It is very fashionable on certain parts of the Internet to dunk on self-driving cars. According to these folks, self driving cars are yet another Silicon Valley scam with huge out-sized risks being placed on an unsuspecting public in the form of poorly tested (or untestable) software just waiting to kill people, and with all of the benefits accruing towards a cabal of tech bro elites filled with hubris and lacking any concern with the consequences of their actions.

Where this discussion begins to go wrong, to my mind, is by catastrophizing the perceived risks of computer-driven cars while never grappling with the obvious question: is this technology safer than what we have now, and if not, could it be in the future? Are there other, realistic, options that are better? Note: specifically not is this technology perfect in every way or we should just make people not be people and get them to pay attention 100% of the time.

It also can’t be about could we completely re-engineer society from the ground up to eliminate cars? In at least one survey, 90% of people said they had “no choice”¹ but to use their cars to get around, which I guess implies they might want a choice, there are two things that I think are unlikely to be resolved any time soon: 1) the built environment in much of the United States presumes car use and is unlikely to be retrofitted to other modes of transportation and 2) it is very likely there will still be some cars out there even if we do mass-adopt mass transit.

Not to belabor the point, but:

1. Human error contributes to the vast majority of motor vehicle accidents (94%² according to the National Highway Traffic Safety Administration)
2. Cars kill and injure an awful lot of people, every year. In the United States alone in 2022, 42,514 people were killed and 2.38 million were injured³.

This would seem to be a target-rich environment for improvement.

Where these discussions completely depart from on-the-ground reality is by not recognizing that we already have self-driving cars on the road. If a car is moving and a human isn’t actively controlling it, that car is driving itself. Perhaps it is slightly more correct to say we have a distracted driving problem, but distracted driving, at least where I live, is getting to a level where this ends up being a distinction without any difference whatsoever. Regardless, it would be nice to have something, however imperfect, in addition to steering alignment to keep cars from running into things.

Caption: Even the local hobos are sick of distracted drivers

1. I’ve been rear-ended while towing a bright red trailer. The person who hit me “didn’t see” the trailer. They seemed barely aware of their surroundings when I spoke with them afterwards, and as soon as they got back in their car, were right back on their cellphone.

2. A large (3/4 ton) diesel truck rolled across my yard and crashed into my neighbor’s garage. A work crew was loading a piece of equipment on a trailer, and, somehow, the vehicle rolled (under power) away from them. Fortunately, no one was injured.

3. I saw a professional limo driver drive with his turn signal on for over a mile without turning. He eventually turned the signal off, then spent the next 100 yards or so straddling his travel lane and the adjacent turn lane. I guess he had more important things to worry about than controlling his 3-ton-plus SUV.

4. When taking a trip of more than a few miles, it is now more common for me to see an accident than not. Most are minor, but seeing a car wrapped around a telephone pole or in a ditch is not an uncommon sight at all.

Sometimes discussions of self-driving cars get stuck on the idea that there will be nobody to ticket or otherwise hold accountable in the event of an accident, or that it will be some low-level employee that gets rolled under the bus⁴ in the event of an accident. This is either a lack of imagination, or, perhaps more likely, cynicism. Holding, or not holding, vehicle manufacturers to account is, fundamentally, a policy problem, and one that self-driving cars present an opportunity to fix and get right. Right now, drivers (and/or owners) of cars, not manufacturers, are generally responsible for what happens when those cars are involved in a collision.

Cars are fundamentally dangerous things. They are extremely heavy, and move at great speed. Those two characteristics practically guarantee high injury and fatality rates when they hit things. Better vehicle design could lower the injury rate, but not eliminate it. I submit that cars are dangerous and will be dangerous no matter what is done to them. So, it seems to me, that in addition to doing whatever can reasonably be done to make these products safer (like eliminating known-unsafe human drivers,) that the vehicle manufacturer should be liable for selling a product with known and foreseeable dangers. Partially, in the case of a car with a human driver. Fully, in the case of a hypothetical completely autonomous car. It is not at all unrealistic or impossible to have regulation that requires manufacturers of these products to hold warranty reserves and/or insurance sufficient to cover expected liability claims over the life of their product.

This would be infinitely better than the current situation, where people just kind of hope that if they are involved in an accident that the other party has adequate insurance. Will there be bad actors that produce faulty products that are unacceptably dangerous, but figure its cheaper to pay settlements than improve their product? Sure. But, there are bad actors now and society isn’t doing a very good job of controlling them. We could do better. We choose not to. A different system of liability does not necessarily mean less regulatory oversight. Or regulatory capture. That would be the outcome of a deliberate choice to not regulate these products properly, not a fait accompli.

In the United States, autonomous cars also present an opportunity to clean up the absolutely absurd police-state system that has developed over the years in relation to vehicle searches and seizures by law enforcement. Today, since drivers are liable for moving violations, and drivers aren’t very good at avoiding moving violations, they interact with the police quite frequently. These stops frequently escalate into policing over other matters, like alleged property and drug crimes, due to the searches police are either allowed to perform, or searches which they trick people into allowing them to perform. With autonomous vehicles, there would be far fewer opportunities for these interactions, and, perhaps, we could all learn to mind our own business again. This state of affairs, like our system of liability, is, of course, a deliberate choice we’ve made as a society. Autonomous vehicles could give us a realistic path to making a better choice here, one more aligned with our putative values.