DRIVER-LESS CARS ARE FAKE NEWS
Part 2: The Fake News
Click to listen on Apple Podcasts or Overcast.
For a menu of other apps, click “Subscribe” above. You won't have to subscribe to listen. Or just listen via the player above.
An unedited transcript is below.
Ali Tabibian: Self-driving cars are mostly fake news.
Trump: Fake news for a long time.
Ali Tabibian: But the real parts are big news. In this, part two of our three-part series on self-driving cars are mostly fake news, we focus on the fake news. Part one was about the big news. And part three will be about the real news.
Robot Voice: Part two.
Ali Tabibian: Welcome, welcome, welcome [00:00:30] everyone to Tech Cars Machines, where we make you smarter about how technology advances everything from sensors to wireless connectivity to artificial intelligence is changing industry and transportation. About those autonomous vehicles, or AVs as they're called, evangelists love to quote the global statistic of approximately 1.3 million accidental car deaths a year, from which AVs will save us because they don't get drunk, fall asleep, [00:01:00] text, etc. This statistic is nothing to ignore, but like your parents told you, you have to take a look at the good with the bad. Consider the following. If we replaced all of the vehicles on the road today with self-driving vehicles of current capability and it's pretty mediocre, in a couple hours, all of the world's traffic would grind to a halt because so many of these vehicles would come across a situation out of their operating capability.
Now really, in the far more important and narrow sense, these accidents aren't happening because the car isn't driving itself. [00:01:30] They're happening because the car is hitting things, tens of millions of times a year around the world, in fact. Collision avoidance, sort of a negative control of the vehicle, is technically much easier than partial or full autonomy, which is a positive control of the vehicle. So affordable components that could make a small percentage difference in collision rates make a difference measured in millions of collisions. That's big news without needing any fake news.
Now right before we dive into some tech talk, I want to spend a few minutes on [00:02:00] why things are fake new-ish. I want to start by sharing a career observation. Outside of the technology industry, massive profitability seems to correlate with massive self-doubt. Companies outside the tech industry, they can be extremely rich, but somehow they just never feel sexy enough. They try to buck themselves up by drinking some tech Kool-aid. Sometimes that means having their executives dress awkward when attempting to dress casual, and then sometimes it means betting the farm like Time Warner did when [00:02:30] it acquired AOL during the dot com era. And the number of emails you get these days from AOL accounts tells you how well that went.
So to this self-doubt, by the way, add the allure of technology valuations. Then further add in the click-worthy nature of robo-car articles and you can hardly blame auto company management for succumbing to the hoopla. That being said, the hoopla has really gotten away from people. It's gotten out of hand. In our travels, it's pretty obvious that there's a very, very big disconnect, not just between [00:03:00] public expectations and technical reality, but between autonomous technologists and with their own managements, they're saying. On the one side, we run into the executives, the press, even politicians who are either championing or wringing their hands about how robot cars and artificial intelligence are either going to save us all or kill us all, both of them soon.
Meanwhile, you run into the scientists and engineers, and you have lunch with them, have a drink with them, and the ones working on this stuff are almost emotional about how beautiful and powerful the brain is, [00:03:30] and how difficult it is to match. Now of course they all say, "Yes, yes, yes," when their high ups make pronouncements because nobody wants to be a naysayer. Despite all this, things are slowly coming around to the way we've been saying here at GTK, but since the belief in the inexorable rise of robots, artificial intelligence, blah blah blah, is like a religion, we're gonna take a few minutes to give you some history. And the most important mistake you want to avoid when you study the history of autonomous vehicles is this. You should not conflate [00:04:00] how rapidly digital technology costs come down versus how other industries work.
We're all used to better, faster, stronger and sometimes cheaper. All of it at astounding rates when it comes to computers and gadgets, but a lot of technologies or advanced equipment just don't behave that way. There's a difference between digital technology and a lot of other types of technology. A big passenger jet still costs a couple hundred million dollars. That's a list price, but I never pay retail. And maybe [00:04:30] it's 10% faster and 30% more fuel efficient than 30 years ago. Ditto your car. Yes, it's safer, more comfortable, it'll massage you, etc., but it's not 10,000 times better than it was 30 years ago. Have you bought a pair of glasses? Have you been startled by how much the lenses cost? It's always been that way.
