Sheesh. And here I thought it would take 5 years for self-driving vehicles to become common.
By my previous estimate, around 5 million jobs in the US alone could be replaced or severely curtailed by self-driving vehicles (about 3.5 million jobs are driving tractor-trailers). I now think that's a low estimate, considering delivery vehicles, taxis, US mail, school busses, and so on.
The first self-driving tractor trailer hit the road about 18 months ago. Yes, they probably won't work in snow. Yes, they probably won't work in some situations, such as finding and backing into the loading dock. You'll still need humans for those situations.
But for the vast majority of cases, they will work for the long-haul across the US. (If you've ever driven across the US at night, you know that the highways are a never-ending chain of tractor trailers in the right-hand lane.) They don't need down time, they don't get tired, they don't get distracted, they can work 24/7. They can learn from each others' mistakes. They don't need salary or benefits.
This is demonstrably better from a safety and cost point of view, and it takes away a lot of tedious work from humans--giving them more free time--but it'll wreck our current economic system.
We currently have about 170 million workers, and sitting at about 10% unemployment. This one technological advance could push that up to 15%. Economically speaking, 10% unemployment is the beginning of the "this is bad, we should do something" level. We only recently dropped below that number from the great(-est) depression.
(How we deal with illegal immigrants is another big chunk of potential workers that could affect unemployment. Not to make this a partisan argument, but if we *do* have amnesty, it should be done in a layered, progressive fashion with an eye on unemployment so as not to tank the economy. Refugees are too few in number to affect unemployment.)