Self-Driving A.I. In Action
youtube.com
Self-Driving A.I. In Action
1–5 of 5 posts
Re: Self-Driving A.I. In Action
#2Looks 99% reliable. I mean it is cool that it tracks all the surrounding items. But there is very little surrounding traffic. And the weather looks perfect. What about dog or small child running out into the street in that neighborhood? I'd like to see it's performance in extreme situations. Still, all told, very impressive. Just makes me nervous.
Re: Self-Driving A.I. In Action
#3Looks 99% reliable. I mean it is cool that it tracks all the surrounding items. But there is very little surrounding traffic. And the weather looks perfect. What about dog or small child running out into the street in that neighborhood? I'd like to see it's performance in extreme situations. Still, all told, very impressive. Just makes me nervous.
Computer Vision tends to fail in surprising ways. The traditional cases are accounted for. It’s the big white box truck against the clear blue sky that will throw it off. It’s logical to think that a large object in the middle of the road can’t be missed but the algorithm has trouble reading subtle hue and saturation differences between the truck and the background. Unfortunately anyone stepping into an AI controlled vehicle is a QA guinea pig.
Re: Self-Driving A.I. In Action
#4The stopping at 0:37 looked scary, as that part of the road looked very easy to recognize. These kinds of mistakes were why Uber engineers made their system more aggressive.
Re: Self-Driving A.I. In Action
#5Looks 99% reliable. I mean it is cool that it tracks all the surrounding items. But there is very little surrounding traffic. And the weather looks perfect. What about dog or small child running out into the street in that neighborhood? I'd like to see it's performance in extreme situations. Still, all told, very impressive. Just makes me nervous.
This is the perpetual problem isn't it?
Even if it could evade/avoid a sudden problem (a child running into the street, etc., as you mention)... what if it cannot avoid disaster and has to choose between two disastrous outcomes. What then?
Disclosure: I'm definitely on the 'maybe not' side of autonomous vehicles (safety-wise), but I wouldn't discount them if the margin of error is reduced "enough". What is "enough"? I'm not sure, personally, but this isn't it.
Humans aren't great at any of this already, so maybe that's an irrational take...?!?