Many of the systems considered necessary for autonomy have similar constraints. For example, most people think that LIDAR, which is kind of a laser scanner is necessary as a vision system for the successful autonomous [00:05:00] vehicles, but just like your glasses, LIDARs have a lot of optics in them. They have a lot of lenses and that makes them expensive. Unless you're using semiconductor based components, your prices are not gonna decline the way prices do in that industry. That's why a lot of LIDAR investment is focused on transitioning these technologies to be based on semiconductors.
Let's start in the late 1980's when MIT and Carnegie Mellon have some awkward looking vans packed with computers and cameras, and manage to [00:05:30] do a pretty good job of off road driving. And I noticed I used the present tense when I just described that to you. That's because I was recently looking at some of the old videos of these things on YouTube. We'll try to put some of the links in the podcast notes. In 1995, a Mercedes S Class, and this is a great one, drove 1000 miles in Europe on public roads. I think it was between Munich and Copenhagen. And this prototype made lane changes, passed vehicles, of course because it's the Germans doing [00:06:00] it, it tested top speeds of 120 miles an hour. Did all sorts of things. Had some famous snafus too, by the way. But once again, it was jam packed with computers and special equipment, and nobody really heard of it again. Although some of the radar technology made it into production S Classes a few years later.
Ten years after that, in the mid 2000's, some things happened that made people wonder if things really were gonna accelerate, namely university teams who had been entering these weird mobile [00:06:30] slash Mad Max vehicles into an annual Department of Defense contest, the DARPA grand challenge for the automobile geeks out there. These teams started making some big jumps on a year-to-year or every other year basis. And meaningfully, it was thanks to the introduction of LIDAR by a company called Velodyne, located here in our neck of the woods.
Now some of these winners went to join Google's auto project, which has since been branded as Waymo. [00:07:00] And since 2009, Waymo has been predicting that these prototypes successes would lead to the demise of the steering wheel in about four years. Now these vehicles, they turned out they were not only autonomous, but they were also stealth vehicles. Only Waymo can see them coming. But jokes aside, despite the tremendous progress and continued status as a leader, even Waymo's ambitions have come down to earth a little bit compared to what they used to be. They're basically pitching a taxi service now, [00:07:30] when seven or eight years ago, maybe five or six years ago, they were demo-ing essentially what looked like a Volkswagen bug without a steering wheel. I don't think it was a coincidence that it looked like a VW bug and the VW was introduced as the Volkswagen, as the people's car. It was a general purpose mass production vehicle.
Let's tie all of this to what's available in showrooms. And I'm gonna switch to a personal example. In 2015, I bought a Mercedes E Class, which at the time was the only vehicle with meaningful limited [00:08:00] self-driving capabilities. And oh yes, Tesla fan boys, it was before Tesla. Now, I know all of you guys think that violates some law of physics and just because I know you won't be able to stop yourselves, we're gonna turn off comments on the podcast website.
In any case, I'm turning that vehicle in for a 2018 model and here's what's interesting. The self-driving features haven't made much progress. There's some automated lane changes, a little smoother operation, partially thanks to integration with the [00:08:30] navigation maps, but that's kind of it. Isn't it interesting that the rate of progress on personally experiencing on a production vehicle, on a widely available vehicle kind of matches what you'd expect when you realize that these prototypes have been around for 30 years? And it's only very slowly that the size and cost of delivering autonomy is being distilled down to the point where it can show up at a showroom near you.
So when you look [00:09:00] at your Teslas and Volvos and Mercedes then in a limited and somewhat nervous fashion, have been doing some limited autonomy stuff for a few years, it's not that surprising when you hear somebody from GM inadvertently let it slip that general purpose autonomous vehicles could be 10 to 15 years away. And this happened mainly because everybody hates Tesla. That's the title of one of our upcoming episodes and newsletters. But in any case, GM was trying to tell off Tesla on its claims about full autonomy was just cameras, [00:09:30] etc., etc. And basically somebody there said, "These guys are crazy if they think this stuff is near term possible at all, especially with Tesla's equipment package." We'll again link to the article on the podcast website. In any case, if you go to GTKPartners.com, subscribe to our newsletter and look at the archives, you'll find it there, too.
Voiceover: Haven't we made you feel like really smart? Let us elevate you to very stable genius. Click subscribe. [00:10:00] Subscribe with the little button in your podcast app, or click the three dots in the little circle. Or visit us at TechCarsMachines.com and GTKPartners.com.