LogFAQs > #953430941

LinkPizza posted...

They are in the right lane and the only way to avoid them is to go to the left (could be a sidewalk with people or even a gas station with people and cars). But the left lane has more traffic. Because it has to be done fast, it swerves to the left toward another self-driving car. And that one has to swerve to the left to avoid that car. It could easily get to the point where one cant swerve without causing an accident. Or one of them dont register the signal to move left. Or that a human driver was to the left, which they either hit or dont swerve to hit causing an accident with the other self-driving car. In the end, its possible the first car that swerved didnt know about the accident that occurred, thus didnt stop Especially in the human took control because it swerved and they got spooked by it.

OK... so how does the presence of a human driver instead of an AI in this situation somehow change the scenario?

With what you've just laid out, you're going to have an accident whether it's humans driving or AI. The AI crash is, honestly, probably going to be less severe, because this network of cars is all talking to each other saying, "Oh crap, need to make an emergency lane change, please move over," where humans don't have the ability to do that.

And if the first car drives off, so what? The rest of the self-driving cars all have cameras recording what happened, including the license plate of the car that drove away. If there's an insurance issue, the owner of that car can be contacted and brought into the process as appropriate.

LinkPizza posted...

But Im sure self-driving cars would try not to hit each other, so

Then you're wrong, much as you have been about this whole scenario.

Self-driving cars will hit another self-driving car if they calculate it is the safest option available. For instance, if they have the option of hitting a pedestrian or another vehicle, they are programmed to choose the vehicle because it is less likely to result in a serious injury or a fatality. Notably, if you've ever taken a defensive driving course you'll know that's the exact advice given to humans in the same scenario: hit a vehicle instead of a person, swerve right into a ditch instead of left into oncoming traffic, hit a stationary object instead of an object moving towards you; hit an object moving the same way as you instead of a stationary object.

LinkPizza posted...

They probably easily can.

They probably easily can what?

If you're going to continue this ridiculous practice of not quoting what you're responding to, you at least need to provide a context or subject in your sentence so that I can figure out what you're talking about.

LinkPizza posted...

AI may have better vision. But thats better vision at what they are looking at. Which might not be the face of another driver in a different car. So, I still say humans can probably tell whether another driver of a different car in asleep

You keep saying, "might" or "may" or "probably" in your sentences, which implies that you both have no idea if what you're saying is true (it isn't) and are suggesting that it isn't a known fact (it is).

AI don't "look at" things the way humans do, because they don't see and focus on things the way we do. They take in everything within their field of view (which is 360 degrees for a self-driving car). They, unlike humans, can pick up details on everything happening around them. If a driver's body language or driving patterns are indicative of someone falling asleep, the self-driving car will be able to pick up on a human falling asleep and react accordingly far faster and more reliably than a human driver will.

---

Kill 1 man: You are a murderer. Kill 10 men: You are a monster.
Kill 100 men: You are a hero. Kill 10,000 men, you are a conqueror!