That rarely works out how you would think:
They are terrible, and always will be. It's simple math: buses have to either have long intervals, or be horribly wasteful (and thus too expensive).
At the moment, the most realistic solution is to just wait 10 years for the self-driving cars to mature and then build infrastructure around them.
Something in between that and buses might be very efficient and practical, especially for things like work commutes, visits to the supermarket and the like. Having say 6 or 8 seats and picking up multiple parties who are heading in the same direction or going to roughly the same destination. It would likely be quicker than waiting for a single trip car and definitely cheaper. Additionally it would likely reduce the number of autonomous vehicles on the road and use less resources.
Extending Waymo to handle ALL possible situations autonomously (you put in "New York" in the address field and the car drives you there from Seattle) is probably a too big ask, but working within a pre-mapped area on semi-fixed routes seems plausible.
Induced demand is a sign of success: it allows more people to move comfortably between points, and it increases the overall economic activity by inducing more trips.
The only reasonable part of that argument is that lane expansion doesn't increase the average _speed_. And this is indeed true. Instead of lane expansion we ideally should just increase the interconnectivity by building new roads.
In the case of the Puget Sound, I-605 that would bypass I-405 bottleneck is sorely needed.
Yet the distance between Seattle and Portland is the same as between Frankfurt and Munich.
There's simply not enough traffic to justify a train. Another lane on I-5 would help a lot more.