By Dr. Lance B. Eliot, the AI Trends Insider
The other day I took my car into my local dealership for an oil change and some other minor maintenance work. Upon doing so, the dealer looked up my car in a special database and discovered that there was a recall on my transmission. The dealer asked me why I had not brought the car in sooner, since the recall was about a year old. I protested that I had no inkling that there was a recall involving my car. Further inspection of the database revealed that a letter from the auto maker was sent to me via the good old US postal service, but the address on file was an outdated one. Undoubtedly, the recall notice went to that address and the person there tossed it away as so much junk mail.
Little did that person know that they might have sealed my fate. Suppose the recall was a very serious and imminently endangering fault or flaw in my car? I could have been driving along on a beautiful countryside winding road, and all of a sudden my transmission gives out. Next thing you know, the car goes nuts and no matter what kind of evasive action I take, the car barrels into a fence and strikes a herd of cows. Well, maybe that’s a bit dramatic, but you get my drift. The aspect that the car had a recall was important for me to know, and likewise doing something about the recall would usually be prudent, since it is a matter of the safety of the car and therefore the safety of those driving the car or being occupants in the car.
Here’s a staggering statistic for you about automotive recalls. There were about 53 million car recalls in the United States last year. Think about that for a moment. Since there are about 200 million cars in the United States, the stat about last year’s recalls means that nearly one-fourth of all cars were encompassed by a recall. Another way to envision this would be to look at your car and if there are three other cars parked next to your car, one of those four cars has a recall on it. Ouch! That’s a lot of recalls. There were about 1,000 recall campaigns last year, meaning that the auto makers identified about a thousand separate recalls and for which the total number of cars impacted was the 53 million cars that came under the recalls.
Sometimes an automotive recall is widespread, while in other cases it is relatively narrow.
Let’s review some of the famous and most widespread recalls. Probably the one that we all have heard recently the most about involves the Takata airbags. This case involved faulty airbag inflators. The danger associated with the fault was that the airbag could rupture upon being inflated, and then it would potentially spew metallic fragments at the driver and occupants of the car. Imagine shrapnel from a bomb, and that’s about what it was doing. The recall started in 2013 and involved nearly 70 million cars. Part of the reason that so many cars were involved was due to the aspect that over 20 auto makers had opted to use the Takata airbags in their vehicles. The number of recalls can get pretty high when the component being recalled is something that multiple auto makers have decided to use in their cars.
A notable recall that involved a smaller number of cars but that got tremendous attention involved faulty ignition switches in various GM (General Motors) cars. Investigations showed that the ignition switch could slip out of the normal engagement mode while the car was actively running and abruptly jump into accessory mode. Doing so would cause the engine to shut down, along with cutting off power, and led to hundreds of people suffering injuries or deaths due to the fault occurring at the wrong time in the wrong place. This recall involved “only” about six million cars. The lethal nature of it and the fact that it had occurred repeatedly made this recall especially notable. In addition, when it became known that the defect existed, there was a big scandal when it was discovered that GM had tried to hide the problem and had not taken proper and prompt action about the recall.
According to the US governmental agency known as the National Highway Traffic Safety Administration (NHTSA): “A recall is issued when a manufacturer or NHTSA determines that a vehicle, equipment, car seat, or tire creates an unreasonable safety risk or fails to meet minimum safety standards. Most decisions to conduct a recall and remedy a safety defect are made voluntarily by manufacturers prior to any involvement by NHTSA. Manufacturers are required to fix the problem by repairing it, replacing it, offering a refund, or in rare cases repurchasing the vehicle.”
In one sense, you could say that there is a bit of a game that is played by auto makers.
They are supposed to voluntarily be proactive and try to identify when a recall is needed, and then take proper action about the recall. Currently, the NHTSA regulations indicate that “Manufacturers will notify registered owners by first class mail within 60 days of notifying NHTSA of a recall decision. Manufacturers should offer a proper remedy to the owner.” In theory, an auto maker will do the right thing, they will dutifully be on the watch for faults or problems that involve doing a recall, they will quickly act to undertake the recall, they will try earnestly to contact those impacted by the recall, and they will ensure that there is a remedy that can be readily applied for the recall.
