Those Escaped Zebras Roaming Fancy-Free In Maryland Provide Eye-Opening Insights For AI Self-Driving Cars
Get ready for some especially sweet irony.
First, some needed background.
There is a famous saying in the medical field that makes reference to zebras. Turns out that in the 1940s a professor at the University Maryland School of Medicine repeatedly exhorted his newbie medical interns that if you hear hoofbeats behind you, don’t expect to see a zebra.
This was sensible advice.
You see, the clever saying alludes to the fact that horses are by far more prevalent in Maryland than are zebras. Statistically, you would be somewhat foolish to guess that you had heard a zebra. In the same vein, as a medical doctor, you are more likely to presumably come across commonplace ailments and diseases more so than rare ones. Thus, when you see symptoms in a patient and want to perhaps immediately leap to a conclusion that the malady is some rare instance, the odds are that it is the common one.
The popular TV series called House, M.D. provided a continual stream of examples that fit this sage wisdom. The fictional character that served as a kind of medical genius, Dr. Gregory House, was quick to suggest rather wildly unlikely diseases when doing his rapid-fire diagnoses. Of course, the creative storytelling usually had him shown to be right. You could argue that this was merely feel-good efforts by the producers and writers, aiming to gain audience infatuation with the show.
On the other hand, there is a counterargument to be made that since Dr. House was typically only involved in baffling medical cases, there presumably would be a heightened chance that the medical issue was particularly rare.
A type of pre-selection was taking place at the hospital.
The everyday circumstances ran the usual gauntlet and were seemingly instances of everyday kinds of medical problems. The outliers and offbeat instances would presumably tend toward more complex and unusual medical explanations, and those were being routed to Dr. House. Anyway, the overall mantra is that you are best to consider the usual, and only once you’ve found a reasoned basis for potentially rejecting or questioning the usual, you can turn your gaze toward the unusual. This inevitably became a popular credo in the medical profession at large.
Hearing hoofbeats while you are walking around Maryland is more likely to be a horse coming toward you than a zebra.
But wait for a second, maybe not.
We are now up to the sweet irony that I earlier promised would be revealed.
Maryland has been in the news recently due to a bunch of zebras that are roaming the streets of Maryland, fancy-free and footloose, as they say.
Yes, some zebras managed to apparently escape from a private farm in Maryland and decided they would take in the sights of the local towns and cities nearby. One can only imagine how much fun those zebras must be having. After being cooped up on the farm and generally semi-restricted in their movement, they have found themselves enjoying the utterly free-range way of living.
Go where you want. See what you want. Freedom at long last. What a wonderful moment to be a zebra in Maryland!
Admittedly, there are practical realities to be contended with.
Finding food is not that easy to do for a zebra that is wandering the streets. The allure of food is indeed how the striped creatures are likely to ultimately get nabbed by the local authorities. The official animal services department is putting out food at a feeding station and trying to attract the zebras to come there. It is hoped that the zebras will let down their guard at the feeding station and be easily corralled.
No-fuss, no harm, no foul.
I’m sure you are thinking that the animal services officials ought to stridently hunt for the zebras and use tranquilizers to subdue the wild beasts. Per news reports, the authorities say that the zebras are too fast and it is overly dangerous to try and shoot them with the tranquilizers. A preference is to do things the friendlier way, thus gently lure them to the feeding station and corner them there.
Okay, given that latest news story about zebras on the loose in Maryland, I want to ask you a question.
If you hear hoofbeats, should you assume they are horses or that they are zebras?
Aha, we return to the witty saying in the medical field.
That being said, we have to be upfront and acknowledge that the odds are still higher that you’ll be hearing hoofbeats of horses rather than zebras. The number of zebras in this Great Escape is quite small, less than a half-dozen, and they are pretty much remaining in a confined rural area. So, unless you perchance live or work in that specific locale, you are unlikely to be witnessing footloose zebras anyplace else.
We’ll have to remain contented with the famous saying, though relishing the lighthearted irony in this instance is certainly a dollop of grazing fun.
A bit of seriousness does enter into the situation when you consider what could happen if those zebras run around on the highways and byways of Maryland. The zebras seem to have opted to stay in wooded areas and luckily keep off the motorways.
There are though alarming photos of the zebras walking, striding, and galloping on busy avenues and car-using streets.
Imagine that you are driving your car and perchance see up ahead a zebra.
If you didn’t know about the zebra breakout, you would swear your eyes were deceiving you. The odds are that you would at first assume the zebras were horses that just so happened to resemble zebras. Or you might guess that the circus has come to town, and they are doing one of those elephants walks whereby they parade a bunch of extraordinary animals down the local streets.
