AI That Provides A Running Narrative Commentary While Driving Could Be The Tell All For Imbuing Safety Into Self-Driving Cars
Trying to teach your teenage son or daughter how to drive a car can be quite a dicey activity.
The usual method seems to be that you let your teenager sit behind the wheel of your car and you sit in the passenger seat, during which you proffer sage advice to the newbie about how to drive. The thing is, you are pretty much in the hands of your own offspring at this juncture in the sense that there is little you can constructively do if the teen goes awry inadvertently.
Some parents believe that if needed they could reach over and take control of the steering wheel. This is a possibility but it is not a sure thing. The teen might not relinquish control of the steering wheel or might be panicking and countering whatever steering you might try to desperately apply. In addition, you don’t have direct access to the brake pedal and nor the accelerator pedal, thus your only means of trying to affect the driving controls is via grasping at the steering wheel.
Truthfully, you have a minuscule chance of averting a dire car incident by a contortionist attempt to physically overtake the driving effort.
Otherwise, the mainstay of what you can do is tell your teen what to do.For anyone that has ever tried to tell a teenager what to do, you know this can be extraordinarily problematic. The teen might abide by your wishes or might be rebellious. Whatever you say might be completely ignored. What you say might go in one ear and out the other, or might be entirely misunderstood and the teen will try to do what they believe you had said to do. The whole interaction can be incredibly messy.
Of course, there’s more than just messiness on the line. You are both inside a multi-ton vehicle that has the immense capability of high speed and humongous physical forces. Something really bad can happen if the teen makes a mistake and rams into another car or a pedestrian. A parent sitting in the passenger seat is on pins and needles, as it were, hoping upon hope that the driving trek with the newbie will go well and safely so.
Some parents are admittedly lousy at being the passenger that is supposed to provide verbalized assistance to their teen driver. I’ve seen some parents that berated their child the entire time, as though that is going to make the teen become a good driver. You can usually bet that if the teen and the parent are already on rough terms, the added stress of the driving chore is not going to suddenly make them into the best of buddies. All of the prior angst gets carried straightaway into the driving lessons.
Indeed, some parents will opt to have just one of the parents take on the duty of serving as the driving instructor. Whichever of the parents seems to have the best rapport with the teen is usually tapped for this somber duty. Another selection criterion is the ability of the chosen parent to remain cool and calm while providing driving advice. If a parent is yelling and screaming frantically at the teen driver, you can bet that the driving lesson is not going to be especially productive.
Some parents opt to outsource the driving lessons. This is handy since the third-party driving instructor is presumably going to be steady and proficient in guiding the teen driver. The odds though are that eventually the teen and a parent are going to get into the car and that the teen is going to have to showcase their driving capabilities in front of the parent. In that sense, the use of a driving instructor is more akin to a leg-up rather than a replacement for the parental effort of driving guidance.
We can ratchet up the pressure by noting that the parent is not necessarily correct in whatever they might advise the teen driver. You might assume that the parent knows perfectly what to do while driving and they will perfectly instruct the teen accordingly. Not necessarily. A parent sitting in the passenger seat can misgauge what the driving situation consists of. They might falsely believe that the teen is about to hit a car to their left when in fact there is sufficient room for the two vehicles to pass each other without touching.
Okay, so you have a parent that is doling out remarks and commands, and you have a teen that is trying to cope with the driving scene and simultaneously placating their parent. That’s a lot for a teen to handle.
Some parents even use a silent treatment approach, whereby the parent remains completely quiet the entire driving journey unless there is some moment that is so catastrophic that verbalization is required. The thinking is that the teen already has enough sensory overload and thus the verbalization from the parent should not be that last straw on the camel's back of overtaxing the newbie.
What about the teen doing verbalization of the driving task?
Yes, that’s an approach that some parents invoke.
Upon starting a driving trek, the parent and teen agree that the teen will explain what they are doing. I am starting the car, I am driving down the driveway, I am making a right turn up ahead, etc. Those are the means of the teen soothing the qualms of the parent. The parent doesn’t have to wildly guess at what the teen is going to do. By having the teen plainly state what is happening, the parent can decide whether the driving is going well or the teen is maybe clueless about what they are undertaking.
Some refer to this as performing a running commentary while driving.
Commentary driving is a handy technique.
