[IanCal]

> a certain amount of, umm, aggression is needed to make any forward progress

They added a slight "aggression" back in 2011, if other cars aren't letting it out it starts pushing forwards a bit:

> Sometimes, however, the car has to be more "aggressive." When going through a four-way intersection, for example, it yields to other vehicles based on road rules; but if other cars don't reciprocate, it advances a bit to show to the other drivers its intention. Without programming that kind of behavior, Urmson said, it would be impossible for the robot car to drive in the real world.

http://spectrum.ieee.org/automaton/robotics/artificial-intel...

I really don't understand why people think small behavioural differences like that require decades of work. Things that humans find difficult and stressful are likely to be the things that are significantly easier when you have a reaction time measured in milliseconds and full 360 degree vision.

[ghaff]

I hope you're right but the very uneven progress of AI generally--especially when it involves real world interfaces--makes me skeptical. As I wrote elsewhere, even voice recognition is very much still a mixed bag.

[IanCal]

But voice recognition is something we take many years to learn, even when are brains are highly plastic. I find it hard to understand people when they speak often. We disambiguate heavily based on context and expectation, very complex issues.

[ghaff]

Oh, I fully agree. The fact that it's been the subject of so much research and remains only very incompletely solved demonstrates just how difficult a problem it is. I'm just saying that navigating a potentially dangerous machine through a physical world populated with erratic human beings is also very complex and difficult--albeit probably far less so in some contexts (limited access highways in good weather) than others (Manhattan, Boston).