I was sitting in the passenger seat of a VW wagon, fiddling with a laptop.
My driving buddy, a Stanford researcher, looked puzzled. We were in a parking lot, and he had set up a couple of pylons to demonstrate how self-driving car technology worked. The car moved slowly, creeping along like a bug on a windshield, even though the parking lot was totally empty. "Must be a glitch," he muttered under his breath.
This was back in 2007, and we barely carried smartphones.
Earlier that summer, Apple released the first iPhone, and I remember using the first Android models that barely worked. Autonomous driving seemed like an exciting idea back then--sensors that scan all around you, on-board processing that determines when to stop and how to park. My friend Jamie was in the backseat, taking photos, and we both joked that the technology seemed like a robotic hive-mind.
That was well over ten years ago. I wrote an article about the experience for a magazine called Business 2.0 that doesn't exist anymore. I didn't have a Gmail account yet, because it was still a fairly new service and most of us insisted on using Outlook. Back then, we used the term "cameraphone" as though taking photos with a phone was still a novel idea.
Seriously, though. George W. Bush was President. My kids were in grade school. I had a lot more hair. The promise of autonomous driving seemed imminent, almost assured. But the truth is that self-driving cars might not become commonplace for another ten years.
A recent crash that killed a pedestrian, setbacks with Tesla and others, delays in building out the infrastructure--it's hard to stay positive when you've been waiting more than a decade, and when you do believe that there is some possibility that cars will drive on their own without any mishaps. Way back in 2007, the experts envisioned a future scenario that involved daily autonomous driving, as though it was second-nature. In that same article, the idea of building highways for autonomous cars still seemed like a distant dream.
I remember discussions about sending the car to pick up the kids across town, and the typical guesses about playing chess, reading the paper, and checking email as the car drove us to work, blissfully unaware of our surroundings and the heavy traffic.
I should have thought a little harder about the problem back then.
In my area, there's a commuter highway that is always crowded with heavy traffic but not quite jammed up bumper-to-bumper on most days. Everyone drives 70 MPH or faster, and it is total chaos. Cars zip in and out of lanes; it's a race to get to work faster than anyone else. Since so many people are in a hurry, if you ever end up in the left lane going a hair slower than the accepted speed, someone will ride your bumper.
We know the human brain is many times more advanced than any computer, and yet human drivers get confused on a daily basis. A sleek BMW slides behind a huge RV, then a Lexus swerves in and out of traffic while a Prius suddenly speeds up faster than anyone expects. It's an algorithmic nightmare. While autonomous cars are constantly analyzing all of these changes, they lack the basic emotional intelligence of even an elementary-age child to really understand what to do or how to adjust.
So how do we get there? How do we move from dream to reality?
One prognosticator suggested that Google and Waymo might have this all figured out in the next two years, and maybe he will be right about that, but the age of fully autonomous cars--driving you every single day to work, in confusing traffic conditions, where the sensors and on-board computers have to process information as fast as a human brain, and with the same assurances about safety--might not happen until 2028.
By then, we might finally realize that robotic cars need dedicated lanes, or even dedicated roads. We'll realize that the only way to allow thousands and thousands of robotic cars to all drive us to work is to put them into a more predictable situation, one that's similar to the highways in Europe where entrance and exit ramps are more rare than in the U.S. Or we'll accept the idea of limiting the speed of autonomous cars at first, perhaps to city streets and tops speeds of about 30 MPH as a way to make sure they don't cause accidents.
Is it depressing? Not really. The technology makes sense in that a bot can scan in all directions at all times, and the human driver can focus on other things. But slow progress with autonomous cars is not a bad thing. Let's do it the right way, and make sure it all works. The alternative is to keep making predictions...and not making progress.