The person doing the driving is articulating aloud what they are aiming to do. A parent can then detect whether the teen is mindfully doing the driving chore. For example, if the teen doesn’t report that a dog is about to run into the street, the parent that has already observed that the dog might do so can then readily alert the teen. Without the commentary taking place, the parent has almost no idea what the teen is seeing, thinking, planning to do, or trying to do.
A significant problem underlying the activity of getting the teen to do a running commentary is that this can be a colossal distractor for them. The teen is already seemingly concentrating with all their being on the roadway and trying earnestly to cope with the terrifying intensity of the driving effort. Having to also explain what they are doing can be an added burden. Some assert that this takes a notable chunk of the thinking activity away from the act of driving per se and devotes it instead to verbalizing what is taking place.
Not a good idea, some claim.
On the other hand, a quick counterargument is that the act of actively generating commentary gets the teen further engaged in the driving task. When the newbie is entirely silent, their mind might not be as engaged. By being forced to explain what is taking place, one belief is that this rivets the attention of the teen and makes them even more so concentrated on the driving activity.
In short, there is the commentary associated with the driver and there is a commentary associated with the parent that is acting as a type of driving instructor. That can be a lot of commentary flying around inside the car. A parent might interrupt the commentary of the teen, due to wanting to point out an urgent matter such as nearing closely to a bike rider. The teen might be talking over the parent since the parent is seemingly blabbering and the teen wants to show that they know what they are doing.
Ugh, the crosstalk and communication woes are potentially troublesome.
Another variant of using commentary while driving entails doing something long before a teen gets behind the wheel of a car. I tried this with my kids. The idea is that you as an adult driver provide a running commentary while driving the car, doing so when the kids are in their pre-teen years or at least early teenager years.
Each time that you take them for a drive, you verbalize what you are doing. You ask them to watch the roadway and pretend that they were driving the car. At some points of a driving trek, you might inquire of the teen as to what they would do if they were at the driving controls. In a sense, you are simulating them driving the car.
You can easily come up with mentally engaging questions. Should I make a left turn or a right turn up ahead? Would it be best to go slower and avoid that truck or try to speed around it? What do you think that red car to our right is going to do next? And so on.
This provides at least a twofold advantage. First, the teen is hopefully garnering the driving act via your verbalizations. You are explaining what you are looking at and how you are handling the driving controls correspondingly. Second, the teen is engaged by the pretense of doing the driving and this gets them to carefully consider what driving consists of.
I sincerely believe this can be a handy exercise.
That being said, if the parent and the youngster are already on bad terms with each other, this could somewhat backfire on everyone involved. The youngster might decide they will do the absolute opposite of whatever the parent is saying. Or the youngster might come to hate the notion of driving, simply because the driving act got mired into the sour relationship between them and the parent.
Anyway, there is no question that the use of commentary while driving is a handy tool that can be used when the circumstances are appropriate to do so.
This brings up an interesting consideration in light of where cars are heading in the future.
The future of cars consists of AI-based true self-driving cars.
There isn’t a human driver involved in a true self-driving car. Keep in mind that true self-driving cars are driven via an AI driving system. There isn’t a need for a human driver at the wheel, and nor is there a provision for a human to drive the vehicle. For my extensive and ongoing coverage of Autonomous Vehicles (AVs) and especially self-driving cars, see the link here.
Here’s an intriguing question that is worth pondering: How might the use of commentary while driving be an instrumental factor in the advent of AI-based true self-driving cars?
I’d like to first further clarify what is meant when I refer to true self-driving cars.
Understanding The Levels Of Self-Driving Cars
As a clarification, 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 vehicles are considered Level 4 and Level 5 (see my explanation at this link here), while a car that requires a human driver to co-share the driving effort is usually considered at 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 contend, see my coverage at this link here).
Since semi-autonomous cars require a human driver, the adoption of those types of cars won’t be markedly different than driving conventional vehicles, so there’s not much new per se to cover about them on this topic (though, as you’ll see in a moment, the points next made are generally applicable).
For semi-autonomous cars, it is important that the public needs to be forewarned about a disturbing aspect that’s been arising lately, namely that despite 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 vehicle, regardless of how much automation might be tossed into a Level 2 or Level 3.
Self-Driving Cars And Digital Commentary Driving
For Level 4 and Level 5 true self-driving vehicles, there won’t be a human driver involved in the driving task.
All occupants will be passengers.
The AI is doing the driving.
