The road to autonomous vehicles has been longer and rockier than many people expected. Tesla Autopilot has been chronically delayed and overpromised. And of course there was the recent Cruise debacle.

But this revolution is still really eventually going to happen. It’s already happening. There really are (still) cars driving themselves around in San Francisco now. They really do already seem to be safer than human drivers.

And I still hope it does happen. Humans are abysmal drivers and should not be entrusted with the operation of multi-ton heavy machinery. Computer systems can and will be much better and safer than we are at this task.

Most of my urbanist fellow-travelers seem to have adopted the position that autonomous cars are wrong and bad and should just be banned, because they perpetuate automotive hegemony. And I am sympathetic to this view. If I were appointed dictator I would redesign our cities to make private car usage mostly optional, with Tokyo density and Dutch street design. But blanket opposition is wrong, misguided, and counterproductive.

Firstly, we aren’t ending automobile hegemony in a meaningful way anytime soon. Even though YIMBY’s are now winning many battles. Even though we’re making good and steady progress on pedestrian and bicycle infrastructure. We’re still stuck for the foreseeable future with land use patterns that necessitate a lot of travel by car. I think it would be monstrous, given this reality, to fully reject autonomous driving technology that can help us avoid avoid literally millions of deaths and far more serious injuries over the coming years.

But. There’s a hidden downside, and I think recognizing it can point us to a higher synthesis between the “ban them” crowd and the uncritical boosters: the combination of cheap electricity from solar power & genuine autonomous driving is going to massively decrease the cost of driving. The fiscal costs will be dramatically lower, and the time and attention costs will be pushed down close to nothing.

And if the price falls, consumption will rise. I don’t know how elastic the demand for vehicle miles traveled will be, but probably dramatic. If the cost of driving falls by a factor of 10, what would that do under our current regime to vehicle miles traveled? Easily at least double. Imagine the traffic nightmares, and the political pressure it will create to spend hundreds of billions of dollars on even bigger, faster, and wider roads.

In short, while autonomous vehicles, especially electric ones, will be better in many respects and worth supporting even if we’d prefer fewer cars overall, some of the externalities of motordom will actually be supercharged; traffic, noise, road maintenance, and the sheer volume of physical space people will demand be devoted to them.

The solution will have to be to raise the cost of driving. Not even necessarily to raise it above what it is now. Even just to raise it above how disastrously cheap it will soon be.

Of course we could and should have done this a long time ago. The social cost of driving is already exorbitant and the negative externalities are grotesque. But it’s always been a political loser. If you were to try to institute a tax on driving that internalizes those externalities fully, even though you could instantly solve nearly all traffic, your proposal would poll about as well as the bubonic plague.

The key point is that we have an extremely rare window of opportunity in which to act on this. Because right now there’s a broad political coalition opposed to this. The urbanists hate them because they perpetuate car hegemony (and again, I’m sympathetic to this even though I think it’s wrong on balance). Meanwhile the right hates tech companies and loves to vroom vroom & roll coal, and many people are just plain uncomfortable with change.

It’s a delicate balance, how to calibrate regulation for the best possible future here. But I think the answer is actually very clear. Just tax them. A modest per-mile tax on vehicles driving at Level 5 autonomy (i.e., with no one in the driver’s seat), that increases exponentially with vehicle weight and size. Since the taxes will be paid initially by corporate operators who already keep careful track of where their vehicles are, we can probably even make the tax dynamic based on traffic conditions.

Then, in the future, when autonomous driving is hegemonic, we will already have the technocratic infrastructure in place to set the tax level where it will need to be to avoid some really nasty environmental and political consequences.

Toll all roads — modestly, and in a targeted manner, but right now, ASAP — in this brief window of political opportunity.