yup, darwin award.

 

Darwin award or not, but this is precisely why fully autonomous cars will never happen, not 15 years, not 15,000 years from now. Because human drivers can actually predict behavior based on circumstances. There are many situations on the road where human instinct makes a human driver slow down and proceed cautiously, and this is one such scenario. Computers can never be programmed to have the same amount of intellect and senses as humans, and if someday they somehow can be, they will probably develop the same negative traits as humans - but that is too far fetched to even think about.

 

What's pretty scary is how there is clearly a position, that if a human makes a mistake in judgement, exercises poor judgement, their life can be ended and somehow it is not of any consequence.

You ever see that teenage prank, where say you get a phone call from Uncle Benny, and when you ask Uncle Benny why he called you, he says wtf are you talking about, you called me? And now you're dumbfounded how this could have happened? Later, your 12 y.o. explains all he did was go to a website and play a prank on you?

In grandpa's days, this would not have been possible, because you, and uncle benny, had a pair of wires physically connected to your phones. Anyone who wanted to play such a prank, would need physical access, which likely would be impossible. What makes you think today that your cell conversation, or home phone conversation, is private? Have you verified the encryption? What I'm trying to say is tech gets too much of a free pass, which, is, actually good for those who are on top of it and want to meddle. Even FB got played, and one would think they are or were at the forefront.

 

Feel free to explain how someone emerging into the street from behind a truck which shields the person from the onboard cameras in front of a car driving 30 mph can stop in time if they are a mere few feet in front of the car so we can all understand.

But read Post #600 above before you respond.

 

How does a self-driving vehicle change this proposition? People get injured and lose their lives every day by exercising poor judgment in situations like this. The self-driving vehicles are programed to react quicker, without being distracted, to attempt to avoid those situations. But no one ever said it would be foolproof.

 

Again, I get that this thread is clearly pro technology, and that a fatality cannot stand in the way of progress.

What if, just what if, it is commonplace for human beings, to attempt to cross illegally--yes, jaywalk, along this stretch of an avenue (Roosevelt Avenue or Queens Boulevard come to mind), because people do not want to go all the way to the crosswalk. A driver, drunk or sober, may be be a little extra careful because it happened yesterday, last week, last month, and many times last year, that people were killed attempting to cross, illegally. They imho rightfully so do not want to be involved in a fatal accident no matter whether the human is crossing illegally or otherwise (it's a hassle, don't like the idea of killing someone else, whatever).

We have to make a judgement, do we legally mow these people down, or do we try to avoid them? I don't see where this autonomous XC90 has such a decision tree in its code. Hey, if it does, show us. I cannot see why tech is given a free pass over humans.

 

Ever take the Major Deegan southbound, to the GW Bridge? I've always been amazed how a tractor towing a 53' trailer is able to go up the right side, and get in at the last minute. Whether we say he's not waiting his turn and it's not fair, or otherwise, that is pure skill. His trailer is 53'. If we were to program a autonomous tractor to do the same, I guarantee it will be involved in an accident, or it will stop to avoid an accident. It cannot make the same decisions and have the same judgement as a human. How many factors and instantaneous decisions were made by these drivers, to be able to barge in like that?

In reality, we're going to have to crush a few motorists to get the autonomous tractor to be able to do the above, for real. It's not going to happen in a vacuum on someone's monitor. If the latter were possible, every car would have been able to handle like a BMW in the 1995-2008 era.

 

When I drive through neighborhoods or areas where there are trucks or objects blocking my view or others from seeing me I drive slower and put my foot over the brake and try to move slightly to the other side to get a better view and put a little more distance if something or someone should dart in front of me, I actually avoided hitting a kid by doing this one time driving in my neighborhood, he darted right out into the street when there was a moving van blocking his and my view but I did see kids running around in a yard and knew it would be better to drive slower and keep my foot over the brake and move further over, I hit the brake hard immediately when I saw the kid and avoided a tragedy, if I did not take that precaution I likely would have hit the kid, possibly seriously injuring or killing him and I would not have thought it was my fault considering he darted out in front of me, I was not speeding, and a object was blocking our view but it still would have been terrible. Is a autonomous car going to be able to read that situation and slow down, move over a little, and immediately brake even if the kid is not right in front, unlikely, it will likely keep driving at a higher speed and only brake when it senses something right in front of it which might be too late. That is the problem with this technology, there are human instints and experience it can't replicate when driving and there are certain situations it is not going to be able to read other situations. What if a bag, mud, rock or debris lands on a sensor or destroys it while in motion and blocks its view/sensing, then what?

 

what do you mean by 'ML'?

 

If I had to guess, it would be Machine Learning

 

not in 15000 years huh? computers can never be programmed? respectfully i don't think you have much depth of understanding of technology. but your point about if they can be they will have the same traits as humans has a grain of truth RIGHT NOW because volvo's self-driving tech simply learns from monitoring human driving. eventually it learns the same 'skills', warts 'n all. but there's a GIANT difference i'll get to later here...

not sure how you get that, it's a conversation, of which you are a part.

respectfully, i'm not sure you're considering how the cloud and machine learning factor into all this. it's possible that in time each autonomous vehicle will learn from the collective experience of all other autonomous vehicles so indeed, vehicles can become 'aware' of incidents' or needs for caution in particular situations. consider also that if enough people were killed by autonomous vehicles in a particular spot that the surviving humans might learn.

self-driving cars absolutely do already and will continue to grow in accident avoidance and better and better rules for endless scenarios. but as ted points out, the laws of physics do still apply, a car can't stop from 60 in 10 feet so if someone leaps in front of a car, human driven or not, the human will be toast.

but think of other scenarios where things are BETTER with self-driving... here in florida with loads of elderly drivers (and all over the world but prevalent here) elderly drivers occasionally get confused and hit the wrong pedal or steer the wrong way and end up plowing into a store front, someone's front yard, off the road, and endangering many. mistaking the pedals would be a thing of the past.

again, collective autonomous car data sharing and machine learning will mean that in short order self-driving cars will be better than ANY human drivers.

i'm waiting for the first formula 1 self-driving car that beats the snot out of the pro drivers. it won't even care about g-forces.

 

That Volvo would have needed to be equipped with night time camera tech and an incredible ability to see what the human eye can't see and then integrate that with the car's emergency braking and alert systems. IMO if a human could not have avoided this accident neither could the self driving car (at least at this point).

Even if the Uber software was learning as it went and slowed down drastically at that particular stretch of road at night every time it travelled there in anticipation of jaywalking pedestrians; it would have needed to slow down quite drastically to be able to even stand a chance of avoiding hitting a person. That would result in complaints from other drivers and the usual "I hate slow drivers" issues. It all comes down to who bears more responsibility here and unfortunately it was all triggered by a decision to not use a marked crosswalk.

 

Agreed lol.

 

i believe the sensors are laser and radar so they don't care if it's day or night.

 

From what I've found it only works for urban low speed environments. There is no tech that would apply to the speeds that the Uber would have been travelling at, especially if someone just darts into the lane. According to the safety driver's report, the car didn't try to even mitigate the collision and that means the pedestrian would have just darted out into the driving lane in front of a car that is thought to be travelling at 40 mph.

That's approximately 60 ft/sec. The subsequent investigation will have to determine at what point did the woman actually get in front of the car. If she was stationary in the median and the vehicle's radar simply bounced a signal off her, then the Volvo would not have detected that as a hazard.