By Lance Eliot, the AI Trends Insider
Special Note to My Readers: At AI World last week in Boston, I was honored to be the opening evening keynote speaker. Throughout the several days of the conference, I was frequently approached by readers of my AI Trends Insider column that came up to chat with me about it and offered various heartwarming feedback about the value they found in my coverage of the latest trends in AI. Thanks goes to those that chatted with me and indeed to all of my readers for their avid feedback and ongoing interest in the column. I vow to continue being topical, covering AI innovations in a practical and interesting way. Wishing you all the best in your AI endeavors. Dr. Lance Eliot.
Consider for a moment your nose. Yes, your nose. The human nose has an amazing capability to detect odors and remains as one of the “last frontier” areas for trying to duplicate this same functionality into an electronic sensor of some kind. If you have an idea of how to make a fully capable e-nose, an electronic nose, you could be sitting on a goldmine.
Light sensors such as cameras are pretty well matured and can readily take the energy wavelengths of light and capture it for us. Sound sensors such as microphones are pretty well matured and can readily take the energy of sound and capture it for us. Odors are trickier because the nose functions by detecting molecules representing odors, which becomes a form of mass measurement rather than energy measurement. Attempts to develop e-nose devices have generally met with difficulty and perfecting a bioelectronic nose or machine olfaction is still an open avenue awaiting a breakthrough.
The interaction of the physiological aspects of the nose and the psychological aspects of the brain are a wonderment.
As an example, the other day I was walking through a farmer’s market in downtown Los Angeles and was joyfully taking a look at the dozens upon dozens of food items that were on display and being cooked, trying to decide what I might eat for lunch that day. Wafting through the air of the marketplace was a cacophony of odors. I could smell slowly roasting meats, I could detect the distinct odor of veggies being cooked, and so on.
All of a sudden, I got a whiff of an odor that caught me by surprise. It was a scent that I had not smelled in many years. My parents used to make a certain kind of unusual soup for special occasions when we had guests come to our home, and I had not ever seen or smelled the scent of that soup since I was a very young child. With just the tiniest hint of the smell in the farmer’s market, my mind raced back to the days of my childhood and I remembered vividly the times when the cooking of the soup took all day long and occurred in anticipation of out-of-town visitors coming to our home for dinner.
It was incredible that with the miniscule whiff of this odor that I suddenly was transported back in time to my childhood and inescapably a somewhat obscure memory took over my mind momentarily.
That’s the grand power of our noses. They can smell pleasant odors such as perfumes, colognes, and other pleasing fragrances. As humans, we obviously seem to give importance to such odors because the estimated global market is about $75 billion being spent annually for fragrances (according to Statista). That’s a lot of money for merely giving our noses something sweet to smell. There must be something significant about our noses to justify that kind of spending.
The other side of the smelling activity involves those rotten and horrid odors that repulse us.
The other night in my neighborhood there was a skunk smell that seemed to last for nearly an hour. When the odor first became apparent, I was outdoors and decided I’d try to avoid the smell by going inside my house. The windows of my house were open and shortly the entire house had the stink of the skunk in it. I even thought that the skunk must be right under my floorboards or be purposely standing at my window and emanating the stench from there. I was torn as to whether to now close the windows, perhaps trapping the odor inside, or even open them wider in hopes that the prevailing winds might blow through the house and get the odor out. What a pickle!
The skunk uses odors as a defensive mechanism. This highlights again the power of smells. We all might think it obvious that an animal might use claws or teeth as a defensive tool, but the idea of simply using an odor seems less useful to our overall sensibility about ways to protect yourself. Emit an odor and somehow prevent other creatures from trying to devour you? Doesn’t at first seem like a credible idea.
Keep in mind that if you get too close to a skunk odor, you can get nauseous and your eyes can burn from the odor. Thus, their odorous defense is more than simply causing you to be discomforted by the unpleasant smell — most animals seem to realize that messing around with a skunk is going to be a bad move for them. Besides skunks, various other animals also use odors to their advantage, including bombardier beetles that use odors for defensive purposes and opossums are also able to use odors for repelling predators (it’s via their excrement).
I had mentioned that while in the farmer’s market I had gotten just a faint whiff of the soup odor that launched me into going down memory lane. Our noses can often pick-up very faint odors that are at a distance or that are very diluted in the air. The skunk odor near my house was at first rather slim, but eventually it became nearly overpowering and I was almost choking on the stink.
