We'll want high-performance driving skills in self-driving cars.
Are you a good driver?
Your answer depends upon what the phrase “good driver” means.
Across the United States, each state provides various licensing requirements for those that want the privilege to drive a car.
Keep in mind that driving is considered a privilege and not some form of constitutional right. If you abuse the driving privilege, such as by committing a transgression against prevailing driving laws, your privilege to drive can be revoked.
The ability to get a driver’s license is so easy that those in their mid-teens can qualify to drive in most states, though there might be restrictions about late-night driving and other facets.
In any case, when you ponder the aspect of obtaining a driver’s license, as a society we have generally agreed to allow a minimum level of driving skills for anyone that wants to be able to legally drive a car on our public roadways.
Novice teenage drivers are often given a wide berth by other drivers that realize there is a newbie at the wheel. Those newcomer drivers are apt to brake suddenly or make a dangerous left turn, and otherwise are still figuring out how to control and properly drive a car.
Driving a car is a serious task.
You have at your command a multi-ton vehicle that can harm others and harm yourself. If you aren’t paying attention, you can run over a pedestrian or ram into a cyclist. If you aren’t comfortable with making tight turns you could swing wide while making a right turn and bash into a car next to you. And so on.
Over time, people get used to driving and usually become better at it, but there’s no way of knowing how many people are especially good at driving versus those that continue to struggle at driving. If you perchance can avoid hitting people or smashing into a pole, you are considered by silent admission to be a good enough driver.
We are a nation of apparently good enough drivers (there are about 225 million licensed drivers in the United States).
Would we have the 40,000 car crash deaths per year and the estimated 1.2 million car-related crash injuries though if we were all possibly better drivers?
It’s hard to say, but one would certainly assume so.
Car crash incidents are typically associated with drivers that are inattentive, or that are intoxicated, or that have some other matter undermining their driving prowess. We don’t quite know whether those drivers would do a better job at driving and have avoided those incidents if they had stronger driving skills.
Of course, you could argue that if they did have stronger driving skills, they would possibly be more attentive and less likely to get behind the wheel in an untoward condition. Perhaps drivers that have a greater sense of how to drive might undertake driving even more seriously than they might otherwise.
Plus, there are certainly some notable number of car crashes that could have been averted if the driver did have a more robust driving skill-set.
For those situations that involve having the car slide on a slippery road, or when taking a turn that the driver lost control of the vehicle, it could be that a driver with a more advanced skill-set could have avoided the resulting calamity.
Indeed, imagine if all drivers had advanced level driving skills, akin to say a high-performance car driver.
All drivers would be proficient in driving skills that could be leveraged to avoid car crashes and thus aid in lowering the volume of lamentable car-related deaths and injuries that occur each year.
Not only would you be better able to avoid your own car mishaps, it would mean that there is an increased chance that you could avoid getting entangled with some other driver that has gone awry.
Presumably, all drivers would be safer drivers due to having an improved ability at driving a car, and it would also mean that when some driver does happen to falter, other nearby drivers might be able to get out of their way.
Unfortunately, we don’t require everyday drivers in the United States to be versed in advanced driving skills.
If you can keep out of driving trouble, and upon periodic renewal of your driver’s license you are able to retake successfully the minimum-level driving test, you can continue to drive in your ongoing muddling manner.
No need to comprehend the principles of tire adhesion and how your tires are crucial to staying on the road. No need to know how to best deal with a car that goes into a spin. There’s no requirement to understand how weight transfer occurs during acceleration, and nor realize how your use of the gas pedal and the brake can simultaneously function in a coordinated manner to gain greater control of the car.
Sadly, our society has decided that you don’t really need to know how to fully drive a car, at least not to the degree that a high-performance driver does.
You could argue that this unwillingness to up the ante on statutory driving requirements makes some sense from an overall economic perspective. The nationwide cost of having every driver take an advanced driving course and be forced to pass it to become a licensed driver would be a doozy of a price. You’d probably see people go up in arms over the added requirement.
One notable thought is whether the estimated national cost of $871 billion that annually goes toward car accidents could be reduced by requiring today’s drivers to be more proficient in driving. And, let’s not forget that the dollars cost doesn’t include the lost lives and disrupted lives due to injuries from car crashes.
Okay, so we know, and can hopefully all agree that we have a preponderance of drivers that are minimally “good enough” to drive, and likely very few that have advanced driving skills akin to a high-performance driver. It seems apparent that any attempt to push those human drivers to a heightened skill level would be arduous and quite unlikely.
