San Francisco supervisors vote to allow police to use robots to kill | Hacker News Reader

San Francisco supervisors vote to allow police to use robots to kill

lite.cnn.com

56 pointsrntn28 comments

> "There is serious potential for misuse and abuse of this military-grade technology, and zero showing of necessity,"

Considering the track record of police I'm inclined to agree. This will absolutely be abused. At least it seems like this will be limited to using "ground-based robots to kill" so it's not like we'll see drones taking out protestors or no-knock-bombing the wrong address just yet.

The headline makes it sound like the robot can make this decision but in reality this is just a RC car with a weapon attached to it that is still controlled by a human

The headline is not very misleading

idk, my first thought was an autonomous machine - that's what my brain thinks when it thinks "robot".

But I wonder if the decision requires the "robot" to have a human controller or can the police unilaterally expand the kind of devices they use to include more autonomous ones?

In the USA, if I understand it correctly, it seems there is a lot of case law establishing criminal or civil liability for autonomous mechanical devices that kill or injure someone - Typically traps like tripwired guns, so it would be interesting to see how it applies to autonomous robots having some degree of "perception" and "judgement"

Do those cases apply to the police though? Qualified immunity is basically a separate, non-overlapping set of case law for police liability that also coincidentally almost always rules that they have none.

Right, an RC bomb is not by itself too different from say a sniper, it's if this establishes a precedent for more autonomous lethal robots.

Sort of like calling in a targeted drone air strike on your own population.

"Suspect killed by robot was caused by computer glitch"

"Report: AI misidentified black suspects as having a gun in 80% of deployments"

Putting distance between the human and the point-of-impact makes it easier to pull the trigger. Will be interesting to see the stats on this.

I would have thought moving LEOs away from danger would _decrease_ the use of unnecessary lethal force.

Their lives are no longer in danger and thus they have no reason to panic and react (with lethal force) to unconfirmed threats.

Thoughts?

It can also escalate the level of violence. For example if you know a robot will come after you with deadly force, probably going to go ahead and get heavily armed, potentially using black market robots. A cat and mouse game

It also removes all other nonviolent options. You’re left with a gun toting tv car and a loudspeaker. Cops other don’t apply force due to threat but due to disobedience of their target.

Danger isn't the reason they kill, it's just how they justify it.

Not always. I recently saw a video of a man attempting to stab a woman and the cop's bullets were clearly the only reason her jugular wasn't sliced open. It was CLOSE.

I disapprove of killer robots though.

One time istg no shit I dropped a bunch of change on the tile floor of the laundromat and a nickel landed on its edge.

If someone says a coin toss is 50/50 I don't jump in with "not always" though.

> "There could be an extraordinary circumstance where, in a virtually unimaginable emergency, they might want to deploy lethal force to render, in some horrific situation, somebody from being able to cause further harm," Supervisor Aaron Peskin said at the board meeting, according to the San Francisco Chronicle.

A lot of qualifiers for a hypothetical scenario. How many will apply when this actually gets used?

I thought SF was one of those places that wanted to send social workers out rather than police. Have they changed their minds now and want to send out murder bots? I can't wait for ED-209

Robot is an (intentionally) ambiguous term. Is it autonomous? How autonomous?

If it's an RC car with a hand gun glued on, who cares. If San Fran secretly has a T1000 (and they don't but let's play boogie man for the fun of it) then that's different.

Of course, we don't have self driving cars or even automatic aiming for real guns so this is almost certainly just a machine with a human operator making all the decisions just at a distance...

>we don't have automatic aiming for real guns

Yes we do. We've had various implementations of it for as long as computer vision has been readily available, as demonstrated by this decade-old video (https://www.youtube.com/watch?v=0WR9g9o1r2M).

There's no difference as far as the turret goes as to what type of gun it uses.

When you send this thing in you have basically decided the execute the suspect. No trial needed

The tools needed to rob a bank just went from a gun, to a baseball bat, an old tv remote, a landline wireless phone, five feet of 5 gauge copper, and a CD (preferably smash mouth).

How long before armed robots are autonomously "clearing sidewalks" that just happen to be obstructed by occupied tents?

This is absolute insanity. Is there anyway to reverse this? A statewide initiative to ban this in all local jurisdictions?

Remotely controlled killer drones get a civilian application.

And so it begins....

Reply on news.ycombinator.com