We can categorize odors by their intensity, such as this scale is often used:
0 = no odor detected
1 = very weak odor and at threshold of detection
2 = weak odor that is considered detected
3 = distinct odor that is considered fully detected
4 = strong odor
5 = very strong odor
6 = intolerable odor that is overpowering (some say “suffocating”)
It’s not just the intensity that governs the nature of our nose and its ability to smell odors, but also the type of odor that counts too.
There is some dispute about whether or not there are “primary” odors, akin to the notion of having primary colors. Smells are not quite as readily categorized as what we can see via light. Some also assert that smells are much more subjective and that the “eye of the beholder” comes to play in the manner of the “nose of the besmeller” (Ok, that’s a Shakespearean word), determines whether an odor is pleasant or not.
When my children were young, I used to change their diapers as babies and the smell was quite rough to withstand. A friend of mine that also had children, he loved the smell of his kids filled-up diapers. He used to brag about the odors and even wished that he could capture the smells and preserve them, similar to taking pictures and putting those into a photo album. I loved my kids, but the packed fully “perfumed” diaper was not the top of my list of awesome odors.
Anyway, some researchers list these below seven aspects as a proposed list of primary smells, take a look and see if you agree that these are the core odors of the world:
- Putrid (such as the odor or rotten eggs)
- Pungent (vinegar would be an example)
- Camphoraceous (mothball-like odors)
- Ethereal (an example would-be dry-cleaning fluid)
- Floral (a rose by any other name)
A catchy mnemonic for considering the nature of odors is FIDOL. The F stands for the frequency of the odor. The I is the intensity of the odor (which I’ve provided an intensity rating system herein). The letter D is for the duration or length of time that the odor exists for you to smell it. The use of the letter O refers to the offensiveness of the odor. And the letter L is used to indicate that the Location of the odor can be significant.
In studies of fruit flies, we are able to learn quite a bit about how odors seem to work and reveal aspects about how the “nose” functions.
An interesting study done at the University of California San Diego and in conjunction with the Salk Institute for Biological Studies revealed that a fruit fly seems to tag each odor that it detects. Tags are then placed into one of two overall buckets, an attractiveness group or a repulsive group. The tags are somewhat sparse and non-overlapping.
The appearance of an odor gets the nostril receptors to become chemically stimulated. This stimulus then generates electrical kinds of biological signals that are then transmitted to the brain. There are Odor Receptor Neurons (ORN) used for this purpose.
The fruit fly study was especially intriguing because it seems that the fruit fly first does the odor detection with about 50 neurons, and then it fans them out to a wider set of about 2,000 neurons. This would seem counter-intuitive because you would usually expect that an odor would fan down, being narrowed so as to figure out the nature of the particular odor.
The researchers likened this fanning expansion as the equivalent to having a bunch of people in a crowded room, whereby you could not readily make heads nor tails of each person because they are crammed together, and instead you repositioned them onto a football field. By repositioning them into a large open space, you would have an easier time of being able to categorize them and figure out what you seem to have in-hand. That’s the working theory on it.
What also made this intriguing was that the researchers suggest that the fruit fly olfactory circuitry uses a kind of locality-sensitive hashing (LSH) function, which is a known form of a computer science algorithm for search-space purposes. Presumably, the fruit fly assigns tags to an odor and for which the more similar the two tags are the greater chances of realizing that those are odors of a similar nature (thus, being “locality-sensitive”). This is akin to being able to numerically find the approximate and nearest neighbors in a highly dimensional search space using a hashing function. Nice to see that perhaps the programming we do with computers can be seen as something that nature itself has honed and uses too (yes, I realize that nature’s version came first!).
What about humans and their ability to differentiate various odors?
Another study done on olfactory senses involved using Machine Learning (ML) to try and identify and predict which molecules would lead to which kinds of human-determined smells. Done by researchers at several universities including The Rockefeller University and Arizona State University, they were trying to see whether it might be possible to predict a “smell” by the molecules that presumably should invoke the human-determined and described odor. This is crucial since right now there is little in the way of well characterizing the dimensionality and size of the olfactory perception space, and furthermore the range and nature of the invoking odors is itself a hazy kind stimulus space.
