The powerful impact of the spoken word.
I was in a local park the other day and someone had let their dog off its leash, allowing the frisky pooch a chance to run around wildly and relish its newfound freedom. At one point, I saw that the canine was about to dart wantonly into the street where cars were zipping along, so I yelled out to the dog and called for it to come back towards the trees and grassy area. Thankfully, the pooch heard me and scooted away from the dangers of the busy byway.
Later that same day, I was walking along on the sidewalk in my neighborhood and saw up ahead a car that was backing out of a driveway. Meanwhile, seemingly unbeknownst to the driver, a child was riding a tricycle down the sidewalk and was inevitably going to end-up directly behind the car, for which I doubted that the driver would realize, and thus there was an imminent and painful encounter about to occur. The driver's window was slightly rolled down, so I yelled out to come to a stop, and the driver did so. Thankfully, no one was hurt, and the youngster continued peddling unabated.
What do these two incidents have in common?
Sometimes the barking out of a verbal command can make a difference, a big difference, including the possibility of choosing between life and death.
Both of the incidents became non-incidents in the sense that nobody got hurt and the world kept revolving without anyone getting injured or killed. Since these non-incidents did not produce fatalities or severities, they pretty much would be considered mundane and unworthy of any news media attention. No one would ever know that the yelling out of a type of command or instruction had saved a child and likewise saved someone's beloved pooch. Just an ordinary activity that had extraordinary and joyful outcomes.
There are some noteworthy characteristics underlying these verbal commands.
First, the command or instruction had to be short and sweet, quickly getting to the point and being abundantly clear cut as to the meaning of what was spoken. This was not the time or place to provide a lengthy soliloquy or quote eloquently from Shakespeare.
The verbalization had to be done with a semblance of strength and conviction. Doing so allows the receiver of the oral indication to realize, nearly immediately, something of great importance that requires their rapt attention. Furthermore, the vocalizing has to be directed towards the subject or target, otherwise, the intended receiver might not sense that the words are aimed at them. Imagine that if I had yelled straight up into the air the resulting confusion on the part of the receivers. They would not likely have realized that my messaging was specifically for them.
An act of this kind also tends to occur in a real-time situation whereby there can be little or no delay. If I had waited to yell out my commands there was a heightened chance that the driver would not have stopped in time, and similarly, an increased chance that the dog would have entered into the street and been in the midst of grave danger.
All told, the verbalization has to happen at the right time, in the right place, aimed at the right receiver, worded in the right way, spoken in a loud and commanding tone, and proffer a succinct and actionable warning or recommended action that will prevent a pending and looming disaster from occurring.
Wow, that's a lot of stuff jammed into a brief moment of time and without any room for error or delay. And yet it seems likely that this kind of activity happens all the time, all around us, and we do not particularly give it much notice or credit. Sure, from time to time there are stories about someone that heroically saved a life by a verbalized command, but by-and-large this is an unheralded act and one that goes on daily without fanfare or limelight.
The other side of the equation about uttering such verbal commands is that the targeted receiver has to be receptive to the utterance.
Consider the instance of the driver backing up his car. Had his driver’s side window not been partially open, I doubt that he would have heard my exhortation. Most cars these days have car interiors so well-built that any noise from outside of the vehicle is essentially muted. The odds are that my yelling would have been entirely rebuffed by modern-day soundproofing. I shudder to think of what would have taken place and how chilling it would be to know that my attempts to avert calamity were for not.
Of course, the receiver of the verbalization has to also be willing to listen and react to what has been stated. The driver of the car could have completely ignored my command to stop the vehicle. He might have thought that I was referring to someone else and opted to ignore what I was saying. Or, he might have heard me and understood my instruction, yet mentally calculated that stopping is perhaps not what he should do.
Believe it or not, there have been circumstances wherein someone was told to stop, and instead they slammed their foot onto the accelerator rather than the brake pedal. This use of the accelerator might be completely unintentional and was a mistaken reaction of missing the brakes and inadvertently landing on the gas pedal. Another possibility is that the driver gets scared at someone yelling stop, perhaps worried that they are going to get robbed or mugged, and deliberately decided to use the throttle to rocket away from the predicament.
Speaking of drivers, consider the future of cars and what will occur as self-driving cars gradually become prevalent on our roadways. For true self-driving cars, there is no need for a human driver. If you've ever seen pictures or videos of true self-driving cars, it is somewhat eerie since there isn't anyone in the driver's seat.
This brings up an interesting question: How will the advent of AI-based true self-driving cars impact the ability to bark out commands at a car in hopes of forewarning or advising a driving action in real-time?
Let's unpack the matter and see.
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 a 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 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 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 Spoken Exhortations
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.
Some people seem to believe that the AI is omniscient and able to drive a car in a miraculous manner, including being able to always avoid car crashes and avert the striking of pedestrians. Perhaps this false impression about the AI is due to science fiction movies or possibly due to self-driving proponents that either exaggerate things or outright misstate what the AI can do.
