But how else do you imagine scenarios like these are tested? Normally there aren't a lot of people willing to jump randomly out in front of cars to test their reaction.

I really don't know. Even with all the hours of recorded road time by Google/Uber used as a testbed for proper driving conditions, most machine learning systems still suffer from figuring out ways of cheating[1], and in those cases it could be fatal.

[1] "The program even learns how to take advantage of bugs and glitches, like timing jumps so that Mario begins falling again at the exact time that he makes contact with a Goomba. Mario's invincible when he's falling, so the touch kills the Goomba, not Mario, and it gives him a further jump boost." http://www.wired.co.uk/article/super-mario-solved

That's a problem for the test design. Maybe we can develop an approach that uses more randomness to test these systems.

But I dont see how you can avoid having a simulation. These events are just too rare to wait until they occur in real life.

It is not ML deciding where to turn, when to slow down.

I think comparing this to human drug trials is an appeal to emotion.

I also think you're saying that bad tests are better than no tests, which I don't agree with. These would be bad tests because they wouldn't constitute the complicated reality of driving, they inherently only summarize the output of such a complex system.

You seem to have confused the concept of bad tests with incomplete tests. Asking to stop the car when pedestrian shaped obstacles appear seems like a bare minimum that we should expect from an AI before letting it drive on real roads. We can't test every driving situation in a lab, but that doesn't mean we can't test some of them.

If you think it’s an appeal to emotion, make a real argument in that direction, don’t bluntly state the contents of your head and move on. As it stands, you just appear morally bankrupt. I’m strongly inclined to believe that’s the case, and your inability to make a cogent argument makes me wonder about what else you’re lacking.

By the way, you may be misunderstanding “appeal to emotion” as “having emotions” and that’s worrying. I can only hope your attitude is literally pathological, and not just your personality. Either way, please never work near people, or I’m a critical system effecting people.

It would be morally bankrupt to rely entirely on a quantitative test for something that is qualitative, I don't know how much clearer with the previous comment I can get?

We've banned you for repeatedly violating the site guidelines.

https://news.ycombinator.com/newsguidelines.html

You're missing the fact that any time someone takes a drug, they're essentially testing it against their unique body chemistry. The same as self-driving cars in unique situations.

The only difference is the car feedback loop is way better instrumented, and doesn't have to rely on mass statistics to inform recalls.

> Half of all humans who have ever lived, died from mosquito-borne illnes[s], about 5,000,000,000 human lives.

Your math doesn't work.

The current population of the world is between 7 and 8 billion.

If half of all humans who have ever lived were 5 billion, the total number of people who have ever lived would be 10 billion, of which at least 7 billion are still alive.

That leaves less than 5 billion who could possibly have ever died for any reason, much less mosquito-borne illnesses.

“Modern” Homo sapiens (that is, people who were roughly like we are now) first walked the Earth about 50,000 years ago. Since then, more than 108 billion members of our species have ever been born, according to estimates by Population Reference Bureau (PRB).

Source: https://www.prb.org/howmanypeoplehaveeverlivedonearth/

I imagine just rolling some objects across the road or some self-riding bicycles should be enough.

But what if the self driving cars fail the simulation? The simulation would be the first thing the car needs to pass, if car passes the simulated, then special tests, then you get approved to test on public roads with a security driver, I have no idea when is the time to test without a driver.