I liked this particular study since it suggests that we can construct models to be able to take a “subjective” smell and reverse engineer it back to the likely molecules that caused the odor. Plus, we can potentially inspect a molecule and be able to somewhat accurately predict what kind of smell that people will think it has. The Machine Learning approach was augmented by a regularized linear model that did nearly as good at the predictions as did a more elaborated random forest version.
These kinds of studies are all small steps toward trying to unravel the mysteries of our noses and the act of smelling.
Technology Here Today to Make an E-Nose
Of course, research has been going on for many decades to try and ferret out how we smell something. I don’t want you to think that research is only just now occurring. It is an age-old quest to unlock the secrets of the nose. What makes these more recent studies a rather hopeful effort, and especially worthy of renewed attention is that we have the techniques and technologies today that can allow us to perform research and potentially make a true e-nose, which in years past would have been nearly impossible to do or much harder because the software and hardware were not yet ready for such an effort.
Odors and our noses are part of our primal senses.
Nature has cooked into us this phenomenal thing that we call a nose. Your nose can detect odors that might warn you of impending dangers. Your nose can help you to locate a suitable mate (that’s why there is so much perfume being bought!). In a caveman and cavewoman kind of way, we have used our noses to find food, along with using our noses to detect potential predators.
Anyone that has gone hunting knows that you should try to stay downwind of a bear or deer, since otherwise they might detect your human odor and scoot away. The same can be said for humans, namely that we try to smell whether a predator is nearby, though our ability to detect odors is not as superpowered as many animals (think about this as you walk your dog around the neighborhood and it is constantly sniffing at the air and at objects).
Humans seem to vary in terms of the ability to smell odors. I’m betting you know people that when you ask them whether they can smell a particular odor, they say they cannot, and yet for you it is overpowering. One such example was an office that I used to work in. Every late afternoon, a fellow worker would put popcorn into the microwave and cook it until it burnt. The burnt odor was excessive and permeated the entire office. When I asked my fellow office workers if the odor disturbed them, most said they could barely smell it. When I asked the person that made the popcorn whether they felt any guilt in imposing the burnt odor onto the rest of us, he said that he enjoyed the odor and thought it perked up the atmosphere of the office.
It seems that we all have our own subjective indicators about odors. Some claim that women can smell odors that men cannot or at least are more sensitive to certain odors. Thus, there might be a gender difference in the ability to smell. Some say that the older you get, the less sensitive you are to odors. This might suggest an age-related factor to odor detection. Health is another factor as to whether your nose can aptly identify smells.
There is odor fatigue that comes to play too. If you are exposed to an odor continuously, there is a chance that your nose will gradually begin to stop reacting to the odor and so the odor will appear to fade, even though it might still be heavily present. For the burnt popcorn, I reluctantly tolerated the daily afternoon smell-fest, and kind of got used to it over time. When we had guests come and visit the office, doing so in the late afternoon, invariably they would react quite strongly upon first detecting the smell. Our first exposure to a smell can be strongest and then fade after we get used to an odor.
Another example of getting used to an odor and no longer noticing it includes our own personal body odors. I went camping for a week as part of the Boy Scouts, doing so as a troop leader. We were in the woods and didn’t readily have available a shower or other ways to bathe. After a few days, the odors of each of the troop members was readily detected. You could be standing behind a tree and indicate who was on the other side of the tree, simply by their body odor. The person themselves often didn’t even think they had a detectable body odor and it was hard to convince them that they stunk like a skunk.
When you ask someone to try and pinpoint an odor, it can be challenging at times. Assuming that they indeed smell the odor, there is a kind of search in their minds about the nature of the odor. Most people can usually tell you rather quickly whether they have ever smelled the odor before. We seem to have a cognitive ability to rapidly do a search of our mental space to see whether an odor has ever been recorded in our minds.
If you then ask the person what the odor means or signifies, they often need to do more mental searching. It might be difficult for the person to remember when or where they last had smelled the odor. It is as though cognitively the “records” associated with the odor are poorly linked together. Perhaps over time the neural connections to the stored odor have faded or changed or been used for some other purposes. Trying to make a reconnection can require some deep concentration, perhaps a means of the mind trying to do a search and then reinvigorate prior connections or create new ones based on the reconstruction of past events.
