We've had pretty good prototypes for more than 20 years. There are two huge barriers:
(1) The progress in computer vision, like voice recognition, has slowed. People had pretty good speech recognition since the 1980's, but the gap between "pretty good" and "good enough to use" remains uncrossed. The gap in computer vision for cars will be even wider because mistakes will be much more costly.
(2) Even with all the tech problems solved, the regulatory hurdles to implementing this are staggering.
Suffice it to say, I would make a "Long Bet" (longbets.org/) against autonomous cars making up 5% of all cars in 20 years, with 10:1 odds.
> Additionally, what would be those massively positive effects?
Car accidents cost the US about $250 Billion per year, with 45,000 killed. Autonomous driving has the potential to decrease this by at least an order of magnitude (most accidents are due to driver error). That's a mind-bogglingly staggering benefit.
10 million Americans commute more than an hour each day, which is $25 Billion per year at minimum wage. There's much more lost among the rest of the commuters, and a factor of 4 to account for realistic wages.
I could go on about how autonomous driving would also (a) drastically reduce the total number of cars we'd need, (b) cut emissions and gas usage by allowing ubiquitous slip streaming, (c) virtually eliminate the need for nearby parking in cities, freeing up vast amounts of extremely valuable real estate.