Let’s talk about common sense. We all have it, presumably. Some more so than others. When we hear about someone performing a seemingly nonsensical act, there is a strong impulse to immediately decry that the person lacked common sense. You either have it or you don’t.
Well, more likely, you sometimes have it and sometimes you do not. There are occasions where we either skirt our reservoir of common sense or it might be altogether blank in a particular setting or context.
A popular way of characterizing common sense is to suggest that there are two primary components involved. First, there is the knowledge itself that underlies the basis for common sense. Those are the bits and pieces of information stored in your memory banks and utilized for tapping into a kind of mental database about the nuances and substance of common sense.
Second, there is the use of common-sense reasoning that comes to play. This entails the enactment of the common sense that is retrievable from within your mind. You reason about a particular matter and are doing so by applying the common-sense cornerstone that you’ve prior established.
We’ll get to this in a moment with some handy examples to illuminate the matter. A person can have a large store of common sense and yet fail to reason with it. This can happen if you are suddenly startled and do not have the presence of mind to mentally process what is happening. The reasoning process does not get engaged or is tardy to the party and ergo unable to timely leverage the available common-sense mental embodiments.
On the other side of that coin, a person might have a notable semblance of common-sense reasoning but lack a sufficient common-sense foundation to do much with it. Suppose that you are a city slicker and not familiar with being out in the woods. Though you might be generally adroit at common-sense reasoning, upon wandering into a wooded forest you realize that you know so little about nature and forests that your reasoning prowess is going to come up short. Until you’ve gleaned more about what kinds of facts and considerations underlie a forested realm, your reasoning ability can only provide a limited advantage.
How do we come upon our building up of common sense?
Children are considered to have only a modicum of common sense and the assumption is that by the time they reach adult age, they will have a more seasoned repertoire of common sense under their belt. In that sense of things, the acquisition of common sense is usually characterized as amorphously acquired over time. You cannot particularly take a class on common sense and somehow gleefully graduate with a heady adoption of common sense now firmly implanted in your noggin.
Trying to pin down the exact facets of common sense is squishy and challenging to finitely define. Some would say that it is akin to the famous line about art, namely that you know it when you see it. The thing is, those without ardent common sense are likely to find themselves in a world of hurt, potentially taking actions that are seriously injurious or harmful, unknowingly so.
Besides common sense occurring in humans and the human mind, the field of AI is eager to crack the code on what common sense consists of.
Those in the AI field are well-aware of the slipperiness of getting a handle on how to codify common sense. Many have tried and are continuing doggedly to find ways to embody this elusive form of cognition into computer-based systems. By and large, the attempts to codify common sense and plop it into AI capabilities have been fraught with great difficulty and exasperating disappointment.
Some initially believed that a gigantic facts database might be sufficient for giving rise to AI-based common-sense implementation. The sky is blue. Clouds can float across the sky. Objects seen in the side mirror are closer than they appear. On and on, the hope was to try and articulate every kind of fact or piece of knowledge that mankind has accumulated.
That’s a tall order.
Plus, even if you could manage to populate a system with tons and tons of those minuscule and seemingly randomly arrayed facts, the question arises as to how they would be made into the goldenrod of common sense (note that some efforts have opted to categorize and try to structure the factual database). This then led to efforts of proposing a myriad of ways to have machine-based common-sense reasoning algorithms that could exploit the (presumed) common-sense embodied knowledge bases.
I dare say, AI is still not even in the ballpark of approaching the human capacity of common-sense and common-sense reasoning. Not anywhere seemingly close to it.
The road ahead is long and bumpy.
Speaking of roads, we can use the everyday act of driving a car as a helpful means of exploring the realm of common sense and common-sense reasoning.
Do you use common sense when you are driving a car?
Absolutely.
You might not be directly aware that you are leveraging your common sense aptitudes, but you are abundantly doing so. I mention this key aspect because some would say that driving is a specific skill that needs to be acquired and in a sense denotes a kind of expertise. To drive a car, you need to learn about how to use the steering wheel, how to use the brake pedal, how to use the accelerator, etc. You need to learn the driving laws that stipulate how you are allowed to drive.
Yes, there is no question that driving does involve quite an amount of specifics in terms of particular knowledge and reasoning needed for the chore of driving. No doubt about that. Indeed, there are driving schools and we test people before they are granted a license to drive.
Meanwhile, silently and generally unnoticed, your common sense is residing resolutely at the core of your driving efforts. You could argue that this is partially why you cannot get a driver’s license until around the age of 16. The logic is that you need the cognitive maturity for driving, and we also are in an unstated way implying that you would have hopefully gained sufficient common sense by that age to warrant the privilege of driving.
Keep in mind that driving is a life-or-death act.
I don’t want to seem sour or downbeat, but the reality is that each time you get behind the steering wheel, you are ascertaining the life or death possibilities for yourself, for your passengers, for other nearby drivers, and their passengers, for pedestrians, and the like. We seem to take for granted the driving of a car and do not explicitly realize the somber and quite risky aspects that we incur daily. Just to mention, there are sadly about 6.7 million car crashes in the US each year, producing approximately 2.5 million injuries and over 40,000 fatalities (see my analysis and coverage about such stats, at the link here).