What’s even less likely consists of having someone tell you what the odor consists of. For a skunk, it is pretty much just the odor of a skunk. If you’ve got a Starbucks fancy coffee, you might be able to pick out the odors of the coffee and the various added ingredients, assuming you’ve trained yourself to do so or otherwise paid attention to the concoction and memorized what the various components smelled like. You’ve perhaps seen wine tasters that claim they can tell you not only the elements of the wine based on smelling it, they often impress by telling you the year of the wine, where it was made, and try to do a James Bond kind of thing by saying they can even tell you the name of the person that stomped on the grapes.
People though that swear they can smell numerous distinct components of an overall odor can at times be falsely thinking that they do. I’m not necessarily saying that the person is lying. It could be that the person genuinely believes they detect various odors, but it is a kind of trick of the mind that they believe this. You can tell a person that they are holding a coffee with certain ingredients, and they affirm the ingredients by sniffing the coffee, but meanwhile you’ve handed them the wrong cup of coffee and it doesn’t have those items in it at all. Their mind had led them to believe that the elements were there, and so it might have imputed the odors even when they were not there at all.
One odor can at times overpower another odor, causing you to lose your ability to detect the now hidden odor. If I put one of those filled-up diapers next to a Starbucks coffee, I assure you that you would not likely be waxing on and on about how great the coffee smelled. In any complex mixture of odors, the ability to pick out specific odors can be quite hard to do. You’ve likely seen people sniff something, tell you an odor, then sniff the item again, and tell you another odor. At times, we can perhaps mentally isolate one odor from a mixture, if we put our minds to it.
Speaking of the mind, once the Odor Receptor Neurons report an odor to the brain, it’s up to the mind to figure out what the odor is and what significance it might have. Since this is a mental matter, you can presumably alter your thinking about an odor and become more attracted to it or more repulsed by it, over time. I had a dog that at first was smelly and I didn’t want to use perfume-like soap to mask the odor. Gradually, I got used to the natural odor and even liked the smell. Even today, if I smell a dog that has a similar such odor, I get a big smile as it reminds me of that favored dog that I once had.
Odors can have a tremendous impact on our emotions and ultimately our behavior. I say this because sometimes when I talk about an olfactory e-nose sensor, there are people that shrug off the sense of smell as a kind of irrelevant topic. If you ask these people about their various senses, they would say that their eyes and ears are crucial, but their nose is not. Were their nose to get plugged up and be unusable, they would say it doesn’t particular matter to them. No big deal.
Aromatherapy claims that you can cure various psychological and physiological ills of the body by paying attention to odors and the proper use of odors. It is at times startling to be reminded out-of-the-blue about the power of the nose and the nature of odors. My farmer’s market experience was an example of how the body and mind work together at times. The moment that I sniffed the odor in that farmer’s market and separated it from the many other odors wafting through the thick air, my mind leapt back to my childhood, resurrecting memories that I doubt I could have otherwise readily recalled, even if you asked me to do so. An odor and the functions of the nose and the mind can be rather incredible.
I would say that sometimes odors can be lifesaving.
During one of the Scouting trips, I was driving my car through a vast wooded wilderness, trying to connect up with my troop. I had been delayed at work getting to the campsite and so was arriving long after everyone else had arrived and setup camp. The roads were all dirt roads and there was no particular signage and certainly no GPS available in this remote area. It was getting toward sunset and the darkness would make it even harder to find where the troop was camped out.
I had the windows rolled down, in case I might be able to hear the sounds of the troop. Often, you can hear the troop leaders barking out orders or hear the sound of the Scouts exploring the woods. My eyes were peeled, looking for any kind of indication that maybe the troop was around the next bend. I begun to smell an odor, the smell of smoke, a campfire.
I realized that the troop would have setup a campfire and it would likely be the only one anywhere near to this area of the empty wilderness. I used my nose to try and catch the odor. I even put my head outside of the driver’s side window, and sure enough was able to figure out which road to take to get to the campsite. The nose came to the rescue!
I realize you might say that this example was not a lifesaver per se, since I would have been unlikely to somehow expire if I did not find the campsite. Yes, I could have survived in my car for the night and then gone to look for them the next day (actually, I would likely have setup my own miniature camp, using my brought along camping gear!).