I really doubt that you would instantaneously believe they were zebras.
Shifting gears, what would an AI-based true self-driving car calculate the wiry beasts to be?
Note that 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: What can these zebras on the loose inform us about 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 Striped Zebras Afoot
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.
Self-driving cars are outfitted with a variety of sensors. The sensor suite is used to examine the driving scene. Sensors can include video cameras, radar, LIDAR, ultrasonic units, thermal imagining devices, and the like. The choice of which sensors to utilize is something that varies from automaker to automaker and depends upon the preferences of the self-driving tech firm involved.
The sensors are used to collect data about the driving environment. Within the AI driving system, there are various algorithms and techniques used to detect objects that exist in the driving scene. The use of Machine Learning (ML) and Deep Learning (DL) are frequently being used for image processing and other object detection capacities.
Envision that a true self-driving car is driving along and up ahead there are some zebras.
Presumably, the sensors are feeding gulps of data into the AI driving system. The object detection facilities are mathematically analyzing the data to figure out what objects are being detected. For example, there might be other cars nearby, all of which are objects that hopefully the AI can discern. Likewise, pedestrians standing on the sidewalk are objects that need to be detected.
The first aspect entails discerning the objects are in the driving scene. It would be like you taking a picture with your smartphone camera and then circling all of the objects that you can spy. I’m sure you’ve played the popular child’s game of I Spy, wherein you glance outside of the car and try to spot or detect various objects.
Once the objects have been generally delineated, the next step involves classifying the objects. Those blobs over on the sidewalk that seems to have two legs and two arms are likely pedestrians. Those large contraptions on four wheels seem to be cars. And so on.
It can be harder than you might assume to try and categorize detected objects.
Humans generally do a nifty job of categorizing or classifying objects. You see an object and almost immediately are able to classify what it is. Rarely are you be up in the air about the thing you’ve seen. A light pole is a light pole. A fire hydrant is a fire hydrant. Anybody used to being in an everyday setting is likely to know what the objects represent.
You can get a semblance of how hard classifying can be if you chat with a toddler about what they see around themselves. A young child will be able to tell you what some objects are, and be at a loss for words about other objects. This is the proverbial “what is that?” question that kids will bombard adults with. They are curious to know what unknown objects signify.
Part of the reason to classify objects is so that you’ll be aware of what those objects portend.
Once you’ve ascertained that a light pole is a light pole, you right away can rest easy about the object. I say that because you would normally anticipate that a light pole will remain steady and upright. It isn’t going to suddenly and shockingly run out into the street. By classifying an object, you associate a bunch of potential actions and events that can arise related to that object.
Take for example pedestrians. The moment that you realize a pedestrian is a pedestrian, you can gauge a lot of nuances about what might occur in the driving scene. A pedestrian can jaywalk, doing so at any time and for any reason. You need therefore to keep your eye on pedestrians. If they are seemingly staying on the sidewalk, all is okay. Meanwhile, a pedestrian that turns toward the street or darts along is a potential danger that might require an evasive driving maneuver in the quick.
Sometimes you aren’t completely sure about the classification that you assign to an object.
Suppose there is a crowd of people at the corner of the street. They are bunched up. Can you be absolutely sure how many people are in there? In fact, someone that has a stroller is obscuring another child standing behind the parent and the stroller. You cannot be fully sure about what you see in this circumstance.
In the case of AI driving systems, the objects are usually assigned a probability associated with whether the object exists or not, along with a probability of what the object represents. We kind of assume that people do the same type of thing.
Let there be no doubt that uncertainty plays a notable role in our world.
Humans need to deal with uncertainty and this rears its ugly head when at the wheel of a car. A car is a multi-ton vehicle with a great deal of power. Misjudgments by human drivers can readily produce car crashes. In the United States alone, there are an estimated 6.7 million annual car crashes, of which produce approximately 2.3 million injuries and sadly 40,000 fatalities (see my coverage of roadway and driving stats at this link here).
I’ve dragged you through this discussion about objects so that you’ll be ready to grasp the delicate matter of dealing with footloose zebras.
Earlier, I mentioned that a human driver would be surprised to see zebras whilst driving along a roadway. Unless you lived in a country or locale that had an abundance of zebras and that were commonly meandering around, you would be caught entirely off-guard to suddenly spy some zebras up ahead of you.
The odds are that most of today’s AI driving systems would likewise be caught somewhat by surprise, though I don’t want to suggest or imply that the AI is “surprised” in an anthropomorphic manner. The surprise is really based on the lack of programming by the developers of the AI driving system being utilized.