It seems like an unsolvable problem.
Wait a second, what about the advent of driverless cars?
Here’s a worthwhile question: Would we be wise to ensure that true self-driving cars have high-performance driving skills as an integral part of their AI driving capabilities?
I say the answer is yes, very much so.
Let’s unpack the matter and see.
The Levels Of Self-Driving Cars
It is important to clarify what I mean when referring to true self-driving cars.
True self-driving cars are ones that the AI drives the car entirely on its own and there isn’t any human assistance during the driving task.
These driverless cars are considered a Level 4 and Level 5, while a car that requires a human driver to co-share the driving effort is usually considered at a Level 2 or Level 3. The cars that co-share the driving task are described as being semi-autonomous, and typically contain a variety of automated add-on’s that are referred to as ADAS (Advanced Driver-Assistance Systems).
There is not yet a true self-driving car at Level 5, which we don’t yet even know if this will be possible to achieve, and nor how long it will take to get there.
Meanwhile, the Level 4 efforts are gradually trying to get some traction by undergoing very narrow and selective public roadway trials, though there is controversy over whether this testing should be allowed per se (we are all life-or-death guinea pigs in an experiment taking place on our highways and byways, some point out).
Since semi-autonomous cars require a human driver, the adoption of those types of cars won’t be markedly different than driving conventional cars, though there are some upsides and downsides worth considering.
Sure, Level 2 and Level 3 capabilities can aid a human driver and potentially enhance their driving skills, but the human can still overtake the driving controls and easily undermine the ADAS.
Furthermore, many of the ADAS features can be turned off by a driver, which some assert should not be allowed since presumably the ADAS is there to bolster the driver.
As I’ve repeatedly exhorted, we are headed into dangerous territory whereby human drivers won’t fully comprehend what the car’s automation is doing, and this lack of a theory of mind when you have a co-sharing driving arrangement, namely a human driver and a piece of automation that’s not sufficiently sophisticated, portends that there will be severe gaps in understanding that are going to lead to car accidents.
For semi-autonomous cars, it is equally important that I mention a disturbing aspect that’s been arising, namely that in spite of those human drivers that keep posting videos of themselves falling asleep at the wheel of a Level 2 or Level 3 car, we all need to avoid being misled into believing that the driver can take away their attention from the driving task while driving a semi-autonomous car.
You are the responsible party for the driving actions of the car, regardless of how much automation might be tossed into a Level 2 or Level 3.
Self-Driving Cars And High-Performance Driving
For Level 4 and Level 5 true self-driving cars, there won’t be a human driver involved in the driving task.
All occupants will be passengers.
Today’s efforts toward developing true self-driving cars are nearly entirely focused on the everyday driving task.
People want a driverless car that can take them over to their local grocery store. This involves successfully navigating neighborhood streets without hitting any children or meandering dogs and gently getting the passengers over to the mall.
Even freeway driving is considered a somewhat everyday driving task for the AI, namely that the AI needs to be able to get onto and off the freeway and stay properly within the lanes while on the freeway.
This is tantamount to what any average “good enough” human driver can do.
Should the aim be higher?
It’s an easy argument to contend that certainly, we would want the AI to be more than simply a good enough driver.
If the AI could be infused with a high-performance driver skill-set, imagine how much better off we would all be.
Presumably, the driverless car would be better at dealing with situations such as slick roadways or situations wherein another car veered suddenly toward the self-driving car.
Some automakers and self-driving tech firms are so overwhelmed with getting the everyday driving nailed down that trying to venture into advanced high-performance driving is considered an edge problem. Edge problems are a kind of industry parlance referring to driving scenarios that are categorized as at the edge of the solution space being solved, often also called corner cases.
Any reasonable AI developer for driverless cars would agree that it would be a nice-to-have of embodying high-performance driving skills, but they would also wince and say that we need to first crawl before we can walk.
In other words, the prevailing view is that once self-driving cars can be the equivalent of a human “good enough” driver, we can then turn our attention toward more advanced aspects such as high-performance driving skills.
The counterargument is that by not including high-performance driving skills now, we are gradually allowing onto our roadways a herd of cars that are unable to drive at a peak level.
Will the passengers inside a driverless car realize that their self-driving car is not a “topnotch” driver and only an “everyday” driver?