Here’s some examples that might be more relevant to the notion of being lifesaving.
Nose May Have Been a Lifesaver on This Camping Trip
I was with a bunch of the troop members while driving toward a campout that we were going to have for just a weekend. The weather conditions were quite dry. It was summer time and the temperature was in the high 90’s or low 100’s during the day. Other parts of the local mountains had unfortunately had fires recently, so we were concerned generally about a potential forest fire. Nonetheless, we had scheduled to go camping for that weekend and figured we’d give it a try.
We had driven up after work on a Friday, and so it was nighttime when we arrived at the wilderness. There was no lighting on any of the dirt roads. The only light available was via the headlights of the car. There were several other cars behind me, all part of our caravan headed up to camping for the weekend.
Coming along a winding mountainous road, we began to just barely detect a smoke-like odor. It was very faint. In fact, at first, I didn’t say anything about it. I thought that my awareness of the other forest fires was playing tricks on my mind. Some of the troop members in my car spoke-up and asked if anyone else could smell smoke. I was relieved that they said something, and it wasn’t just in my own mind.
I was going very slowly on this winding road since it was replete with ruts and holes, and we were on a cliff that if you made a wrong move would mean a car would go over the side and down to its doom. As we inched ahead, everyone agreed that the smoky odor was getting stronger and stronger. We didn’t see anything up ahead and were perplexed that we could not see flames shooting up into the air. Nor could we see the smoke, though it was nighttime and so dark that we could barely see our hands in front of our faces.
I decided that where there is smoke, there is fire (well, that’s true a lot of the time), so I opted to turnaround the car. I slowly drove back down and reached the other cars that were still making their way up. Since I was the lead car, there weren’t any other cars ahead of me. Upon meeting up with those other cars, we told them about the smoke odor and how it was progressively getting stronger. We all agreed that it made sense to head back down out of the mountains as a precaution.
The next day, we found out that a fire had gotten started up in that region and it eventually became a widespread forest fire. Had we kept going and opted to campout, we might have gotten stuck in it and the thought of endangering our lives still somewhat haunts me that we might have done so, had we not luckily smelled the smoke when the fire was just starting to burn.
Okay, so that’s hopefully more of a lifesaving example.
Here’s another one, a bit quicker of a story. I was driving and pulled into a gas station. I was going to stop and pump some gas, but I immediately upon getting near the pump, I could smell a heavy odor of gasoline. I decided to drive straight out of the gas station and not stop. As I drove onto the street, I looked back and could see that someone had spilled a bunch of gasoline onto the ground near to the pumps. All it would have taken was one spark and the situation could have been awful.
I don’t want you to think that detecting odors while in your car is only about lifesaving. The other morning, I was driving on a highway and came to a red light. As I sat there, I could distinctly smell the pleasant scent of baked goods. Sure enough, there was a bakery at the corner, one that I had never particularly noticed because it was a small and inconspicuous mom-and-pop bakery. I made a U-turn and came back over to check it out. Turns out it is now one of my favored locations to get freshly baked blueberry muffins and other pastries.
What does this have to do with AI self-driving cars?
At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars. A key element for AI self-driving cars are the various sensors that detect the world surrounding the self-driving car. These include sensors for capturing video and still images, along with radar sensors, LIDAR sensors, ultrasonic sensors, and the like. One futuristic sensor that is intriguing notion would be an olfactory sensor, an e-nose for your AI self-driving car.
For more about sensors for AI self-driving cars, see my article: https://aitrends.com/selfdrivingcars/lidar-secret-sauce-self-driving-cars/
For the IMU sensor, see my article: https://aitrends.com/selfdrivingcars/proprioceptive-inertial-measurement-units-imu-self-driving-cars/
For what can happen when sensors fail, see my article: https://aitrends.com/selfdrivingcars/going-blind-sensors-fail-self-driving-cars/
For the dangers of relying too much on just one type of sensor, see my article: https://aitrends.com/selfdrivingcars/cyclops-approach-ai-self-driving-cars-myopic/
I’d like to clarify and introduce the notion that there are varying levels of AI self-driving cars. The topmost level is considered Level 5. A Level 5 self-driving car is one that is being driven by the AI and there is no human driver involved. For the design of Level 5 self-driving cars, the auto makers are even removing the gas pedal, brake pedal, and steering wheel, since those are contraptions used by human drivers. The Level 5 self-driving car is not being driven by a human and nor is there an expectation that a human driver will be present in the self-driving car. It’s all on the shoulders of the AI to drive the car.