Allow me to explain.
Things are quite hectic for those devising AI driving systems and self-driving cars. The primary goal is to make self-driving cars that can drive in our everyday neighborhoods and communities. As such, the AI is shaped around detecting the type of objects you would expect to see on city streets.
Anything else is likely rated as an outlier, oftentimes referred to as an edge or corner case. The notion is that the first place to get the AI figured out involves the core aspects of driving. The oddball or rarer situations can be dealt with later on. Some fervently believe that we are never going to turn the corner on true self-driving cars because there are always more edge cases to be dealt with, as per the so-called long-tail problem. See my analysis at this link here.
Bottom-line is that zebras are not likely a fully recognizable or classifiable object by most of today’s AI driving systems.
Let me clarify that assertion.
A zebra is certainly detectable as an object, and likely so as an object that has four legs and moves around. But that object is not matchable to its aligned virtual representation if there isn’t a defined aspect of zebras inside the AI driving system and its object recognition software.
In essence, act as though you had never seen a zebra in your life. You hadn’t seen any pictures of zebras. You hadn’t seen them in movies or on TV. Nor did you see any zebras when you visited your local zoo. A zebra is a completely unknown creature to you.
That’s what happens when the AI developers decide that getting the AI to recognize zebras is not something worthy of attention right now. The same could be said about recognizing elephants. These are animals that a U.S. traversing AI self-driving car is highly unlikely to encounter and ergo the AI object detection is not trained to gauge.
This though doesn’t mean that the AI will ignore the detected object.
There is an object up ahead, and it has four legs. It appears to be moving. The shape of the object and its motions are akin to those of a horse. Most of the AI driving systems are programmed to recognize horses, which makes sense because the chances of at some point encountering horses while driving a car are relatively assured.
I’m not just talking about when driving in remote farmlands where you are bound to come across a horse. There are horses even in the bustling heart of New York City. Some of the horses are used to take tourists around for a festive ride in the major parks via horse-drawn carriages. There are also horses being used by police that patrol various areas. Etc.
It makes inarguable sense to be able to categorize objects that are discernable as horses.
This takes us back to the medical profession and the catchphrase about horses and zebras.
When it comes to the crafty saying that if you hear hoofbeats then assume they are horses rather than zebras, most of the existing AI driving systems have no other option but to assume that the object is a horse. In that manner of consideration, there is pretty much zero chance of categorizing the object as a zebra.
This will continue to prevail until the developers of AI driving systems decide it is time to incorporate a familiarly about zebras into their wares.
Here’s something that you might have been pondering.
Would it make a difference that an AI self-driving car misconstrued a zebra as being a horse?
In the Maryland scenario, the odds are that this faulty categorization would not make any notable difference. The AI driving system probably is programmed to anticipate that a horse could suddenly move this way or that way. As such, the AI is attempting to predict what this kind of object might do next. Whether this is classified as a horse or a zebra is not of substantive consequence, as long as the horse classification comes to bear.
We could reasonably guess that the actions of a horse and the actions of a zebra are approximately the same when it comes to considering how they will react on roadways. The two types of animals might have other fundamental differences such as dietary or other facets, but that isn’t especially significant to the act of driving when near them.
Let’s wrap up this discussion with some handy lessons learned.
Being at the steering wheel and upon hearing hoofbeats, you can allow yourself some latitude and make a reasoned guess that those are the sounds of a horse.
When you get home and your family asks you if you saw those escaped zebras, you’ll undoubtedly be a bit ashamed and chagrined to report that you thought they were horses. You missed your big chance, perhaps once in a lifetime, to drive nearby escaped zebras (please do so cautiously!).
For peace of mind, I’d bet that by the time you read this article, the Maryland zebras will have been adroitly returned to their farm. They will have had an adventure they can delightfully tell their offspring about.
For medical doctors that are working in Maryland, they are probably having a good laugh about those footloose zebras. As a medical professional, after all those years of medical school and being knocked over the head about the zebra aspects, they now had a modicum of insider mirth about it all.
Well, if I somehow end up visiting Maryland and gosh-forbid have something adverse occur such that I need to be hospitalized, I still want them to be thinking about horses, though I’m fine with them also considering zebras to make sure that whatever ails me is getting covered on all the bases. Calling Dr. House, stat.
We’ve covered this topic extensively and I’d loath to beat a dead horse on it. But I was wondering if those zebras are happy that we are potentially mistakenly classifying them as horses. This might greatly irk them. How in the heck can we not notice those awe-inspiring stripes?
Hey, humans and AI, wake up and get your act together, they collectively exhort. Zebras have a right to be recognized.