Of course, not all auto makers will act in such an idealized manner. Some might not be watching for faults that are a sign of a need for a recall. Some might ignore faults that are brought to their attention. Some might try to do a cover-up and act like there isn’t a fault and thus no need for a recall. Keep in mind that an auto maker will be looking at quite a cost to deal with a recall, since they will need to provide a replacement or fix to however many cars are impacted. This cost is bound to hurt their profits. Furthermore, they realize that the mere act of announcing a recall can also hurt their sales, including sales for the car models directly impacted, along with all of their other car models since consumers might believe that all cars from that auto maker are faulty.
If an auto maker drags their feet about a recall, on the one hand it could be handy for that auto maker in the short-term since it delays the potential impact of the recall from a cost and immediate public relations blowout perspective. But, if they get caught about having done little or even covered-up, there are chances that it could become an even worse public relations nightmare and even a costlier issue. There could also be the potential for criminal charges against the company and those that knew about the dangers and did nothing or covered-up the matter. And, let’s not forget about the loss of lives that can occur because of a faulty aspect on a car.
What does any of this have to do with self-driving cars?
At the Cybernetic Self-Driving Car Institute, we have been pointing out that there are going to be automotive recalls involving self-driving cars. This sometimes catches some of the self-driving car pundits by surprise. They seem to be living in a nirvana world that includes the fanciful belief that self-driving cars will never break down and never have any faults or issues. It is both disheartening and scary that there are some so fervent over self-driving cars that they actually seem to believe that self-driving cars are going to be utterly error free.
It’s a crock.
First of all, a self-driving car is still a car. By this I mean that it still has all of the normal aspects of a car, including that it has an engine, a transmission, tires, etc. All of those are components that today are subject to faults and failures, and will continue to be subject to faults and failures. We are going to have recalls on those aspects of self-driving cars. There isn’t some magic fairy dust that just because the car is a self-driving car that all of those automotive parts are going to never fail. They will fail.
Next, we need to consider the specialized components that are going to be in a self-driving car. There will be lots of new hardware involved, including LIDAR devices, radar devices, sonar devices, cameras, and so on. This adds lots and lots of opportunity for mechanical and physical faults and failures. In essence, we can anticipate that a self-driving car is likely to have even more chances of a recall than a normal car, due to the addition of all of this nifty new hardware.
Another aspect will include the hundreds of likely microprocessors needed to underlie the AI and smarts of a self-driving car. We could have recalls impacting those microprocessors. Many of those microprocessors will be very specialized and have been made for the specific purposes of aiding a self-driving car. In that sense, they will not have stood the test of time in terms of being generalized microprocessors that have been in say our mobile phones and laptops. Their specialization will make them relatively unique, new, and generally untested in the field.
There is also the opportunity for a recall on the software of the self-driving car. There could be defects in the software that need to be fixed or replaced. Think about all of the defects in Microsoft Windows and you’ll know what I mean when I say that we need to be realistic and assume there will be lots of defects in the software that runs and controls a self-driving car.
For the software recalls, I realize that many of you are likely saying to yourself that those can be fixed “easily” by doing an over-the-air replacement. Just as Tesla today does fixes for its car software by pushing new updates directly to the car in an Internet-like way, so too we should anticipate that most self-driving cars will be able to equally be fixed remotely in terms of the software.
This is both a yes and a no. Yes, in theory, the auto maker should be able to push patches and updates to the self-driving car and thus not need to physically have the self-driving car come to a place to have the software fixed. The no is that doing so will require remote access to the self-driving car, and there might be instances whereby the remote access cannot occur or is limited. For example, suppose the self-driving car is in a location that does not lend itself to remote access. Or, suppose that the communications components of the self-driving car are failing and it cannot do external communication. Admittedly, those will be rarer circumstances, but I am just pointing out that there are exceptions to the notion of readily doing software updates over-the-air.
Some potential twists that can help with recalls are indeed possible in an era of self-driving cars.
For example, I had mentioned in my story earlier about my transmission having a recall and that I was not aware that the recall even existed. This was due to the auto maker sending a snail mail notification to me, which I never received. With self-driving cars, presumably the auto maker can notify me more directly via my self-driving car. In essence, the auto maker informs my self-driving car, and it then informs me upon my next encounter with my self-driving car, such as when I get into it to go to the grocery store, it might verbally tell me that there is a recall on the transmission or whatever.
