Advertisement
REGISTER | LOGIN
Breaking News
News & Analysis

Robocars: What We Saw in Vegas

1/17/2018 00:01 AM EST
6 comments
NO RATINGS
Page 1 / 8 Next >
More Related Links
View Comments: Newest First | Oldest First | Threaded View
sixscrews
User Rank
Author
Re: Although this sounds too good to be true....
sixscrews   1/18/2018 6:54:50 PM
NO RATINGS
Point well taken.  I like the idea of getting at them through their insurance.  And a 'driver rating system' might be useful.  Big data - or is it Big Brother? ss/wb

realjjj
User Rank
CEO
Re: Although this sounds too good to be true....
realjjj   1/18/2018 6:31:43 PM
NO RATINGS
Don't stop after detecting a problem, try to also find solutions. You operate a large fleet of AVs with a rather long memory, what do you to about hostile drivers? You can record, identify and classify, report, adjust behavior, retaliate. You can work with law enforcement agencies, with insurance companies or "driver rating" companies (think credit rating). If someone is trying to mess with your billions, you are gonna deal with it. Ofc it's preferable to avoid retaliatory actions - robots harassing humans - but some will likely do that too.

sixscrews
User Rank
Author
Re: Although this sounds too good to be true....
sixscrews   1/18/2018 4:12:10 PM
NO RATINGS
Bert, your point is well taken. I'm not too concerned about AVs that can't recognize an oncoming hazard (i.e., a reversing vehicle) or chemically impared drivers. It's 'normal' drivers who game the system that bother me. Think of AVs on Interstate highways keeping a strict following distance.  This offers a huge opportunity for lane-changers going faster than the flow of traffic to cut in - while the AVs will then slow down to keep their following distance. How long will it take for most of the drivers to recognize this 'weakness' in AVs and use their regulated following distance as an opening to gain a (likely illusory) advantage in the game of cars 'n trucks? ss/wb

EELoser
User Rank
Freelancer
Re: Although this sounds too good to be true....
EELoser   1/18/2018 12:18:22 PM
NO RATINGS
So how much more did it cost to repair the autonomous vehicle that was backed into compared to a regular vehicle that was backed into?  

Bert22306
User Rank
Author
Re: Although this sounds too good to be true....
Bert22306   1/17/2018 5:55:05 PM
NO RATINGS
Not sure I understand the argument. Presumably, the autonomous vehicles will be replacing some of the human drivers on the road, rather than only adding to traffic, and will be replacing these sometimes irrational and unpredictable operators with more predictable ones. I don't see how this can make matters worse.

One example might be that recent incident, where a truck at a red light backed into an automous vehicle waiting at the same red light. People made a big fuss about that, but how exactly is that any different from what would have occurred with a human driver? The only objection I heard was, why didn't the automous car blow its horn? Well, that's a simple enough fix to the algorithm, but it would likely have made no difference anyway. No one wondered why the truck driver decided to back up. Had the truck been autonomous too, chances are good that the minor accident wouldn't even have happened. If for no other reason, an autonomous truck would not back up blind. It would check its rear vision first. Easy to program into an algorithm, seemingly impossible to teach human drivers.

And never mind the more obvious and egregious examples. How can it be worse, if a drunk decides maybe it's a better idea to set his car on autonomous driving, before going home from the bar on Friday night? Or, how can be worse if the guy who feels compelled to text while behind the wheel, first sets his car to autonomous mode? For that matter, why assume that a sober driver is better able to avoid that drunk or texting idiot behind the wheel, who is driving manually? Just the better 360 degree vision of the automous car should make it more capable of avoiding sudden obstacles, without causing another accident to happen.

sixscrews
User Rank
Author
Although this sounds too good to be true....
sixscrews   1/17/2018 4:32:32 PM
NO RATINGS
OK, it's too good to be true - so it ain't. Computers do what you tell them to do, not what you want them to do.  AI and machine learning are dependent on deterministic systems that are only as good as the instructions given by their human developers.  And Kurt Gödel showed back in 1931 that any logical system cannot encompass all possible outcomes. I would eat my intel i7 processor (better than a shot glass) if someone convinced me that a deterministic system can cover all contingencies.  And the biggest contingency is the behavior of other humans operating vehicles in the same space as an AV. I recently read an article about car crash fatalities in the US that assigned 95% of the fatal crashes to human error.  Although definitive data are not available for 2017, it is estimated that 40,000 people died in car crashes in the US.   Taking the 95% human error number, only 2,000 of these deaths were due to 'other causes' - weather, mechanical failure, road hazard or acts of assorted deities. IMHO, unless we segregate AV traffic from human-operated traffic the future is grim. Think of fleets of AV taxis or load haulers overwhelming our already strained transportation system then add in the behavior of annoyed humans in non-AV vehicles. It almost makes me considering joining my Amish neighbors and going back to horse-and-buggy transportation.  Fat chance.  ss/wb

Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed