People leaving their homes less combined with development of high quality public transportation will hopefully greatly reduce traffic and eliminate the need for todays extensive street networks, with the rest being well served by small scale autonomous electric taxis or bikes.
While this is a valid conclusion, the status quo in the United States has considerable inertia behind it and will remain rather car dependent for some time to come. Even if a critical consensus around
densification could be magicked into existence overnight it would take many decades and the expenditure of considerable wealth to re-engineer the urban and suburban spaces in the US to realize.
cars in the future will be self driven
While no trivial volume of development is going into the field and it seems likely that autonomous cars will emerge in the future, this is not yet a certainty. Almost the entire industry - except an automaker whose name rhymes with "-esla" - is betting on LIDAR, which can deliver reasonably-accurate raw
geometry to the navigation AI to provide short-range fine-scale visualization of the environment ... which has been relatively slow to hit requirements for capability and price. Everyone in the field is also realizing just how
hard the problem is outside of the relatively predictable environment of the limited-access highway - seems the human brain has powerful visual-processing and visualization heuristics that billions of hours of data and machine-learning processing have yet to catch up with.
Should this future ever emerge, I expect the model to be more like
aviation - the vehicles will be expensive and operations will be extremely regimented subject to strict certification requirements. I vert much doubt that you're going to be able to install sketchy suspension mods, "chip" a vehicle for better performance, or install smoked tail lights on such vehicles. Whether the maintenance follows an aviation model or not is uncertain, but I expect it to be more regulated than the present status quo.
Initially I expect ownership to be almost exclusively fleets - the purchase cost of the first generations will likely be so high that the only way to make usage affordable will be to minimize downtime and charge users by the hour/mile. Maintenance requirements might also be initially prohibitive as the likes of sensor calibration and the like eventually matures into something predictable that can be baked into more static processes, schedules, onboard diagnostics.