For self-driving cars less than a Level 5, there must be a human driver present in the car. The human driver is currently considered the responsible party for the acts of the car. The AI and the human driver are co-sharing the driving task. In spite of this co-sharing, the human is supposed to remain fully immersed into the driving task and be ready at all times to perform the driving task. I’ve repeatedly warned about the dangers of this co-sharing arrangement and predicted it will produce many untoward results.
For my overall framework about AI self-driving cars, see my article: https://aitrends.com/selfdrivingcars/framework-ai-self-driving-driverless-cars-big-picture/
For the levels of self-driving cars, see my article: https://aitrends.com/selfdrivingcars/richter-scale-levels-self-driving-cars/
For why AI Level 5 self-driving cars are like a moonshot, see my article: https://aitrends.com/selfdrivingcars/self-driving-car-mother-ai-projects-moonshot/
For the dangers of co-sharing the driving task, see my article: https://aitrends.com/selfdrivingcars/human-back-up-drivers-for-ai-self-driving-cars/
Let’s focus herein on the true Level 5 self-driving car. Much of the comments apply to the less than Level 5 self-driving cars too, but the fully autonomous AI self-driving car will receive the most attention in this discussion.
Here’s the usual steps involved in the AI driving task:
- Sensor data collection and interpretation
- Sensor fusion
- Virtual world model updating
- AI action planning
- Car controls command issuance
Another key aspect of AI self-driving cars is that they will be driving on our roadways in the midst of human driven cars too. There are some pundits of AI self-driving cars that continually refer to a utopian world in which there are only AI self-driving cars on the public roads. Currently there are about 250+ million conventional cars in the United States alone, and those cars are not going to magically disappear or become true Level 5 AI self-driving cars overnight.
Indeed, the use of human driven cars will last for many years, likely many decades, and the advent of AI self-driving cars will occur while there are still human driven cars on the roads. This is a crucial point since this means that the AI of self-driving cars needs to be able to contend with not just other AI self-driving cars, but also contend with human driven cars. It is easy to envision a simplistic and rather unrealistic world in which all AI self-driving cars are politely interacting with each other and being civil about roadway interactions. That’s not what is going to be happening for the foreseeable future. AI self-driving cars and human driven cars will need to be able to cope with each other.
For my article about the grand convergence that has led us to this moment in time, see: https://aitrends.com/selfdrivingcars/grand-convergence-explains-rise-self-driving-cars/
See my article about the ethical dilemmas facing AI self-driving cars: https://aitrends.com/selfdrivingcars/ethically-ambiguous-self-driving-cars/
For potential regulations about AI self-driving cars, see my article: https://aitrends.com/selfdrivingcars/assessing-federal-regulations-self-driving-cars-house-bill-passed/
For my predictions about AI self-driving cars for the 2020s, 2030s, and 2040s, see my article: https://aitrends.com/selfdrivingcars/gen-z-and-the-fate-of-ai-self-driving-cars/
Returning to the topic of an olfactory sensor or e-nose for an AI self-driving car, let’s consider what this sensor might consist of and how it might be utilized.
No ‘Machine Olfactory’ Device Exists Today
First, be aware that there is not yet an e-nose or true “machine olfactory” device in existence today.
You can today purchase a low-cost electronic sniffing device that might be devoted to a particular scent and performs a very rudimentary version of an e-nose (I wouldn’t even consider it an e-nose, maybe a small part of an e-nostril, at best). One such example would be a meter used to detect natural gas. These kinds of devices are usually handheld, meaning they are mobile, and you often need to bring to the attention of the device an odor, which it might or might not be able to register (you often have an antenna-like stick that acts as the electronic sniffer). These devices can be used by those working in the field that need to quickly gauge how much natural gas might be in the air. There are also versions that can be outfitted to be used in a manufacturing facility or in an energy plant (sometimes mobile, sometimes affixed or anchored in place).
