Yes, of course it will be cheaper to pay for taxis (not more taxis) for drunk people than it will be to have cars drive themselves. You're talking about converting the entire US car fleet, developing a non-existent technology, and refining it to the extreme (less than 2 deaths per 100 million miles traveled on average, which is the human driving record right now). It is MUCH, MUCH cheaper to implement a program where you make sure that drunk people leaving bars get into taxis.
>It's obvious to any observer that driving deaths are a major problem, and that an intensive effort to solve this problem is appropriate. We've made tremendous progress over that last 50 years. Let's continue that path.
This is disingenuous. Driverless cars are not part of an effort to reduce driving deaths, that's just how they're being sold. It's just as likely that driverless cars will increase driving deaths, until their safety record is refined over a period of many years and lots of testing (the same way it happened with humans).
Driverless cars are being pursued because tech companies see an opportunity to use their machine learning expertise to turn driving into a commodity service that they control and can make billions off. It has nothing to do with driving safety, and if that were the goal (which, again, it is not, the goal is to make money by turning driving into a service and eliminate the cost of labor for drivers), there are a hundred cheaper ways to do it. If increasing driving safety were a goal that society actually cared about, Tracy Morgan would not have been hit by a truck.