2 ms·

I've never really expected self driving cars to lower the rate of accidents simply because cars are dangerous whether or not you have the reaction time of a com

I've never really expected self driving cars to lower the rate of accidents simply because cars are dangerous whether or not you have the reaction time of a computer. Can it save the human in a moment of inattention? Absolutely, my prius has shouted at me to brake when I was messing with the radio, but it doesnt matter how good your radar is when you're flying around a blind turn or coming up behind an invisible, stopped object. (%50 reduction would be good, drunks taking robo taxis home is great, but IMO driving is easy and robots wont be better at it)

al_borland 3y ago
In an interview with Boston Dynamics, one of their robotics engineers said something like, “the hard stuff is easy and the easy stuff is hard.” There are things that humans can learn and do extremely quickly, or figure out on the fly, that robots require extensive training for.
- usrusr 3y ago
  Stuff like checking the rear view mirror when slamming the brakes upon some surprise uncertainty, true. Humans will always be bad at this, but for a computer it's trivial to include the tail situation in the brake decision.
  - al_borland 3y ago
    I think the issue is that basic ACC doesn’t understand intent and the larger picture. If someone merges in and I have control, I can usually just let off the gas a little bit, as they will likely speed up, or move out of the lane shortly. Actual braking is almost never required. But the ACC just sees some slow moving object inches from the front of the car and flips out to try and get back to a good state.
- FabHK 3y ago
  AKA Moravec's paradox. > Moravec wrote in 1988, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility". > Steven Pinker wrote in 1994 that "the main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard." https://en.wikipedia.org/wiki/Moravec's_paradox https://en.wikipedia.org/wiki/Moravec's_paradox
  - seanmcdirmid 3y ago
    https://xkcd.com/1425/ https://xkcd.com/1425/ Although that has obviously changed in recent years.

al_borland 3y ago

In an interview with Boston Dynamics, one of their robotics engineers said something like, “the hard stuff is easy and the easy stuff is hard.” There are things that humans can learn and do extremely quickly, or figure out on the fly, that robots require extensive training for.

usrusr 3y ago
Stuff like checking the rear view mirror when slamming the brakes upon some surprise uncertainty, true. Humans will always be bad at this, but for a computer it's trivial to include the tail situation in the brake decision.
- al_borland 3y ago
  I think the issue is that basic ACC doesn’t understand intent and the larger picture. If someone merges in and I have control, I can usually just let off the gas a little bit, as they will likely speed up, or move out of the lane shortly. Actual braking is almost never required. But the ACC just sees some slow moving object inches from the front of the car and flips out to try and get back to a good state.
FabHK 3y ago
AKA Moravec's paradox. > Moravec wrote in 1988, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility". > Steven Pinker wrote in 1994 that "the main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard." https://en.wikipedia.org/wiki/Moravec's_paradox https://en.wikipedia.org/wiki/Moravec's_paradox
- seanmcdirmid 3y ago
  https://xkcd.com/1425/ https://xkcd.com/1425/ Although that has obviously changed in recent years.