Hacker News Reader

Comment by saidajigumi | Hacker News Reader

> I think the ethical question is greatly overblown.

Agreed. At least at the level these cars are likely to enter the market, the whole concept of what the car "knows" isn't likely to extend to "other people", "my passengers", "inanimate objects" and a probability matrix of injury/death/damage among these categories. Instead of this heavily anthropomorphized view, cars are much more likely to have a highly simplified, "bug-like" level of information to work from.

This isn't harping on a "technology is neutral" vibe, which I view as obviously false. Just that many of the high-minded philosophical models presented to date don't appear to map well to a likely actual embodiment of "ethical" decisions that may get embedded into vehicular software. If the model is a bad fit for the real-world problem(s) encountered, it's simply not going to be very useful for informing solutions to that problem.

View on HN

> Instead of this heavily anthropomorphized view, cars are much more likely to have a highly simplified, "bug-like" level of information to work from.

Not even close to "bug-like". At the top of modern AI, we've got deep-learning and reinforcement learning.

You probably underestimate the very complex brain that even a bug still has. Maybe also overestimate what today's AI is capable of. You can't build a house with two lego-bricks, just as you can't build a (real time) bug's brain with current-day AI.

It's a humbling process, AI research, finding that even quite "simple" organisms are still miles ahead of our technological state of the art.

More likely than one might want to think.

A car traveling the same direction swerves into your lane, crowding you off the road.

Does your self-driving car allow the collision and push the offending car back in order to hold your lane? Or does it avoid the collision and run you off the road?

(As a person who drives in the mountains, this could have deadly results. If you live in the plains of Kansas, maybe not so much.)

It does not run off the road... And it pushes the break hard.

What kind of ethical dilemma should it have on that example?

It has 0.5 seconds until collision so no hope of stopping. Push the brake all you want, unless that car gets back in its lane, people die.

.5 seconds is both incorrect, and plenty for a car that sees and responds instantly to make a car-length braking correction.

A collision may not be avoidable, but every decrement to the speed will reduce its severity. It's the same idea as having crumple zones on a car body.

Do you have a specific example of this situation occurring as you've described it? If so, was there an ethical dilemma when a human was in control? If not, don't worry about it.

Yes, this situation is in the realm of the plausible, which has little to do with my comment. I'm not arguing against an analysis of the decisions made by a self-driving system. That's essential. I'm arguing against applying ethical models that utilize knowledge outside the realm of the knowledge or "reasoning" capacity of the drive algorithms.

To continue your example, I'll pose that any system with unilateral behavior of self-endangerment is extremely problematic. Motivating example: assume that self-driving cars will always avoid head-on collisions with other vehicles, even if there's no safe alternative. Including driving off a cliff. Attack: manual driver drives down the wrong side of the road (or maybe the middle of a two-lane road), forcing all of the self-driving cars off the cliff during evening commute hours to a hillside residential neighborhood. I suspect all systems that don't hew to some baseline of self-preservation are subject to similar modes of attack.

There's a whole kind of unpleasant but interesting topic about attacks against behavior of self-driving cars. I think I saw an earlier HN thread that suggested that they might put occupants at greater risk of kidnapping because the cars could refuse to take evasive actions that endanger or injure a kidnapper (or just violate a traffic law). This risk might be farfetched, but it may illustrate the concern about people knowing the limits that self-driving cars will respect and then exploiting those limits in some way.

>I think I saw an earlier HN thread that suggested that they might put occupants at greater risk of kidnapping because the cars could refuse to take evasive actions that endanger or injure a kidnapper

What alternative would a human take? Would they kill the kidnappers by driving over them?

I guess that's the simplest one that a self-driving car's software might not approve of. Or intentionally collide with a kidnapper's vehicle, or just make a prohibited turn or run a red light to try to escape.

These scenarios are kind of vague because they don't make it clear whether the self-driving car lets a human driver take over and take full responsibility.

"Would they kill the kidnappers by driving over them?"

Of course, in many countries that would count as self defense (if kidnappers were armed and pointed guns towards your car).

I think a human would drive over a kidnapper. Say your car is bullet proof. A man in a ski mask wielding a machine gun jumps in front of the road and orders you to stop. You refuse, he starts shooting at your car as you get near him (this is why I made your car bullet proof, to keep you alive but unequivocally demonstrate clear and present danger to you). In such a scenario, if your only escape route is over the dead body of the man in a ski mask, I bet a human will take that option.

Wow, where did the machine gun come from!

As this is happening in the future, the gun was probably 3d printed by the kidnapper.

Manslaughter is a likely situation that needs to be addressed in the next Tesla OTA update.

> I think I saw an earlier HN thread that suggested

I might also have heard about some of these problems from Brad Templeton, who describes "Abusing robocar collision-avoidance" among many other interesting problems at

http://www.templetons.com/brad/robocars/downsides.html

I don't remember if he also mentioned the kidnapping scenario when I heard him present about this.

> Does your self-driving car allow the collision and push the offending car back in order to hold your lane? Or does it avoid the collision and run you off the road?

Neiter of those. The first and foremost rule in traffic is to act predicatable. And the only predictable reaction to something blocking your way is to BREAK!

Seriously, what is it always with people that makes them think that a change in direction (swerving) was a good collision avoidcance strategy?

Newsflash, it's not! When swerving you don't change your momentum, you just redirect it (apart from friction losses, of course, the the teensy bit of angular momentum transferred to the planet). More importantly, swerving combined with braking may cause loss of control, which is where there are Anti Break-lock Systems. But ABS will increase your stopping distance, so modern ABS will kick only if the steering wheel does not point into the forward direction.

Of all the scenarios thrown at self driving cars ethics, none of them is an issue if basic traffic rules are followed. And the first and foremost one is, that you always keep so much distance in front of you, that you can come to a full stop within that distance by breaking. And the other first and foremost rule is you never drive faster than road conditions, and more important visibility allow. TL;DR: Don't speed and Don't tailgate!

Any suddenly appearing obstacle in a vehicles way poses the same problem for a human or a computer driver. The computer driver may have the shorter reaction time, but physics is physics and the best course of action remains the same: Braking!

Not always. I've been in a scenario where I swerved into an open breakdown lane because someone swerved into my lane at high speed to avoid running into the car in front (because they had slammed on their brakes after being inattentive). Lots of swerving high-speed vehicles but no one ended up colliding.

But I agree as a general rule. A lot of this trolley problem stuff is way overblown. Far and away the best action in the vast majority of cases is going to be to slam on the brakes. Even if someone or something still collides, the energy is going to be reduced and therefore the consequences of the collision.

I certainly don't want my car flying off the pavement to avoid hitting something on the road without a really good model of why that's a safer approach for me.

Reply on news.ycombinator.com