Tesla released video of a car driving itself
tesla.com
tesla.com
Winter is here, however, I'm looking forward to roads covered with snow, slush, ice, and animals. I'm looking forward to winds in excess of 30mph driving snow across the highway, limiting visibility to a hundred feet or so. I'm looking forward to inexperienced folks driving on ice and snow and being a good 20mph below the rest of the traffic.
People can barely handle these roads (a two day snow storm last week put dozens of cars and semis in the ditch); I'm not sure how a Tesla will handle them. What will it look for when there are no lines on the road to be seen? How will it know the difference between slush, powdered snow, and plowed snow (which is only slightly less hard than a concrete barrier)?
I'm sure the answer is currently "don't allow automatic driving", but these kinds of conditions can reign over half the year in the mid-west; what value is a self driving car which can't half the time?
Remain optimistic.
For every successful technology, another ten technologies have fallen by the wayside due to cost, complexity, or unknown unknowns. I firmly see self driving cars in that category until a fundamental shift in our ability to program occurs.
My ultimate fear is that the programs that will be responsible for making hundreds of decisions every second are still being written by people like us. Not NASA engineers, who will spend millions of dollars verifying all of their code; who spend years writing a few thousand lines of code to an incredibly tight spec.
Instead, they are people who are looking at data and going "that's our integration testing right there." Who write big, complicated bits of code which takes several seconds to render text on our screens and think "Yeah, that's good enough." Who look at accidents their cars have been in and go, "That's totally the driver's fault," instead of saying, "Here's how we will stop that from ever happening again."