Hacker News Reader

Comment by incomingpain | Hacker News Reader

You can never hold the robot accountable. It's in direct violation of asimov's laws where it's intentionally harming humans. Bad idea.

On the opposite side, the robot never gets to fear for its life and end up shooting innocent people. There ought to be high definition recordings proving lethal force was necessary. Freely available to all public for review. Absence of video being immediate summary judgement for misconduct charges to the operators which includes jail time. Police unions thusly aren't getting involved.

So I can see some pros and cons. The big question I have, why is San Fran the first to do this?

https://en.wikipedia.org/wiki/Homelessness_in_the_San_Franci...

View on HN

> You can never hold the robot accountable. It's in direct violation of asimov's laws where it's intentionally harming humans.

It's remote controlled by a cop. There's no AI.

It's not asimov's laws of AI, it's the laws of robotics.

And those laws are not actual laws in the eyes of the government, why would they care

Firstly Asimov's laws aren't inviolable physical laws. They were just fictional idea that were suggested could be programmed into robots as a plot device in his books. The stories were frequently about unexpected side effect caused by intersections of the laws and specific situations that caused some kind of ambiguity in how they should be followed. Real world robots could be programmed with or without them, or with entirely different behavioural 'laws'.

Secondly, the 'laws' were about behaviour that applies mostly to the AI side of things, the fact they are referred to as laws of robotics is mostly immaterial, they are about behaviour rules for AI.

I was going to disagree, but re-reading the laws would mean that the robot would be required to be capable of overriding any input command from a human which might cause it to harm another human.

I'm inclined to agree that violating this is generally a bad idea. I don't know that I would go so far as to make it a law, as I am not a pacifist and would prefer a robot to retain the capability to override the first law exclusively in the interest of obeying the second law, when there is an audit trail of commands given such that someone- whether the robot or the human behind the robot- can be held to account.

You've got to be kidding. Law enforcement robots are completely remote controlled by a human operator. They have no capability to override an input command, or even understand what harm to a human means. We're talking about the real world here, not science fiction.

The robots have video cameras so those recordings can be used to hold the human operator to account.

No, this thread is talking about whether or not violating asimov's supposed laws are a Bad Idea.

Naturally, the current implementation is a long, long way from complying with those concepts.

The real question, for today and now, should be: Should we be focusing on implementing more safety measures (getting us closer to the principles behind the laws) or should we deploy what we have in a situation where they can be used for violence?

Are the video feeds from the robots recorded? Are those recordings accessible to the person manipulating the robot, or their direct superiors? Are there backups in case those "get lost"? Are there on-board backups in case of network loss? Some sort of black-box maybe? Can the robot's weapons be deployed if the recording is not in progress?

Even if the robot itself doesn't have an independent will, it can still be equipped with safeguards. Hell, automatic doors almost always have sensors that prevent the door from crushing something if it is in the way of the door closing. Surely we should be more stringent with weaponized robots than we are with garage doors.

Asimov was just a mediocre author, nothing more. Nothing in his novels can be taken seriously as a starting point for real public policy discussions. Come on.

That's just ad hominem. As I've been saying, I don't even 100% agree with them. I do, however, find some value in the idea that weaponized robotics should have failsafes against human instructions to commit violence.

Again, consider the garage door example. You push a button, the garage door closes. If something should trigger a safety sensor, the garage door reverses rather than crushing whatever tripped the sensor.

How many times have police been involved in shootings where the body camera was present but not running, or the recording was "lost"? I think it is perfectly reasonable to insist on more safety measures, such as failsafes against the robot firing its weapons if it isn't recording or detects interference in the recording and backup mechanisms.

This is something I hadn't fully considered when I reviewed the three laws; if you go back to my first comment on the thread, I had initially been of mind to dismiss it.

Probably because cops don't want to show up to calls where they may have to use lethal force and possibly have their lives destroyed and decisions Monday morning QBd.

Definitely a bad idea though, I don't like the idea of a guy in a suit being able to hit a button. You also don't need to convince a robot to violate my rights and put themselves in danger while doing it.

Most of the problems with policing as we see them today involve police behaving as humans do in life or death situations, even "bad shoots" have fundamental elements of reasonableness to them most of the time, whereas a robot does not have to worry about being killed by the perp or any number of things that require use of force that we consider excessive.

Just guessing this would be used in proscribed circumstances, such as a hostage situation or someone holed up so you can further avoid hurting bystanders. The advantage is you don't have to act in a fraction of a second because in the person + robot system the person representing the law is not in immediate mortal danger and is not put into a life-death decisionmaking pressure. So presumably the outcome for law abiding citizens is improved.

Surely the explosions will ignore the innocent bystanders

How many 'hostage' situations are there in San Fran? I expect none or damned close to none.

No, these bots are going to be used in a more liberal fashion to justify all the effort being used to arm them.

This wont be for rare situations, this will be used for their ongoing crisis.

Yes, police response is typically prescribed when people do something that is proscribed.

Reply on news.ycombinator.com