[brians]

If that were true, we’d expect the nonunion carriers to do fantastically. But Frank Lorenzo has broken N unions and driven N airlines under. We’d expect the regional carriers, which often pay under $40k to pilots, to do great—but they fail more often than the big carriers paying $100 to $150k.

But your last point is interesting. AlphaGo Zero suggests that for many situations, the return to more data levels off very quickly—after all, humans all learn to drive reasonably well with a certain amount of training. The early phases may have this self-reinforcing bit, and there may be an early dominant player, but we should expect that twenty years in anybody will be able to make a self-driving car that is as safe as modern air travel.