While many social media users are blaming the pedestrian for reportedly crossing against the light, the incident highlights the challenge autonomous driving faces in complex situations.
This doesn't seem that complex to me. If there is a pedestrian in front of your car when the light turns green, you wait. Pretty fucking simple. This isn't some offshoot of the trolley problem where an incident was unavoidable. The car made the active decision to proceed when it was not safe to do so.
Why have we programmed our self-driving cars to emulate the psychotic behaviour of a typical road ragin' car-brained human? Isn't that the problem these projects should be trying to solve?
I'm going to inject some unpopular nuance here, so I'll preface by admitting that I haven't looked further into this event than the information provided in the linked article, which isn't much. Nevertheless, a few points:
No system is perfect, including exclusively human drivers. Obviously zero accidents is ideal, but as you said, road ragin' car-brained behavior is typical. How many people are killed every year by human drivers?
Obviously driverless system development should aspire to dynamic reactivity comparable to the best human driver. But when running a cost-benefit analysis for driverless adoption it's worth considering if, normalizing each by their respective total hours-on-the-road, the mistakes made by driverless cars due to rigid adhesion to traffic laws outnumber the mistakes made by drivers due to their own flagrant disobedience.
That's an interesting comparison and something I've wondered about quite a bit. I would be surprised if machine drivers were not categorically safer than human ones, and if safety is (rightly) a priority in the cost-benefit analysis of driverless car adoption, then it's hard to imagine not concluding that we ought to proceed in that direction.
But I think this specific incident illustrates very well that the human vs. machine driver debate is tragically myopic. If an infallible machine driver adhering perfectly to traffic laws is empowered to accelerate from a standstill directly into a violent collision with a pedestrian, then maybe it doesn't matter how "safe" the driver is. I take it as evidence that car travel the way we have it set up is inherently unsafe. Our traffic laws emphasize the convenience of car traffic above everything else -- including safety -- and only really serve to shift blame when something goes wrong. Despite its certainty, there is very little builtin allowance for human error aside from the begrudging mercy of other parties.
To be fair, human drivers are an unmitigated disaster which we really need to do something about, but I think if we're going to go through the messy process of reforming how we think about cars, we might as well go farther than a marginal improvement. We could solve the underlying problem and abolish the institution of car dependency altogether, for instance. Otherwise it just amounts to slapping a futuristic band-aid on a set of social and economic issues that will continue to cause unimaginable harm.
To repeat myself from the other post where I'll probably be downroaded:
The car should be programmed to self-destruct or take out the passengers always. This is the only way it can counter its self-serving bias or conflict of interests. The bonus is that there are fewer deadly machines on the face of the planet and fewer people interested in collateral damage.
Teaching robots to do “collateral damage” would be an excellent path to the Terminator universe.
Make this upfront and clear for all users of these “robotaxis”.
Now the moral conflict becomes very clear: profit vs life. Choose.
Why would you hold self-driving cars to a standard that we don't hold drivers? If you are a driver and realize you are about to harm a pedestrian, there is no circumstance when the law suggests you ram a car into a building or pole instead of the pedestrian. Your insurance would rather you hit the pedestrian, usually. Because in an animal strike, hitting the animal is comprehensive (in America) and swerving to hit a fence is collision. You can't be at fault for comprehensive. A pedestrian is a different mater and not comprehensive, but they'd rather you mitigate liability, and then mitigate cost. And there's a chance the pedestrian was at fault, at least partially. The building/pole can't be.
But all of this is a moot point. Self-driving cars will NEVER be programmed to harm the driver before an outside person. Simply for the fact no one will ever buy or ride in a car that chooses to kill the passenger over others. No one will ride in the Suicide Car.
Self-driving cars should absolutely be held to a higher standard than humans. They are not humans and cannot be held accountable for their actions, therefore the benefits of their use over human drivers should be overwhelming before we allow them in the streets.
As for the trolley-esque problem being discussed, it's actually an incredibly complicated problem with even more complicated solutions. A statement like, "hit a wall instead of a person", seems obvious to a human but just adds a million complications to the situation. How do you detect if it's a safe wall to hit? What if it's a fence on a schoolyard with 30 children sitting on the other side.
No. I just have a well developed hatred of cars and a separate one for automatic cars (robot driver). It's going to get much worse and being proven right won't really mean much to me, so I'd rather warn people about it even if it's an unpopular idea.
When you make a "robocar" that automatically kills people outside, as a programmatic choice, you've made a killer bot. Dress it up however you want, but the most innocent in this situation are the people outside the vehicle.