Hacker News Reader

Comment by sfusato | Hacker News Reader

The weapons, perched over a crowded Palestinian refugee camp and in a flashpoint West Bank city, use artificial intelligence to track targets.

AI + guns, what could go wrong?! Towards the end of the article, they make it sound like it's for the good of the oppressed. I guess we have a good contender for a future Nobel peace prize award.

It also said the technology allows soldiers to target "less sensitive" areas of the body to minimize harm and avoid shooting bystanders.

"In this way, the system reduces the likelihood of inaccurate fire," it said.

View on HN

Sounds good to me. Soldiers controlling the thing can make smarter decisions if they aren't in imminent danger.

If you are standing post and perceive a threat you cannot afford to wait and see. This is just a piece of equipment and ROEs can hopefully be more relaxed.

Oh wow..

double wow to people seeing it like that.

I did a double take too, but I think they meant the rules of engagement can be “more relaxed” as in, to relax the envelope in order to make engagement less likely, rather than relaxing the rules of engagement in order to make engagement more likely.

Could you please elaborate?

>Towards the end of the article, they make it sound like it's for the good of the oppressed. I guess we have a good contender for a future Nobel peace prize award.

Isn't that the point a well written PR article? To inform and to persuade?

As someone who have done some reputation management work with the defense sector and firms that offer private peacekeeping solutions, it is crucial to shape the narrative early to avoid misunderstanding from the general public.

> someone who have done some reputation management work with the defense sector and firms that offer private peacekeeping solutions

I'm inclined to believe you, as that's a rather... uh... colorful way to say "propagandist".

Anyone curious how a Palestinian refugee camp can exist under Palestinian authority 70 years after the war that made those ppl refugees? Are they going to stay refugees forever?

>to avoid misunderstanding

Whether it's misunderstanding or proper understanding, is up for debate.

Its important to lie and promote the idea that its not an apartheid state running a gigantic prison.

You mean to avoid actual understanding? I think everyone intuitively understands what ai guns leads to. Wars that the gov can do without its public caring so much, and govs that can't be toppled by civil unrest (read the military/cops switching sides)

Big difference between automated tracking and automatic firing.

The difference is a few lines of code

TBF the difference between many does-good-things and does-bad-things systems is a few lines of code.

A few lines of code can turn an X-Ray machine into a death machine. A drivers assistance system into a kamikaze car. An industrial automation system into a bomb.

Those things are not built for the express purpose of killing people. There is no manual button on the outside of the xray marked "kill patient" that's at risk of being automated away. No "aim for pedestrians" button in the Tesla that a developer will choose to finish implementing in code.

Like, yes, this is a thing that could happen, and has, we call them bugs. But the companies building those products aren't in the business of making that outcome occur. These developers and companies are in the business of helping make people die and I'd bet significant sums this is already tracked in a backlog issue or behind a //.

So what is your point? Why is automated target tracking bad exactly? I’m not saying it’s not, but I don’t see how your argument is different from “all weapons bad”.

The point is that "you can invert the rewards signal on a Tesla and turn it into a people mower" isn't exactly a dunk on the point above them. We're not worried about Tesla having (someday accurate) people detection because we're not worried about them using it to kill people more efficiently. Indeed, generally people want Tesla to get very good at that!

So automated tracking is bad when attached to a gun because it more easily enables the rest of the concerns laid out in the thread. It's not some hobbyist "paintball the squirells in my yard", but a government-purchased, corporation built project with an explicit, obvious goal in mind. Disregarding that because many other things _could_ be used for bad with a couple lines of code is not the right call, imo.

So don’t you want a gun to hit the target with no collateral damage? Imagine a different use-case, for example a mass shooter in a crowd. Where an AI controlled gum can take out the shooter and not the victims, the way police occasionally do.

Sounds like nyouve forgotten the history of failures in this field relating to identifying who is or is not human based on skin color.

Some of the victims will invariably be dissidents rather than mass shooters when nasty regimes get their hands on them.

So we’re back to the arguments against weapons generally. Yes, you can use a weapon for bad things.

Hope fail-safes get thought out more than this one[0].

[0] https://youtube.com/watch/Ip7GGf2_b6Y

This simply isn't the case for the X-ray machines you'd find in a hospital or at a dentist's office. Industrial equipment though? Sure

See https://hackaday.com/2015/10/26/killed-by-a-machine-the-ther...

> the absence of international norms for "killer robots" is problematic

nonsense, they're totally bug-free

Reply on news.ycombinator.com