There are lawyers already lining up to make the case that if a self-driving car gets into a car accident, and if it had not been infused with high-performance driving skills which might have avoided the crash, the automaker or tech firm ought to be sued for not having appropriately skilled the AI.
A presumed defense would be that the AI contained a necessary and sufficient skill-set to drive on public roadways, and in the desire to make progress on self-driving cars, it was a necessity to start there.
The counterclaim might further be extended that if there had been an attempt to include high-performance driving skills, it would have delayed the advent of driverless cars.
It would be straightforward too by the respondent to point to the other driverless car makers and note that their self-driving cars by-and-large also lack high-performance driving capabilities.
We’ll have to wait and see if those defensive postures are adequate.
The general public might not buy into such arguments and be disappointed and outright disturbed to discover that the vaunted driverless cars they’ve been riding in were not at a level beyond everyday drivers.
Taking a slightly different angle on the topic, those automakers or tech firms that opt to implant high-performance driving skills into their driverless cars can leverage the facet by pointing it out as a strategic differentiator.
If you believe that all self-driving cars will essentially be alike, meaning that consumers and riders won’t distinguish any one type of driverless car from another, it might be notable for a self-driving car maker to emphasize that their AI is a better driver or at least more skilled at driving.
Of course, this differentiation might also raise people’s eyebrows as they suddenly become aware that apparently some driverless cars have advanced driving skills, and some don’t.
Regulators might be prodded into action to establish requirements for the AI driving skill-sets that are to be used in driverless cars.
Conclusion
Don’t misinterpret these remarks to suggest that there aren’t efforts underway of adding high-performance driving skills into self-driving cars.
There are indeed those kinds of efforts taking place.
In fact, the recently released NOVA documentary entitled “Look Who’s Driving” (see my review here) briefly showcased such an effort taking place at Stanford University. Stanford Professor Chris Gerdes and the team at their Center for Automotive Research (CARS) lab have been exploring the inclusion of high-performance driving skills into self-driving cars.
And, let’s be clear, most of the automakers and self-driving tech firms have on their To-Do lists these kinds of skill sets and are eager to pursue those elements.
It’s a question of trying to crawl before you walk, or to some degree perhaps mixing the two together, the crawling and the walking. You see, the definition of high-performance driving skills is so loose that you could argue that some of the presumed everyday driving skills being imbued into driverless cars are in the same ballpark as high-performance driving.
There are some added twists involved in imbuing high-performance driving skills into the AI.
If you just toss the high-performance driving skills into an existing set of everyday driving skills, you could end up in a situation of dueling driving actions. When a slippery road is encountered, the everyday driving skill-set might refuse to handover to the high-performance skill-set, and the driverless car either fails to leverage the added feature or maybe hands-over control belatedly.
In that sense, the high-performance skill-set needs to be seamlessly implanted within the overarching AI driving skill-set.
Another difficulty involves the classic Trolley Problem that faces driverless car makers. The Trolley Problem refers to an ethical debate about how and when the AI ought to make driving decisions that involve life-or-death choices (see my detailed explanation here).
Here’s how that applies.
Suppose the high-performance driving capability could whip around the driverless car and potentially avoid slamming into a wall, perhaps coping with a patch of black ice on the roadway. If there are passengers inside the driverless car, the act of making such a radical movement could injure them, causing a whiplash effect. On the other hand, slamming into the wall is going to injure them too.
Which is the right choice, make the radical move or allow the car to smash into the wall?
Many such scenarios can be crafted that are not perhaps so clear cut as to which choice is the better one.
One final twist for now.
Should the human passengers be allowed to invoke the high-performance driving skill set of the AI system?
You are in your nifty sports car that is driverless. Deciding to go up into the mountains, you wonder how it might do on the upcoming mountainous roads. You tell the socio-behavioral Natural Language Processing (NLP) element of the AI driving system to go ahead and punch it.
Hey, you think to yourself, let’s see what this baby can do, and invoke the high-performance driving capability to speedily take those curves and give you a thrill ride.
As a society, do we want passengers to be able to activate such a mode?
It’s an open question.
Well, overall, I’ve tried to bring to your attention that self-driving cars are eventually going to have high-performance driving skills, which bodes well for us all, though the timing of getting those skills into the AI and making sure that they work appropriately will be a tough road to travel.
Say, is that driverless car over there peeling out and shredding it’s tires?
Could very well be.