You could go even further and anticipate that the self-driving car might take itself in for the recall. Suppose that I am not going to be using my self-driving car for the next day or two, and it might then self-drive itself to the dealership. At the dealership, they replace the recalled part, and then they send the self-driving car back to me. All of this being done without my having to drive the car. Instead, the self-driving car goes to get the recall part replaced or fixed, and then drives itself back to me. That’s nice!
There’s another interesting aspect too that we’ll likely see.
With GM’s Chevrolet Bolt electrical vehicle, they recently were able to use their over-the-air telematics to remotely detect a battery related problem in some of the vehicles. It turns out that some of the Bolt EV’s were misreporting the battery levels and causing human drivers to believe that there were more miles available to go than were actually available (the Bolt has an advertised range of 238 miles on a single full-battery charge). Not all of the Bolt EV’s were having this issue. Normally, GM would have had all of the car owners take the car into a dealership to diagnose whether their particular car was one of the ones impacted. Instead, in this case, they were able to remotely detect which cars had the specific problem.
For situations wherein it is a software specific problem, as mentioned previously, presumably a software update can be remotely applied. Suppose though that it is a problem of a hardware nature?
In some cases, it might be possible to adjust the software (doing so remotely) to accommodate a hardware fault. If the ignition switch can slip from one status into another, such as the example mentioned earlier, suppose the software was updated to deal with the situation. Maybe the software could override a physical movement of the ignition switch, and decide that if the engine was running and the car was in motion that it made no sense to suddenly switch into the car accessory mode. Thus, a software related “workaround” might be possible when dealing with certain kinds of hardware or physical related faults.
We need to be mindful though that if a self-driving car has various software workarounds to contend with various hardware faults, it might not be prudent over the long term to just allow those hardware faults to continue to exist. It could be that the software workarounds are short-term measures to keep a self-driving car viably in action. You might think of this as the same as having run-flat tires, yes, you can still drive a car while using run-flat tires when you have a flat, but you should replace the tires with proper ones when you get the next chance to do so. Likewise, for some kinds of car parts that are faulty, even if the software can act as a workaround, it might still be needed to ultimately replace the faulty part.
Another aspect to consider about self-driving cars will be the safety of the driving of the car, under circumstances of parts recalls. When a human is informed about a recall for a conventional car, they might decide it isn’t safe to drive the conventional car, and therefore have it towed to a dealership to make the needed repairs. For a self-driving car of a Level 5, will the AI be able to make a similar kind of determination? If it alone is informed about a recall, is it able or even right that it would somehow make that decision? Furthermore, suppose an AI driving self-driving car tries to drive even though the fault has serious potential repercussions?
If that seems a bit like a quandary, it is one that we definitely need to figure out and not just allow auto makers to decide arbitrarily. By the way, once we have self-driving cars, I would anticipate we’ll have self-driving tow trucks. Thus, your self-driving car, which maybe should not be driving itself to the dealership for a recall, could have a self-driving tow truck that comes over, hooks up to the self-driving car, and takes it over to the dealership for you. That seems like a nice way to deal with things, avoiding having to get involved as a human per se. But, I point out, those same self-driving tow trucks might also be prowling neighborhoods and city streets looking for illegally parked cars, then automatically tow those to the police impound. Imagine that those self-driving tow trucks can do this twenty-four hours a day, seven days a week. For those of you that are scofflaws about where you park, this might not seem very dreamy.
Anyway, the emphasis here has been that we will have recalls that impact self-driving cars. They are cars. They will be subject to parts that are badly manufactured or otherwise have some kind of issues or faults. Furthermore, with the added components to make the car become a self-driving car, the odds of having a recall goes up. Those components, especially the hardware ones, are bound to also have some kinds of faults or failures. I don’t want to seem like the bearer of bad news, but the notion that there will be no recalls is out-the-window, and even that there might be less recalls seems somewhat farfetched, though the aspect of communicating about recalls will certainly be improved as will at times the ability to workaround a recall or have the self-driving car itself go in automatically to take care of a recall.
This content is originally posted to AI Trends.