Electronics will eventually supplant humans for most driving duties because they can do it better than we can
Before that happens there are still plenty of issues to overcome. The autonomous vehicles that are already being tested can deal with most of the predictable stuff. Two of the biggest issues left to address are roads that can be seen such as when they are covered in snow and those that are under construction or facing other diversions.
When +General Motors showed off its EN-V concepts at CES in 2011 http://youtu.be/WAP0Bx8m6Fw , Christopher Borroni-Bird, director of advanced technology vehicle concepts explained that fully autonomous vehicles would need to take advantage of next-generation GPS systems with sub 1-meter precision. When that sort locational precision is available, it's conceivable that vehicles could follow a road that has been mapped even if the road itself isn't visible due to the weather.
The second major issue of dealing with road diversions due to accidents or construction could be dealt with by V2X communications. Vehicle to vehicle and vehicle to infrastructure communications are already being tested in the real world including a pilot program in Ann Arbor. GM also has a potential solution that was demoed at the 2011 ITS World Congress. http://media.gm.com/content/media/us/en/gm/news.detail.html/content/Pages/news/us/en/2011/Oct/1017_v2v.html
They are looking at extending the same communications system to smartphones. A pedestrian could use a phone with a transponder as a beacon to warn vehicles of their presence. The same mechanism could be used by police offers or construction crews guiding traffic to alert autonomous vehicles of the diversion.
There are no doubt issues to address before our vehicles drive us around, but they will eventually be solved.
Reshared post from +Lauren Weinstein
From Business Insider // I have not actually experienced a G self-driving car yet (though I have seen them in Mountain View). The issues with accident sites, hand-directed (or yelled!) police instructions, and other unusual situations, I've thought about quite a bit. They suggest an alert driver ready to take over will be necessary for the foreseeable future (at least until there is major smart road infrastructure and integration with emergency services, and most cars, trucks, and the like are autonomous — many years, likely — but what of bicycles and motorcycles also sharing the road? Hmm.) Major advances in AI will impact this of course. It does seem clear that (as the article suggests) the initial major benefits of this tech will be in driver assist and safety systems, and on specially designated roadways. The long lonely stretches of Interstate 5 here in California between L.A. and Mountain View come immediately to mind in this latter context! – Lauren