We should have self driving car F1, ralley and nascar races... with betting on who's going to win - and the proceeds going to fund the research further for the winners.
but wait until the scandal comes out that one of the winners was actually being remotely piloted by a real person :-)
"Ever" is a long time. I'm certain it's a matter of time before an AI drives better than an average human in all conditions. And then after a few years it will drive better than the best humans.
I believe solving level 5 self-driving is equivalent to solving AGI -- i.e. it's "AI-complete".
In that case, there will be a lot of other things to be excited about besides self-driving -- it will basically be the equivalent of aliens landing on Earth.
Also, while I agree "ever" is a long time, "a matter of time" doesn't tell me very much. That is, wouldn't surprise me if it were 100+ years away, in which case there other things we should be talking about instead.
People also can't drive in all conditions, so it's kind of a trivial statement.
I learned to drive around Ottawa, where freezing rain and white-outs happened occasionally. At some point the wise choice was to put the hazard lights on, pull to the side of the road, and wait for it to be over.
Presumably self-driving cars will also be able to do that. So it's not clear what the difference will be, other than that self-driving cars are likely to be more analytical about the risks of proceeding vs. stopping.
Many people have been trumpeting "superhuman" performance of deep learning on narrow tasks, and extrapolating that to a future where self-driving cars are ubiquitous and saving our lives routinely.
That performance is technically true, but unfortunately it doesn't translate to real-world engineering problems like driving cars.
The question isn't if computers will be able to drive motor vehicles like the best people. The question is if computers will be able to drive well enough to make companies money and save companies money. If the answer to both of those is yes, then autonomous cars are inevitable.
While this comment is strictly true. Once self driving cars become common in cities, cities will start modifying roads to be more accessible to self-driving cars. This will result in "close to perfect" self driving cars in urban areas within a decade or two.
On the other hand, driving on remote country roads in ice storm conditions is something that self driving cars are never likely to be ready for, at least until tires come with automatically deployable half inch studs.
11 comments
[ 4.1 ms ] story [ 34.0 ms ] threadbut wait until the scandal comes out that one of the winners was actually being remotely piloted by a real person :-)
(AI Doping)
Where we’re at is painfully slow https://youtu.be/7NGEM5jcXV4
In that case, there will be a lot of other things to be excited about besides self-driving -- it will basically be the equivalent of aliens landing on Earth.
Also, while I agree "ever" is a long time, "a matter of time" doesn't tell me very much. That is, wouldn't surprise me if it were 100+ years away, in which case there other things we should be talking about instead.
I learned to drive around Ottawa, where freezing rain and white-outs happened occasionally. At some point the wise choice was to put the hazard lights on, pull to the side of the road, and wait for it to be over.
Presumably self-driving cars will also be able to do that. So it's not clear what the difference will be, other than that self-driving cars are likely to be more analytical about the risks of proceeding vs. stopping.
That performance is technically true, but unfortunately it doesn't translate to real-world engineering problems like driving cars.
On the other hand, driving on remote country roads in ice storm conditions is something that self driving cars are never likely to be ready for, at least until tires come with automatically deployable half inch studs.