Trust and Traffic Lights (and Robots)

Suppose you stop at a traffic light. There are, say, four cars ahead of you. Eventually, the light turns green. When that happens, the first car goes. The second car waits to see the first car go, then it goes. The third car waits until they see the second car go before it goes, and so on. By the time you are able to start moving, several seconds have elapsed from the light turning green -- a lag that results from us double-checking that the car in front of us will actually move when it is supposed to. All those seconds could be saved if we all just trusted each other to go when the light turns green.

Now, obviously, there are many reasons why one shouldn't do this beyond just "trust". There's the obvious possibility that someone won't react immediately to the green light. There's also potential differences in how fast a car would move even if all cars started moving at the same time. And then there's complications if one car was turning instead of going straight.

Still, it does seem to me that robot cars, in communication with one another so they know how fast they'll be going, when exactly they'll start driving once the light turns green, and if any car is turning or doing something wonky, could circumvent many of these problems (if we trusted them to do so, that is). I wonder how much commute time would be saved by that change -- an impossible task for humans, but well within the capacity of a hypothetical fleet of driverless cars.