" How many people do you need to be in a place to detect a traffic jam? If even 5-10% of people are willing to share the information then you'll still get 90% of the benefit"
Those "benefits" frequently have tradeoffs.
The road I live on (1/2 mile long, 30mph speed limit, 3 schools, 2 housing estates and a 70mph dual carriageway bypass road paralleling) had a major problem with night time speeding (80k cars per week, 60% speeding, 95% at night exceeding 40mph, 5k/week over 50mph). On the other hand it got majorly congested for 90 minutes at a time in the morning and evening.
Cue the council ignoring resident complaints about speeds and rolling out changes "to reduce congestion" - including so many parking restrictions that there are now 3 times as many houses as car parks.
We now have 100k cars per week, with even more night time speeding and the daytime snarlups last for 2 hours each morning and afternoon.
The rat runners who use the road outside of peak periods are prone to road rage incidents too. There have been a number of incidents of pedestrians on one of the three pelican crossings on the road being nudged by impatient drivers and I've witnessed a lot of drivers screaming abuse at anyone who dares slow them down (mostly women being aggressive too...).
My pick is that automated driving will reduce vehicle ownership levels and density (which will reduce the number of streets clogged up with parked cars) and remove the problem of 'entitled' types treating the roads like their personal racetracks even in peak periods (which will reduce speeding and congestion issues respectively, given that the vast majority of holdups are caused by drivers who won't be polite causing everyone else to have to stop or slow down to deal with their selfishness). We may even be able to reclaim our residential street as a residential street where it's safe for children to walk, with routing algorithms designed to treat ratrunning as the antisocial behaviour that it is.