I love new technology as much as anyone. My dad was an inventor, and I spent many fervent hours in my youth absolutely convinced that I could make a perpetual motion machine out of egg cartons and string. But I just can’t take self-driving trucks seriously.

Partly this is because self-driving trucks aren’t self-driving — not in the way that regular people have every reason to understand that term, which is “requiring no human presence in the truck.” They’re not actually going to be self-driving anytime soon, either. In breathlessly enthusiastic news story after ludicrously inaccurate P.R. announcement, the headline is completely undermined by the practicalities.

Here’s a typical example, proudly announcing that “UPS self-driving delivery trucks are on the road.” “The driver and engineer are only on board as a fallback,” assures the narrator. So, wait, what? We now need a driver and an engineer? I thought one of the points of self-driving trucks was that they would consign concerns over the truck-driver shortage to the dustbin of history. If I were to come up with a way to make driving a truck long distances even less attractive than it already is, offering someone the role of “fallback” to a computer would be a splendid place to start.

Picture the scene. You’re snoozing in the cab or eating lunch or Skyping while the computer takes care of driving, but then — quick! There’s a situation that requires human intervention! Before you can throw your fork down, it’s most likely going to be too late. At present, the estimated reaction time of a human truck driver is 0.7 seconds; longer than that and he or she may face liability claims for failing to prevent an accident. Anything that extends that time is surely less than desirable.

UPS and other proponents claim that savings will abound by “likely removing human occupants once the technology has been perfected.” Just how perfect (and cheap) does the technology need to get? At present, if it needs a human backup, it’s less good than a human driver. So wouldn’t you surmise that there’s quite a long way to go?

Another question: In situations where a snap decision has to be made about how to cause least harm, whose interests will this technology prioritize? MIT released a report last year charting the kinds of moral choices people make when confronted with choosing whom or what to harm in a traffic situation such as the famous trolley problem. Such decisions will now be in the hands of private enterprise. Will a computer-guided truck full of Panasonic TVs prioritize the commercial interests of the cargo owner over, say, the lives of a family? What about where a truck driver actively causes an accident in order to save an unconscious car driver? I’m not saying these tricky questions can’t be answered, but they need to be addressed, head on, as it were. And there’s absolutely no sign of that happening so far.

The UPS video urges us to get ready for more robots delivering things, and to “keep your prejudices to yourself.” Sorry, I will not. I’m pretty sure we’re all going to be converted to avatars in a virtual gaming environment created by Sony in the next few decades, freed from cancer and traffic accidents (although not heartbreak). But while the laws of physics hold sway, I retain the right to make a huge distinction among the hazards presented by a drone the size of a dinner plate dropping off light packages, a knee-high R2-like vehicle ambling down the sidewalk, and an 18-wheeler barrelling down the highway at 70 mph with no one inside.

One suggested solution to the self-driving problem is to send trucks in convoy, attached to one another via Wi-Fi, controlled by a single human driver and augmented by computer power. Here’s an idea: How about we physically attach the trucks together? Then we could put one driver in the first, and one in the last. And then, since the combination of trucks and passenger vehicles on the same road is undesirable for safety reasons, why not build a separate road system just for trucks? For long stretches of road where there’s no need to deviate from the route, they could run on fixed ruts, so there’s no danger of swerving off course.

Oh, wait. We already have this technology. It’s called rail – cheaper over long distances and, in many countries, as fast or faster than road freight. Further, it’s arguable that a highly automated and even completely autonomous train is realistically achievable much sooner than a genuinely self-driving truck. What a shame the U.S. government shows no sign of rescuing our elderly, sclerotic railroads as a safer, greener means of moving freight. At least in Europe, they’re trying.

Helen Atkinson is a contributing writer with SupplyChainBrain.