7 ms·

Waymo CEO: Autonomous cars won't ever be able to drive in all conditions

18 points by sethbannon 8y ago

samstave 8y ago
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 :-) (AI Doping)
- thejrk_ 8y ago
  I could get behind that. Let race teams develop virtual weapons to use during the race too!
  - samstave 8y ago
    that actually sounds fun. jam the others control signals.
- RobLach 8y ago
  RoboRace is a thing. They aim to compete on Formula E tracks with similar performance. https://roborace.com/ https://roborace.com/ Where we’re at is painfully slow https://youtu.be/7NGEM5jcXV4 https://youtu.be/7NGEM5jcXV4
bufferoverflow 8y ago
"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.
- chubot 8y ago
  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.
  - glenrivard 8y ago
    I agree. But have not heard anyone say the same. But I think you can get L4 where it can drive better then humans in bad weather.
tlb 8y ago
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.
- chubot 8y ago
  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.
stcredzero 8y ago
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.
sunstone 8y ago
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.

18 points by sethbannon 8y ago

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 :-) (AI Doping)

thejrk_ 8y ago
I could get behind that. Let race teams develop virtual weapons to use during the race too!
- samstave 8y ago
  that actually sounds fun. jam the others control signals.
RobLach 8y ago
RoboRace is a thing. They aim to compete on Formula E tracks with similar performance. https://roborace.com/ https://roborace.com/ Where we’re at is painfully slow https://youtu.be/7NGEM5jcXV4 https://youtu.be/7NGEM5jcXV4

bufferoverflow 8y ago

"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.

chubot 8y ago
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.
- glenrivard 8y ago
  I agree. But have not heard anyone say the same. But I think you can get L4 where it can drive better then humans in bad weather.

tlb 8y ago

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.

chubot 8y ago
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.

stcredzero 8y ago

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.

sunstone 8y ago

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.