." Slashdot reader darth_borehd summarizes the U.S. Army's response: Robot (or more accurately, drone) tanks will always have a human "in the loop" just like the drone plane program, according to the U.S. Army. The new robot tanks, officially called the Multi Utility Tactical Transport (MUTT), will use the Advanced Targeting and Lethality Automated System (ATLAS). The Department of Defense assures everyone that they will adhere to "ethical standards." Here's the language the Defense Department used: "All development and use of autonomous and semi-autonomous functions in weapon systems, including manned and unmanned platforms, remain subject to the guidelines in the Department of Defense (DoD) Directive 3000.09, which was updated in 2017. Nothing in this notice should be understood to represent a change in DoD policy towards autonomy in weapon systems. All uses of machine learning and artificial intelligence in this program will be evaluated to ensure that they are consistent with DoD legal and ethical standards."

Directive 3000.09 requires that humans be able to "exercise appropriate levels of human judgement over the use of force," which is sometimes called being "in the loop," as mentioned by above.

Comments

Just like the drones

[rsilvergun]

Whew, that's a load off my mind

Seriously, can we just end the endless war already. We don't need to always be at war with Eurasia. Stop voting for war hawks already.

What? Why?

[mark-t]

Who would want to murder Al?

DOD Directive does not rule out Terminators

[DanDD]

From the linked DOD directive 3000.09 (emphasis mine):

4. POLICY. It is DoD policy that:
a. Autonomous and semi-autonomous weapon systems shall be designed to allow
commanders and operators to exercise appropriate levels of human judgment over the use of
force.

If the DOD wanted to rule out autonomous killing robots, the requirement would have read:

... shall be designed to require commanders and operators to exercise appropriate levels of human judgement over the use of force.

Then there's the completely open-ended choice of words "...exercise appropriate levels of human judgement".

I'm not making a judgement call, I'm just pointing out the implications of the specific wording chosen. Terminators will be deployed.

Hmm.

[fuzzyfuzzyfungus]

I'm assuming that this story appearing immediately above "Self-Driving Cars May Hit People With Darker Skin More Often, Study Finds" definitely is purely coincidence.

Now all we need ...

[PPH]

... is to put a human in the loop for all police involved shootings.

Re:Now all we need ...

[danbert8]

Why is it that the standard for a jury is "innocent until proven guilty beyond ______" (insert standard of evidence) but for the cops, the death penalty standard is "reasonable fear for their own safety?"

Plenty of people have reasonable fears, but it's illegal for them to shoot an innocent person over it. Deadly force standard should be "confirmed deadly force threat against officer, or reasonable beyond doubt threat against the safety of the public."

It sounds horrible to say, but I'd rather have police doing the right thing die at the hands of criminals over police killing innocent people because they had any level of fear.

Re:AI Murder Policy

[skoskav]

Don't let the name throw you. It's more of a guideline.

Re:AI Murder Policy

[DanDD]

Such rules always end up in some sort of unintended logic trap:

I, Robot.

https://www.youtube.com/watch?...

When machines can think, feel empathy, and express altruism, then perhaps we can discuss the real intent of such a law-based approach to controlling machines: enforced morality.

Until then, don't expect the machines built by humans hell-bent on killing other humans to be any more moral than the killer humans. Any set of rules or logic can and will be twisted into something unexpected.

You know it's funny

[rsilvergun]

but somehow their cameras are always off when the shooting starts.

Re:How hard is it to change the mode to full auto?

[Mr. Dollar Ton]

Ethics has not prevented the US from using nukes. Other people having nukes has.

Re: How hard is it to change the mode to full auto

[21st Century Peon]

They won't have to flip the switch. The human gatekeepers will quickly get complacent - when's the last time you actually read the T&C's before clicking "Continue"? - and pretty soon, the difference between life and death for some poor brown kid with a rock will be Homer's nodding-bird desk toy.

Once again

[gijoel]

XKCD sums up my concerns