In any case, please be aware that today's AI is not sentient, it lacks entirely any semblance of common-sense reasoning and otherwise is arguably closer to a monkey-see-monkey-do than it is to the full gamut of human level of intelligence. Also, it is crucial that we prevent the misleading portrayal of AI as though it does have human qualities, which can lead to a lot of confusion and adverse consequences, so all in all let's stop anthropomorphizing the existing spate of AI.
The point herein is that you cannot expect the AI of a self-driving car to always avoid running into someone or something. I've repeatedly stated in my columns that the notion of zero fatalities due to self-driving cars is malarkey and has a zero chance of happening. There will still be car crashes and sadly deaths and injuries, though the hope and expectation are that it will be a lot less than the existent annual 40,000 fatalities and some 2.3 million injuries in the United States alone from car accidents (see my indication about the major driving statistics at this link here ).
This emphasizes that there is still room for the use of spoken commands to alert a driver when a pending endangerment is about to happen.
Were somehow magically the AI to on its own always and perfectly be able to avoid any car collisions, presumably the providing of any external warnings would not be warranted. I'm saying and declaring that those external commands are still going to be valuable and that we should not delude ourselves into thinking that the AI is all-knowing.
Recall that earlier I had mentioned that the receiver of the verbal commands has to be receptive to hearing the spoken words, else the act of yelling or providing the instruction is null and void. For the car driver that I alerted about the toddler on the tricycle, he heard me due to his window being partially rolled down and he apparently was able to hear what I exhorted.
What about AI?
You might be pondering whether the AI of the self-driving car is going to be a willing receiver of externally urged instructions. Well, similar to the case of the human driver, some key elements have to come to play for this to happen.
One crucial aspect is whether the AI will even hear a spoken command that has been urgently barked from somewhere outside of the vehicle.
Self-driving cars are typically including internal audio capabilities such as microphones to be able to hear the passengers that are inside the self-driving car. The AI is using contemporary Natural Language Processing (NLP) that is slightly adapted to interact with riding passengers. Currently, this NLP is extremely crude and a bit like the stilted interaction you might have with Alexa or Siri. As such this interaction is very limited and usually focused on obvious facets such as where a passenger wants to go. In the future, it is anticipated that the AI will be much more socio-conversational and be able to engage in quite extensive dialogue related to the driving task and the wishes of the passengers (see my predictions at this link here ).
Few of the self-driving cars are being outfitted with exterior-based audio microphones.
Doing so is currently considered relatively unnecessary and an added and unneeded cost. The use of external listening is overall ranked as an edge or corner case, suggesting that it is a matter that is a low priority and might someday rise on the list of things to do. When you are knee-deep trying to get an AI driving system to simply drive from point A to point B, doing so without hitting anything, the prospects of coping with listening to the outside world offers little added value in comparison to the other meat-and-potatoes capabilities.
Some self-driving cars are adding external audio capabilities such as speakers that will allow the AI to speak at people, such as telling pedestrians to stand away from the curb or advising them to get out of the street when in the path of moving cars. Also, these audio capabilities are including microphones. The most prominent use to-date is to listen to the sounds of sirens coming from ambulances, fire trucks, police cars, and the like. As you know, human drivers are supposed to be listening for such sirens and then are supposed to pull over, though it seems that more and more human drivers are ignoring the sirens (a quite dreadful trend).
For those self-driving cars that do not yet have any external microphones, those brands and models have little chance of hearing any externally spoken commands or instructions. There are small odds that the interior microphones might pick-up an externally uttered human command, which is kind of what happened with the human driver (he was sitting inside the vehicle at the driver's seat), but this is highly unlikely and also not being programmed with any prevalence anyway.
One supposes that a passenger riding inside a self-driving car might hear a person that yells out a command that is directed at the vehicle, and then the human passenger might repeat or attempt to relay the command to the AI system. This is fraught with difficulty. The person inside the self-driving car is bound to be startled at the externally yelled instruction and therefore not be alerted to what is happening, especially since they aren't driving the car and they naturally assume there is no need to be conscious of the driving actions.
Even if the passenger hears the spoken command, there is a significant chance of a pronounced delay between them hearing it and then opting to repeat the verbal command to the AI driving system. Keep in mind too that the passenger is not sitting at the driving controls. The rider can only implore the AI about the driving act and cannot themselves directly use the driving controls. For now, the approach of having an externally exhorted command become relayed via an internal passenger is essentially untenable and relatively unlikely.
Bottom-line: Barking commands at a self-driving car is not going to do you any good, not today, though perhaps someday in the future this feature will be provided.
In the instance of the human driver backing out of their driveway, if the vehicle had been among the current crop of self-driving cars, I would have been better off yelling at the toddler rather than the car.