A driver needs to have the proper skills associated with driving to appropriately drive a car. In addition, a driver needs to have sufficient common sense to be able to adequately drive a car. I think you would be hard-pressed to argue otherwise.
Consider the following scenario.
You slide into the driver’s seat of your car and decide to go for a cheery Sunday drive.
Using your honed driving skills, you start the engine, you back out of your garage, you continue in reverse and drive down your driveway, finally edging slightly into the street. Looking both ways, you decide it is safe to fully enter the street. After backing a little way further and slowly turning the wheels of the car, you are ready to put the car into forward drive. You then nonchalantly drive down the block and begin your joyous driving trek.
Was there any use of common sense in that seemingly perfunctory effort of getting underway on a driving road trip?
Zillions upon zillions of common-sense instances arose, though few would realize the importance and infused nature of common sense in this situation.
Let’s unpack it.
You knew what a car is and what a car can do, which otherwise you wouldn’t have gotten into the car, to begin with. You knew that you would have to go in reverse and back out of the garage since going forward would be disastrous and you would summarily ram into the wall of the garage. You knew that there could be obstructions or possibly kids or dogs nearby to your driveway and thus you scanned around to make sure that you weren’t going to hit something or someone. Likewise, when you reached the street, you looked for cars or bikes or anything that might strike your vehicle.
You know that objects such as bikes can be in motion. You know that things in motion can come toward you. You know that when one object encounters another object, this gives rise to a collision. You know that collisions can cause damage. You know that the type of damage can include the injury. You know that injuries can potentially lead to deaths.
I trust that it is readily apparent that you are using a lot of common sense about the world and how things operate and interact in the real world, doing so during the routine effort of getting underway on your driving journey. Multiply this by all the other common-sense aspects that will come to play as you drive down the street, and as you drive on open highways, and so on.
Another means of revealing the cognitive subtleties of common sense can be shown via our use of language.
Here is a sentence that I’d like you to carefully read: The moving car hit the Stop sign because it was in the way.
What did you glean from that sentence?
I’m assuming that you figured out that a car was in motion and it came upon a Stop sign, perhaps sitting in the street and somewhat in the path of oncoming cars. The car struck the Stop sign. We don’t know how hard the blow was. We seem to be told that the driver was not necessarily at fault. The Stop sign was seemingly not where it should have been in terms of normally being outside the realm of where cars might be going.
Here is the same sentence with a notable change: The moving car hit the Stop sign because it was speeding.
Is the meaning different?
Yes, gobs yes. We are logically going to infer that the car was speeding and this presumably reckless act led to the driver ramming the Stop sign.
I want you to reinspect those two sentences and keep your eye on the word “it” within the sentences.
In the first sentence, we have “because it was in the way” and you are left to your own devices to semantically grasp what the “it” is. Does the “it” refer to the car or the Stop sign? We don’t know for sure, but you do know by common sense that the Stop sign is more likely to be the object that was in the way, rather than the car.
In the second sentence, we have “because it was speeding” and you are left to your own devices to semantically grasp what the “it” is. Once again, I ask you, does the “it” refer to the car or the Stop sign? In this case, our common sense suggests that since the “it” is speeding, the car is most likely the object being referred to rather than the Stop sign.
That exercise of cognitive analysis is insightful.
We have discussed the role of common sense in an everyday real-world setting of backing up a car and going for a drive. This involves numerous dimensions of common sense. A vital dimension or frame of reference entails the physics of the world. Objects and their manifestations underlie our common sense regarding physical interactions.
The language exercise proffered another variant of how we use common sense. I gave you some sentences and you mulled them over in your mind. There wasn’t something physically that happened akin to the aforementioned effort of driving a car out of a garage and into the street. This was the use of common sense in a pure thought experiment fashion.
You are perhaps convinced by now that common sense is a core component of human reasoning. And, you are hopefully also persuaded that common sense is hard to delineate in any codified way. It is somewhat nebulous and everywhere, largely hazy and mushy to get your arms around.
We have also used driving as an exemplar of the importance of having common sense. Furthermore, the field of AI is seeking to decode common sense and include this sorely needed capacity into AI systems.
Whoa, this brings us to a monumental juncture or conjunction.
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: Do we need common sense reasoning to achieve AI-based true self-driving cars, and if so how will we get there?
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 The Need For Common Sense
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.
Some assert that the AI of today will not become bona fide AI until we can ascertain how humans embody common sense. There is even a naming that some use, referring to today’s AI as so-called narrow AI. This is meant to convey the notion that AI is narrowly focused and not able to work in any broadly based manner.
One strong case being made is that without AI systems having a common-sense component, they will forever be limited in what they can do. They will be narrow. They will be fragile and brittle. They won’t at all approach the vaunted goal of attaining artificial intelligence in the full sense of what we aspire to achieve.