I’m aiming to discuss herein instead a true electronic nose that could potentially detect the range of odors that a human could detect. It is said that humans are able to have around 10,000 odors that they gradually become familiar with. This is also why as you get older you tend to get less excited about odors per se, since you’ve potentially already detected them and mentally cataloged them. At an older age, encountering a new odor is often a rare treat and one that you ought to give some due attention to, since if you’ve had much of a traveled life you likely have already gotten used to most everyday odors.
Imagine a true e-nose device that is somehow part of or attached to an AI self-driving car. It is to be considered “equal” as a kind sensory partner in comparison to any other sensory device on the self-driving car. Just as the cameras have their own particular capabilities and limitations, so too would the e-nose device. Similarly, just as the cameras and radar and the rest all have specialized software that collects their data and interprets it, so too would the e-nose device.
There might be multiple e-noses associated with the AI self-driving car. You might have ones that are placed at the front and the rear of the self-driving car. Besides being outside of the AI self-driving car, there could also be e-noses inside the AI self-driving car. The number of such e-noses would partially depend upon the cost of each one, along with its size and electrical power requirements.
This brings up a point that some would argue provides reasons to not put an e-nose device onto or into an AI self-driving car. Namely, if the device or multiple such e-noses are going to be physically in or on the self-driving car, it adds weight to the self-driving car. It also chews up electrical power. The size might also mean that you need to reshape areas of the car to accommodate the devices. Will the added weight, the changes in the shaping, and the power consumption be worth whatever the e-noses devices might provide as a benefit for the AI self-driving car?
The jury is still out on that question. Admittedly, since we don’t have these e-noses today, it is hard to imagine an ROI (Return on Investment) calculation that would lend itself toward concluding that yes, the e-noses would be overall handy to have on-board an AI self-driving car. Thus, even if we had the technology to do it, the value provided for an AI self-driving car might not warrant having them.
You also need to consider the AI effort related to the e-noses. Once the sensory devices do their detection, they then feed into the sensor fusion portion of the AI system. This is when the various sensors are considered in terms of trying to tie together their respective indications. Maybe the camera shows a shadowy image ahead, but it is fuzzy, meanwhile the radar indicates that there is an object there and the LIDAR likewise indicates the same. The sensor fusion tries to assess the validity of each of the sensor indications and combine them in a manner that helps with the AI driving of the car.
If we add the e-noses to the pack of available sensors to be considered, you are now increasing the effort for the sensor fusion. You are also then presumably going to need to enhance the virtual world model to be able to include indications about odors that are detected. This might include the suspected nature of the odor, the intensity, the direction and distance of the source of the odor, and so on. The AI action planning portion would then need to inspect the virtual world model and try to ascertain whether the added layer of odor information provides any indications about how to best direct the motions of the self-driving car.
Does the added effort for the AI add-up to something useful? If not, you are then further burdening the AI, and you are using up potentially needlessly the precious processing cycles and memory on-board the self-driving car. There is only so much on-board processing bandwidth available and presumably it should be devoted to the primary mission of driving the car.
We need to then consider to what use the e-nose would be put and whether it adds to the mission of being able to drive the self-driving car.
If the e-nose does not add to the overall aspects of being able to properly and safely drive the self-driving car, it does not though ergo mean that we can therefore entirely opt to not consider the e-nose devices. It could be that we might want the e-nose for purposes other than as part of the driving task, such as perhaps to just notify the human occupants of the self-driving car on an informational basis or to alert them, and not otherwise because it needs to be contributing directly to the driving task aspects.
Suppose the e-nose device detects smoke that is wafting outdoors. This could be an indicator of a nearby fire. I gave earlier the example about being in the woods and smelling smoke, but of course a fire can occur anywhere. One day I was driving through my community and smelled fire and realized that one of the homes was on fire. I was about to call the fire department but then the fire trucks came roaring into the community and I realized that someone else had already fortunately alerted them.
The AI could potentially consider whether the detection of smoke outdoors of the self-driving car was of potential concern. It would need to consider the intensity of the smoke and other factors to try and ascertain whether the smoke was of any particular danger. This might then involve changing the intended driving path or might involve conversing with the human occupants and trying to jointly decide what driving changes might be needed, if any.
