Actually, a software-driven car could follow cars much closer than (mph / 10) car lengths. That particular rule of thumb is based off of human reaction times.

Actually, a software-driven car could follow cars much closer than (mph / 10) car lengths. That particular rule of thumb is based off of human reaction times. A safety buffer is still a good thing, but software will react several orders of magnitude faster than a person. Also, if software-driven cars become common enough, you will probably get places a lot faster, at least in heavy traffic. Traffic flow could be much better than it is with humans driving. There have been several studies on freeway gridlock traffic that show that small driving mistakes (e.g. slowing down more abruptly than necessary) can butterfly-effect out into massive bottlenecks. And of course if software-driven cars cause fewer accidents, there will be less gridlock due to those as well.