One aspect to immediately discuss entails the fact that the AI involved in today’s AI driving systems is not sentient. In other words, the AI is altogether a collective of computer-based programming and algorithms, and most assuredly not able to reason in the same manner that humans can.
Why is this added emphasis about the AI not being sentient?
Because I want to underscore that when discussing the role of the AI driving system, I am not ascribing human qualities to the AI. Please be aware that there is an ongoing and dangerous tendency these days to anthropomorphize AI. In essence, people are assigning human-like sentience to today’s AI, despite the undeniable and inarguable fact that no such AI exists as yet.
With that clarification, you can envision that the AI driving system won’t natively somehow “know” about the facets of driving. Driving and all that it entails will need to be programmed as part of the hardware and software of the self-driving car.
Let’s dive into the myriad of aspects that come to play on this topic.
First, some assert that self-driving cars ought to include a driving commentary feature or function overall.
The notion is that a human passenger can find out what the AI driving system is doing. This might help human passengers feel more confident in what the self-driving car is undertaking. You can liken this to the earlier discussion of having a parent ask their teen driver to explain what is taking place during a driving journey.
So far, most of the driving commentary for self-driving cars is nearly entirely visually based via the use of a display or onboard computer screen.
The human passenger can watch the display and see a visual indication of what roadway objects are being detected. Via the use of a range of colors and shapes, the display can overlay where the self-driving car is headed. This can also include blinking indications when something untoward is detected such as another car that is illegally running a red light in front of the self-driving car.
Most of these commentary driving aspects are visually portrayed and only sparingly verbalized. The oral explanations are usually only at key points of a driving trek. Using an Alexa or Siri kind of verbalization, the AI driving system might emit a sentence or two that you are nearing your destination or that there will be a delay in reaching your destination due to heavy traffic.
Some believe that the human passenger should be able to activate a more verbose version of the commentary driving. You might want the AI driving system to report on every itsy-bitsy act it is undertaking. For some passengers, this could be especially reassuring, particularly if they are unable to readily see the display or want to concentrate on the roadway rather than looking at the display.
Some naysayers suggest the verbalization might scare passengers more than comfort them. If the verbalization doesn’t seem to cover all the myriad of driving scene aspects, the skittish passenger might become overly worried that the AI driving system is unaware of what is happening. As such, the provision of explaining the driving act should be kept to a minimum.
I’ve covered these matters in my columns, including at this link here.
We’ll have to wait and see how this plays out.
The odds are that some automakers or self-driving tech firms will include a rather complete commentary driving capability, while others will restrict their own versions. In the end, it will likely be the human passengers that will determine what is provided since the fleet operators are going to want people to use their self-driving cars and will undoubtedly seek to provide capabilities that the riding public desires. For more on how getting people to use self-driving cars is not necessarily going to be an easy slam dunk, see my analysis at this link here.
Shifting gears, slightly, there is a somewhat different reason for the commentary driving of self-driving cars that have been percolating and gradually rising in interest.
Suppose that we could use commentary driving as a means of inspecting the AI driving system and do so to try and ensure that the system is performing the driving chore successfully. Proponents of this approach argue that it could lead to self-driving cars achieving greater safety versus without this type of capacity.
All told, we might require self-driving cars to emit a Digital Commentary Driving (DCD) indication.
The DCD would not necessarily be a vocal verbalization. Instead, the DCD would be a data stream that indicates what the AI driving system is undertaking. By using a data stream, you can avoid the problems associated with having to do verbalization. It is in a sense cleaner and more revealing to have the raw data than to have the vocalized transformation that has been using the data as a basis for making the resulting utterances.
By and large, the AI developers associated with budding self-driving cars are doing something like this already. They tend to use various internal technical tools to get data dumps and find out what the AI driving system was doing during a driving journey. This helps in debugging the AI driving system and can also be instrumental in making modifications and improvements too.
For those that are already doing this DCD, the odds are that it is entirely proprietary. The type of data, the data elements, the values associated with the data elements, are entire of an idiosyncratic nature. Whatever the AI developers happened to come up with, that’s what their digitalized commentary contains.
A call to standardize the DCD is being bandied around. This would make life easier for being able to compare different self-driving cars from the various automakers and self-driving tech firms. By having a single standard, everyone would be emitting the same kinds of data and using the same predefined sets of values.
The kicker is that this could be used for regulatory purposes.
