Many people seem to think the arrival of fully autonomous cars is right around the corner; when I say that it will likely take at least 10 years, but probably more like 20, until there will be significant numbers of those on the road they are very surprised.

I admit I recently upped that number from "5-10 years minimum" after seeing the media reaction to recent driver-assist systems related accidents — it became clear to me that the old adage "the autonomous car does not need to be perfect, it only needs to be better than an {average, good, skilled} human driver" is simply wrong, because autonomous car brand X will be regarded - and judged - as if all X cars are one driver. Thus, these cars actually need to be much better than humans in essentially any traffic situation, because it simply won't be acceptable for "autonomous cars of brand X" to kill a few thousand people each year in every country. People would equate this to a single human driver mowing down thousands.

This effect is not necessarily bad, I don't have the inside information, but it's likely (or at least probable) that manufacturers invested in this technology observed the same and drew a similar conclusion.

The question now is whether the car makers make autonomous cars that good, or invest into massive PR campaigning to change public perception towards "each autonomous car is like an individual driver and it doesn't need to do much better than that".

We've seen the latter before. Jaywalking and avenues are just two examples.

Gill Pratt, the head of auto-car R&D at Toyota, has been agreeing with you since he joined in 2015, essentially saying that auto-cars must be better than very good human drivers in all parts of driving, even those that occur rarely. But mastery of those rare events that will be hardest to acquire, since few useful examples are available from which to train. Thus those last few yards to the goal will add years to the development time in subtle ways that will be maddeningly invisible to the public -- until seemingly clueless crashes arise on public roads, perhaps like the one last month which terminated all of Toyota's auto-car testing on public roads until further notice.

Well Jaywalking is only illegal in 10 countries. So PR campaigns won't be enough for that. The benefits of self driving cars as so massive that it will trump anything like that - but we are still a long way away from practical self driving cars.

Self driving cars are not almost there, they are already a reality. Waymo is testing self driving cars hailing passengers without a safety driver since last October and soon it will start a commercial service.