For conversing with an AI self-driving car to give driving commands, see my article: https://aitrends.com/selfdrivingcars/car-voice-commands-nlp-self-driving-cars/
For the socio-behavioral aspects of humans instructing AI self-driving cars, see my article: https://aitrends.com/features/socio-behavioral-computing-for-ai-self-driving-cars/
For humans helping to teach AI self-driving cars via Machine Learning aspects, see my article: https://aitrends.com/ai-insider/human-aided-training-deep-reinforcement-learning-ai-self-driving-cars/
For more about Machine Learning and AI self-driving cars, see my article: https://aitrends.com/ai-insider/occams-razor-ai-machine-learning-self-driving-cars-zebra/
The e-nose might detect the smell of gasoline. It might detect the smell of natural gas. It might detect the smell of a skunk. It might detect the smell of a bakery. These are all examples that I had mentioned earlier about using my actual human nose and could be applicable for the use of an e-nose.
There are lots of other smells that could be handy for the AI self-driving car to be aware of.
The AI self-driving car might have a library of smells that the e-nose device is familiar with. Over time, perhaps via Machine Learning (ML), the e-nose would get better at detecting a wider range of odors and also get better at determining what those odors represent. Also, via the use of the OTA (Over-The-Air) updating electronic communications with the cloud of the auto maker or tech firm that made the AI self-driving car, the e-nose could get updates and potentially leverage what other e-noses across an entire fleet of cars has gleaned.
For more about OTA, see my article: https://aitrends.com/selfdrivingcars/air-ota-updating-ai-self-driving-cars/
For federated Machine Learning and AI self-driving cars, see my article: https://aitrends.com/selfdrivingcars/federated-machine-learning-for-ai-self-driving-cars/
For the Turing test and AI self-driving cars, see my article: https://aitrends.com/selfdrivingcars/turing-test-ai-self-driving-cars/
I suppose the e-nose could be used for other larger-scale purposes, such as for environmental studies. You might ask people with AI self-driving cars to consider volunteering to use their e-nose devices to try and ascertain various odors in a particular area. The odors might be an indicator of pollutants or something else going awry in the environment.
For the use of the e-noses within an AI self-driving car, it might be able to detect if somehow deadly gases are leaking into the interior cabin of the self-driving car and do so before it reaches a killer level. It might be able to identify who is in the self-driving car by their own distinct body odors and then use this as a form of personal identification for giving commands to the self-driving car. I am sure we could all come up with other uses for the e-noses inside a car.
For my article about edge problems, see: https://aitrends.com/selfdrivingcars/edge-problems-core-true-self-driving-cars-achieving-last-mile/
For more about privacy, see my article: https://aitrends.com/selfdrivingcars/privacy-ai-self-driving-cars/
For the burnout of AI developers, see my article: https://aitrends.com/selfdrivingcars/developer-burnout-and-ai-self-driving-cars/
When I mention this idea of e-noses to AI developers that are making AI self-driving cars, they raise their arms in protest and say that they do not need another thing to worry about when it comes to sensors and sensory devices. They’ve already got their hands full. Right now, the focus is to get an AI self-driving car to successfully drive down a road and not hit anyone. If an e-nose won’t directly help with that focus, it is pushed down on the list of things to consider.
Indeed, you could rightfully list this as an edge problem. An edge problem is at the edge or corner of what you are otherwise trying to solve. Since we don’t even have e-noses today, there is not much need to be considering what to do once they arrive, assuming they do.
For some AI developers, they are intrigued by the possibility of an e-nose. It is an as yet untapped avenue of a new kind of sensors. You could be on the ground floor of something that one day we take for granted. Rather than the snickers and laughter that I get today when I bring up the topic of electronic noses, it could be that in some future time we will take them for granted. We’ll wonder what life was like beforehand, without having an e-nose device on us all the time, perhaps your smartphones and smartwatches will have them built-in.
The e-nose could raise privacy questions. Why should an electronic device be sniffing me? What business is it of yours that your self-driving car has detected odors in an area? It might be a device that has disadvantages in addition to advantages. Perhaps the ready detection of odors could lead to some form of widespread discrimination. Who knows?
For the moment, I’ll just point out again that our ability to smell is a primal sense. Humanity and most creatures rely upon the ability to smell things. It is the last frontier for being able to replicate our biological senses into something electronic. I’d say it is worthwhile keeping our eye on progress being made. Let’s just hope that the whole thing doesn’t end-up stinking.
Copyright 2018 Dr. Lance Eliot
This content is originally posted on AI Trends.