You might also have noted that when I yelled to try and stop the dog from running into the street, I opted to call out to the dog rather than the cars. Why? I could see that the cars were all moving fast, and their windows were rolled up. Trying to yell to those drivers would have been futile. My best bet was to hope that the dog would care about a human yelling at it, despite my not being the actual owner.
All of this suggests that whenever you see a self-driving car that is about to get into trouble, the odds are that you'll have better luck at proffering a verbal command to an endangered human or animal rather than to the AI of the vehicle.
That's the current state of affairs.
Do not lose hope.
As mentioned, there will eventually and in my opinion inevitably be self-driving cars that are extensively making use of external microphones and actively listen to the world around them. I would anticipate that all self-driving cars will ultimately leverage such technology.
There are added twists though that need to be given due consideration.
Suppose that when I yelled to the human driver to come to a stop for avoiding the toddler, I was somehow mistaken. Perhaps it would have been better for the driver to quickly accelerate and get out of the way of the oncoming toddler (this is a bit ridiculous in this instance, but you can readily envision situations whereby the verbal command is askew of what needs to be done).
If an AI driving system hears a spoken command, should the AI obediently abide by the command?
Your initial answer might be that yes, the AI ought to always obey humans. Period, end of the story. But this belies the chances of a human that has unintentionally provided an incorrect command. Or perhaps the human is being dastardly and purposely wants the AI to drive adversely or do some other devilish act.
I believe we would all reasonably agree that we expect the AI to assess the given command, even when a command has been expressed directly by a human being. The possible downsides are just too high and risky to not evaluate whatever might perchance be spoken. We certainly expect a human driver to do so, such as the driver backing down the driveway. I am assuming that he momentarily thought about my exhortation and then decided it made sense to come to a sudden stop. Indeed, no matter what I might have said, we generally would all concur that the driver had the final say in the matter and would be held responsible for the driving actions undertaken.
Assuming you concur that the AI would need to first assess any spoken command, you have now taken a turn into an abyss or morass. This opens a veritable can of worms.
When should the AI unquestioningly do what it is told versus possibly countermanding or ignoring a human uttered instruction?
Can we actually hold the AI responsible for the driving of the car, as though it is the legal or moral equivalent of a person responsible for the act of driving?
And so on.
One of the most fascinating aspects about self-driving cars is not merely the mobility that can be had via the use of AI driving systems, and thus no longer needing to have a human as a driver and enabling a mobility-for-all future, but this also brings up extremely important questions about AI and ethical considerations (for my columns about the rising realization of AI & Ethics , see the link here ).
We might ask overall questions about AI that is being used for deciding on the granting of home loans or the AI that helps you to figure out how to invest your money, though those matters are not especially life-or-death per se. The AI for self-driving cars is in fact about life-or-death. We already know that every time a human gets behind the wheel of a car, the human is entrusted with life-or-death choices, for themselves and other nearby drivers and pedestrians. The same absolutely happens when we let AI autonomously drive a car.
Besides the excitement and delightfulness of seeing self-driving cars on our roadways, we also need to somberly consider how the AI is doing the driving and how it will be making life-or-death choices. That's a routine part of driving, yet the repercussions are far from routine.
Let's keep the AI from barking up the wrong tree, doing so by all of us listening to each other about how to best devise AI for self-driving cars.
That's verbal command or instruction that we can all live with.
- Waymo invites Arizona residents to apply for limitless, free rides in its self-driving cars
- See A Car Crash From the Perspective of Google’s Self Driving Car
- How Self-Driving Cars Will Change Transportation Forever
- Google is testing self-driving cars in a third city: Kirkland, WA
- Amazon, Sequoia invest in self-driving car startup Aurora
- Google veterans head off on their own to work on self-driving trucks
- This startup sees the highways filled with self-driving big rigs
- Self-driving spacecraft could protect Earth from asteroid impacts
- Volkswagen and Nvidia want to imbue future cars with artificial intelligence
- Top car trends of CES 2018
- Uber and Lyft Induced Congestion Give a Preview of Driverless Car Hell
- 15 awesome flying taxis and cars currently in development
- Study Says Driverless Electric Cars Could Cut Emission by 90%
- Should you hope your child never has to drive a car?
- The Craziest Cars and Futuristic Vehicles of CES 2019
- Norwegian Armed Forces See Security Issues With Self-Watching Tesla Cars
- Best Buy slashes prices on car electronics and GPS devices for Cyber Monday
- Why car companies are trying to imitate Uber and Lyft
- 2019 Volvo XC40 first drive review
- No, Low Gas Prices Are Not the End of the Electric Car
Barking Commands At A Self-Driving Car Won’t Do You Any Good have 3582 words, post on www.forbes.com at December 13, 2020. This is cached page on Europe Breaking News. If you want remove this page, please contact us.