Self-driving cars are not happening

I started a new job in February of 2015, in an established software group. For my new-hire lunch, we piled into cars and went to a Chinese place. The discussion on the way was about self-driving cars, which were everywhere in the news. There were various guesses about when self-driving cars would be for sale. As a long-time software engineer and manager of software engineers, I was skeptical, as I would be of any new software. I said that I didn’t know when they would appear, but I would be happy to use them, once they were on release 3.2 of the software.

I thought more about the subject after that, and aside for my overriding skepticism about the software, two other practical concerns occurred to me. First, what about the liability associated with this product? Any product short of perfection would invite a swarm of lawsuits. Second, what about the all-important sensors on those cars? Who would maintain them well enough to be the eyes and ears of the near-sentient beast? I only wash my car once a year, and that’s because we’ve been in a drought.

It has been eight years, and numerous cracks have appeared in the edifice. The signs are everywhere, if you are paying attention to this kind of stuff.

The main reason that people crave self-driving cars is that humans are so error prone. Surely intelligent cars could reach a higher safety bar than fallible humans? But maybe that bedrock assumption is a bad one to start with.

The Deadly Myth That Human Error Causes Most Car Crashes

Every year thousands of Americans die on the roads. Individuals take the blame for systemic problems.

It turns out that “human error” is the catch-all bucket when trying to determine the cause of an accident, and this attribution is seldom questioned.

Remember that Elon Musk was one of the early pitchmen for self-driving car technology. How is that going?

Musk’s recent Twitter misadventures, and a nudge and a wink from the movie Glass Onion (spoiler alert: this is one of the best scenes in the movie, so see the movie first), casts his long-time pitch of imminent self-driving technology in a different light, doesn’t it? We were warned at least as early as the “pedo guy” tweet that this visionary genius had more than one type of voice in his head.

Listening to some other self-driving tech prophets would probably be in order, especially someone closer to the technology development itself:

Godfather of Self-Driving Cars Says the Tech Is Going Nowhere

With all the hype around self-driving cars, one key-but-now-disgraced pioneer of the industry says the tech is far, far from being viable.

When trying to figure out where a complex issue is, especially if you are lazy like me, and want to take advantage of someone else’s research, it is good to look at the betting markets. I don’t know if there are any formal betting markets for this (there goes that laziness again), but some very large, important companies are placing, or more accurately, withdrawing their bets:

Apple scales back self-driving car, delays debut until 2026

Apple executives are grappling with the reality that the company’s vision for a fully autonomous vehicle — without a steering wheel or pedals — is not feasible with current technology.

Where it says “delayed until 2026”, read “delayed through the next planning cycle, aka the foreseeable future”.

Finally, the New York Times checks in on some of Tesla’s latest efforts:

What Riding in a Self-Driving Tesla Tells Us About the Future of Autonomy

We took a ride with Tesla’s experimental self-driving system. It was by turns intriguing, impressive and scary.

We are left with the picture, appropriately, of Elon and his hardcore software engineers striving to solve the nefarious “left turn across traffic” problem for one very specific case. A quick calculation on how many very specific cases exist out in the wide world should serve as a warning to us all.

Take special note of this section of the Times article:

Experts say no system could possibly have the sophistication needed to handle every possible scenario on any road. This would require technology that mimics human reasoning — technology that we humans do not yet know how to build.

Such technology, called artificial general intelligence, “is still very, very far away,” said Andrew Clare, chief technology officer of the self-driving vehicle company Nuro. “It is not something you or I or our kids should be banking on to help them get around in cars.”

When I went to college and studied computer science, in the late 70s and early 80s, what is now known as “artificial general intelligence” was just called “artificial intelligence”. It has proven to be a problem so difficult to solve that AI has been redefined as a baby step in that direction – 40 years later. It turns out that humans are full of erroneous processing, but machines are not nearly as good at the on-the-fly processing involved in driving a car.

So let’s give ourselves a round of applause. Or a slow sarcastic clap, if you prefer.