I think AI drones make us more willing to kill indiscriminately. When our people are also at risk, we pay more attention to whether there is sufficient information to safely operate with minimal casualties on either side. Drones invite "spray-and-pray" warfare that breeds further aggression.

They definitely have value, drones would have been great at D-Day, but realistically considering human behavior... I suspect these "discriminate" weapons will result in a net increase of collateral damage.

I'm not sure that I buy your first paragraph. When our people are also at risk, there is more incentive to shoot first rather than risk being shot. When our people are not at risk, we can take the time to be sure, because our downside is smaller.

Nobody's talking about AI drones, or anything close to that.

The proposal was better image recognition algos for surveillance drones. https://dod.defense.gov/News/Article/Article/1254719/project...

Let's take a step down the slippery slope and assume that in the future it was repurposed for manually triggered weapons drones. Ever see a story about how the US drone-bombed a wedding or a goat farmer? Maybe better (or any) image recognition could have saved innocent lives.

It's not that clear cut.

You aren't thinking about AI drones, other people are. Project Maven was a first step towards securing broader DoD contracts. Pretending drones were not part of the picture is either naive or disingenuous.

You can't do any better than slippery slope? Who's being disengenuous?

There are several rubicons between here and AI killing machines. Why draw this particular line? Why not just halt all AI research?

If I said that banning bump stocks was "basically gun confiscation" or that implementing universal healthcare was "basically the great leap forward", would you be convinced?

I'm really trying to understand your argument here, please help me out.

What makes you think these weapons will be discriminate? Who will be deploying them and against what will they deploy them? Drones are dual use technologies, the user matters. AI is also a dual use technology, the trainer matters. AI (poorly defined term almost always, but whatever) can be just as evil as a human, if not more so. It's hard enough to figure out what's a moral right vs. wrong as a human, I've yet to see a computer do that, and if it did, I'd argue it's just as easy to tell it to always choose wrong vs to always choose right.

Um, use your imagination a little. This has the potential to make murder-for-hire and assassinations much easier. The body count may be lower but everyone is at risk. It won't just be US government drones and Russian poisonings anymore.

This is likely to require more stringent security. But heck, who really needs to go outside anyway?

My imagination is failing me, how would improving image recognition on surveillance feeds make murder-for-hire easier?

The drone-joystick-operator combo already exists and would seem to be exactly what you wanted for that scenario. Weaponry is a bit of an issue if you're not the government, but the project at issue didn't have anything to do with that.

Maybe try reading some of Daniel Suarez's earlier novels.

Good image recognition is a step towards not needing a human operator and that could change a lot. It's no longer a 1:1 or even 1:10 relationship with a human organization. They're more like intelligent mines that can be manufactured and deployed in bulk. They're no longer dependent on a radio signal and can lie in wait indefinitely on a roof somewhere for a good ambush.

No doubt he makes it sound easier than it is; tech often has unexpected limitations. But, it seems like it would be hard to show that it couldn't happen. It seems well worthwhile to try to ban the things (similar to mines).

.. so it's just because it sounds spooky?

Because I thought we were talking about CNN based image recognition in the real world, here.

Could you explain to me, in a few short steps, how we get from there to skynet? I'm assuming this involves top military brass just blithely taking humans out of the loop in drastic contradiction to 75 years of established US aerospace culture?

Seriously, I want to hear something convincing about why I'm wrong on this. There must be something.

I don't really expect the U.S. military to take humans out of the loop any time soon. But the tech industry doesn't seem to be very good at keeping secrets. And IED's and car bombs are already a thing, right?

It is a bit weird that people will work on drones as open-source tech and not for the U.S. military. With the open source stuff, anyone could use it.

The long term outcome of this is not what you want. The less risk there is to going to war, the less incentive there is to avoid war in the first place. More automation will inevitably lead to more frequent, bloodier (especially in terms of collateral damage) wars.

Right. The world is more complicated than "weapons bad love good". Do people really think that fewer people would die if we could somehow roll weapons technology back to 1914? God help us if we did that unilaterally.

We live in an age of unprecedented peace and prosperity, and it's not artificial limitations on technological development that got us here.

I see the adoption of simplistic and absolute moral stances by people smart enough to know better as a form of entitled selfishness. It's as if people feel entitled to performative first-order moral posturing while foisting on other people the job of avoiding the foreseeable negative consequences of adopting simplistic feel-good policies in a complex world.

> Do people really think that fewer people would die if we could somehow roll weapons technology back to 1914?

Nope, that's a red herring.