Let’s consider the advent of AI-based true self-driving cars in that bailiwick.
Level 4 self-driving cars are considered to be bounded by a pre-defined ODD (Operational Design Domain). This consists of an automaker or self-driving tech firm specifying the context in which their AI driving system can properly operate. For example, an AI driving system might be established to work in a particular geographical area such as a designated part of a city, and do so only when the weather is moderate (won’t operate in rain or snow).
Some believe that we might be able to muddle along and get self-driving cars to generally operate in various strictly prescribed ODDs. One implied basis for keeping the ODD’s strictly bounded has to do with the lack of common-sense reasoning capabilities of AI.
In short, we aren’t likely going to attain Level 5, which essentially ditches the ODDs concept and be able to have AI that can drive in a wide variety of open-ended settings (within limits, see my discussion at this link here), unless we conquer the embodiment of common sense and common sense reasoning into AI systems.
But as earlier stated, progress has been stunted and many are wondering what other outside-of-the-box ways could we pursue to overcome the existing hurdles. That being said, if trying to ferret out the deep secrets of human common sense is arduous and having troubles progressing, there is another avenue that might be considered.
Sit down and prepare yourself for a bit of a surprise.
Animal common sense.
Say what?
The argument is that animals also have common sense.
Of course, you can quibble with this and contend that the common sense exhibited by animals is not nearly as sophisticated and advanced as that of humans. Maybe so, but the notion is that we might start there and see if that gives us a leg-up toward figuring out the fuller blown human-based common sense.
A clever research article entitled “Artificial Intelligence and the Common Sense of Animals” published in Trends In Cognitive Sciences makes an intriguing case for making such a pursuit.
Authored by researchers at DeepMind, and the Imperial College of London, plus the University of Cambridge (Murray Shanahan, Matthew Crosby, Benjamin Beyret, and Lucy Cheke), they provide this salient point: “Here, we argue that common sense in humans is founded on a set of basic capacities that are possessed by many other animals, capacities pertaining to the understanding of objects, space, and causality. The field of animal cognition has developed numerous experimental protocols for studying these capacities and, thanks to progress in deep reinforcement learning (RL), it is now possible to apply these methods directly to evaluate RL agents in 3D environments.”
Is this farfetched?
Not at all.
It makes solid sense. I’ve covered previously that there have been efforts to see if dogs could drive a car, see the link here. There are studies of mice brains that are hoping to reveal how the thinking process works, possibly revealing insights that will aid in explorations of the human brain. Etc.
You can readily make the case that animals do have a semblance of common sense. Perhaps by figuring out their form of common sense, we can generalize and make added progress on human-based common sense.
It is the proverbial take a step back and rethink our approach solution, namely that we can start small or smaller and ultimately make our way to the bigger machination.
Conclusion
There is a famous old fable about a donkey that might be relevant to consider.
It seems that a donkey had fallen into a well. The farmer hears the crying sounds of the trapped donkey. Upon looking closely at the situation, the farmer decides that there is nothing much that can be done since the donkey is too heavy to be lifted out of the well. With great sadness in his heart, he decides that he will at least cover the donkey with dirt and provide a suitable burial.
As each scoop of dirt is summarily dumped into the well, the donkey at first is surprised but then starts shaking off the dirt. This continues repeatedly. Eventually, to the surprise of the farmer, the donkey is near to the top of the well and easily leaps out, continuing on its merry way.
Turns out that the donkey had iteratively shaken the dirt to the floor of the well, pounded on the dirt to compact it, and gradually the floor level rose and rose, reaching a point that the donkey could readily jump out.
The moral of the story that we’ll consider here is the donkey employed better common sense than did the farmer. This illustrates that not all common sense is necessarily bright or inspiring. As they say, sometimes common sense isn’t as common as you might think it is.
Plus, this is a handy parable since it highlights what we hope to glean from studying animal cognition and the role of common sense.
For those in AI development or AI research that find themselves stuck down inside a well and seemingly at wit’s end about how to figure out common sense, the lowly and stubborn donkey might provide insightful clues about discovering the heart and soul of common sense and inexorably lead to instilling that into more advanced AI.
Per the classic proverb, the patience of donkeys is a laudable virtue.
originally posted on forbes.com by Lance Eliot
About Author: Dr. Lance B. Eliot is a Stanford Fellow and a world-renowned expert on Artificial Intelligence (AI) with over 4.5+ million amassed views of his AI columns. As a seasoned executive and high-tech entrepreneur, he combines practical industry experience with deep academic research and serves as a Stanford Fellow at Stanford University.
Formerly a professor at USC and UCLA, and head of a pioneering AI Lab, he frequently speaks at major AI industry events. Author of over 50 books, 750 articles, and 400 podcasts, he has made appearances on media outlets such as CNN and co-hosted the popular radio show Technotrends. He’s been an adviser to Congress and other legislative bodies and has received numerous awards/honors. He serves on several boards, has worked as a Venture Capitalist, an angel investor, and a mentor to founder entrepreneurs and startups.