Regulators that want to try and ensure that a particular model of a self-driving car is suitable for public roadway use might require that the automaker or self-driving tech firm provide copious amounts of DCD that could be reviewed. The review would go hand-in-hand with the actual roadway results. In the end, a kind of roadway readiness certification might be issued.
A recently released report by the BSI Center for Connected and Autonomous Vehicles (CAV) provides an overview sketch of how a DCD standard might be initiated (the report is entitled “A Review of CAV Safety Benchmarking and a Proposal for a Digital Commentary Driving Technique” and was co-authored by Professor Nick Reed of Reed Mobility, Bryn Balcombe of ADA, Paul Spence of PKSInsights Ltd, Dr. Siddartha Khastgir of WMG, University of Warwick, and Nick Fleming of BSI, Head of Sector Transport & Mobility).
Here’s a bit of an overview of their vision: “Commentary driving is a technique used to train and assess human expert drivers, in which they are required to verbalize relevant information in the driving scene. This is used to determine that they can perceive, prioritize and act effectively when driving. Digital commentary driving (DCD) is proposed as an objective measure of CAV safety performance. DCD does not entail a CAV verbalizing or describing the driving scene in the same way as human drivers, rather, it involves the continuous collection of data from a CAV on its perceptions, decisions, reactions, and feedback whilst driving. This data effectively probes the ‘understanding’ the CAV has of its environment.”
The report points out that not everyone is going to be thrilled with this DCD type of regulatory requirement if one is so divined.
Probably the most notable concern is that the DCD could potentially reveal the inner secrets of the AI driving systems. An automaker or self-driving tech firm that has invested millions or even billions of dollars into their development efforts could be rightfully queasy that the DCD is going to give away their secret sauce. The Intellectual Property (IP) rights issues of devising a DCD standard will be onerous as an attempt is made to avoid veering into IP qualms and yet still ensuring that the DCD is revealing something substantive.
A watered-down DCD is probably not going to achieve the lofty goals of what the driving commentary is intended to produce.
A slew of other concerns is generally less defensible.
For example, there might be a decrying that the shunting of data out of the AI driving system while in the act of driving could end up diluting the use of onboard computing resources. Presumably, you want every ounce of the limited computer cycles to be devoted to making sure that the AI driving system can fully operate. By adding the data export aspects, perhaps this usurps those revered onboard resources.
The counterargument is that you can seemingly devise a pragmatically feasible means to keep the data export aspects from somewhat distracting or overtaxing the onboard processing. Sure, there might be some tinge of an impact, but the aim of attaining the greater good of trying to ensure the overall safety of the self-driving car outweighs that minimal impingement.
One potentially tricky consideration is whether the DCD data exportation might reveal something of a private nature related to the passenger use of a self-driving car. I’ve noted previously that we are going to enter into an age of the “roving eye” associated with the advent of self-driving cars. This is roving eye notion is a catchy moniker that I use to indicate that we are facing a tradeoff between what self-driving cars beneficially provide and the privacy intrusions that might also result (see more at the link here).
Conclusion
There is a somewhat more insidious reason that some might balk at using a DCD approach.
If you aren’t fully convinced that your own AI driving system is up to snuff, you certainly would not want it to be baring its soul.
You would prefer to keep things quiet and only known internally. The aspect that you might be forced into revealing the inner workings and perhaps showcase to the entire world that there are incorrect or feeble driving capabilities would strike fear into your heart and those of your investors.
Nobody in their right mind would say this aloud. They would instead stand firmly and overtly on the stance that the highly valued and earnestly earned IP is going to be undercut by the DCD revealing efforts. It might be hard to discern when the resistance due to a legitimate concern about IP is actually hiding the fact that the AI driving system is not yet adequately adroit enough for public roadway application.
But this is something that can potentially be dealt with by adequately defining a standard for DCD that can sufficiently protect the IP considerations.
A final comment for now.
If you got into a car with your teenage newbie driver and they refused to explain what they were doing, this would abundantly give you pause for concern. A parent that continued anyway and let their offspring drive, well, that’s either an amazing act of pure devotion and unbridled love toward their wayward teen or something that you might regret the moment that the teen makes a wild turn or perchance runs through a red light.
Being a parent sometimes means employing tough love, including at times asking their teenage driver to provide a semblance of driving commentary (as long as doing so is not overly distracting). As an added plus, the words that your son or daughter utters will always be a memorable part of those days that they first learned to drive and you were there to witness and guide them in this most life-changing of skills.
Post a Comment