It's the classic induced demand phenomenon - when roads are expanded, people drive a lot more, and congestion remains the same. Same thing with transit - people

It's the classic induced demand phenomenon - when roads are expanded, people drive a lot more, and congestion remains the same. Same thing with transit - people ride it up to the point where it's miserably crowded. Both are signs that our infrastructure in either form is unable to service the maximal demand society places on it, so people ultimately decide not to go places because of the cost of doing so (in terms not just of cash, but also comfort and time). I think people will use rideshares only to the point that congestion is precisely miserable enough those who choose to rideshare to tolerate it. So what might happen is that mass transit remains equally as congested, just more people are taking trips than they were before because the capacity of the whole system has increased. It's also the case that, unlike self-driving car infrastructure, trains are far more scalable - in places that actually invest in their mass transit infrastructure, I think mass transit can compete pretty well the more the municipality invests in train lines and stuff. But to assume that self-driving cars will save cities that don't have transit infrastructure—that might be true for very sparsely populated, small cities, but for all the American cities that lack public transit infrastructure, I think the situation will only change marginally.