Keep up with new content on the site, receive exclusive content and commentary, and learn about activities within the Straight Talk community.
By Dr. Lance Eliot, CIO and CTO, Techbrium
This article is by Featured Blogger Lance Eliot from his AI Trends column. Republished with the author’s permission.
As head of the Cybernetics Self-Driving Car Institute and a frequent speaker about self-driving cars and autonomous vehicles, I often get asked the question of why are we now seeing such a widespread interest and advancement in self-driving cars. Some inquirers feel that this is like suddenly discovering a new movie star and wonder what sparked that person to vault into stardom. Tagging onto that analogy, I explain that just like the proverbial small-time actor that starved and took on any off off-Broadway acting roles they could find, it has taken many years of toiling in research labs and universities that has preceded the now more visible appearance of self-driving cars.
Since the invention of the horseless carriage, there have been dreams of someday having a car that can drive itself. During the pre-computers era, attempts to develop a self-driving car were pretty much DOA (Dead on Arrival), since the kind of technological capability to achieve self-driving cars did not yet exist. During the early days of the introduction of computers, researchers realized that the potential for a self-driving car reasonably now existed, doing so by harnessing computers to act on behalf of a human driver. If you take a look at the body of literature on autonomous vehicles, you’ll see that there have been hundreds of academic and research institutions and thousands of professors and researchers that have been pursuing the dream of a self-driving car for years and years. One of the most famous instigators towards self-driving cars has been the Department of Defense (DoD), for obvious reasons of battlefield purposes, along with the DARPA sponsored competitions that have helped to push innovations in robotics forward immensely and that are directly pertinent to self-driving cars.
So, my first point is that it is not as though we all woke-up in the last year or two and suddenly decided to invent self-driving cars. The desire for a self-driving car has been around for a long time, and the advances toward it have been incrementally advancing. That being said, it has been a slow and snails paced progress toward a self-driving car. No overnight successes here. Inch by inch, we continue to make our way toward the self-driving car. I say this because some think that maybe there was a “silver bullet” that finally opened the door for a self-driving car to emerge. I know that some will claim that perhaps neural networks should be the winner for anointing self-driving cars as viable, while others would say that it is instead LIDAR (see my column on LIDAR for self-driving cars), and some would offer other singular aspects of technology to assert that is the “it” that triggered the self-driving car craze.
Though it is often easiest to try and simplify the world and make the claim that one particular innovation led to a new world order, in this case I argue that anyone laying the credit at the feet of just one advancement is either ignorant about the field of self-driving cars, or miscomprehending things, or pushing a particular love-fest piece of high-tech, or has not taken a contemplative moment to reflect on what has taken place and continues to take place in the self-driving car arena. If you really step back and take a macroscopic look at the self-driving industry, you would come upon the notion that where we are today can be described in two words.
There, that says it all. It is a grand convergence. There have been a slew of key high-tech advances, combined with societal and business aspects, all of which have come together to create a circumstance and ecosystem that allows for the emergence of self-driving cars. Each of the members of this grand convergence have contributed mightily. No one in particular reins more supreme. At the same time, if some of the members were not present, it is questioned whether we would now be as far along as we are. Self-driving cars would still be toward infancy rather than maturing toward practical reality. Like links in a chain, each member of the grand convergence has made a contribution. Any contribution that had been missing would have left a missing link and we might not be at this pivotal juncture of nearing the realization of self-driving cars.
Allow me to also clarify that when I refer to self-driving cars, you need to know that there are an array of differences of meaning about what constitutes a self-driving car. As stated in my column on the Richter scale of self-driving cars, we are only in the mid-way range of the levels of self-driving cars. Right now, self-driving cars are around levels 2 and just poking into level 3 (per the official SAE scale). We still have a fight on our hands to get to level 4. And, getting to level 5 is like a moonshot. Don’t let anyone trick you into thinking otherwise. Even though each day there seems to be wild claims about a level 5 self-driving car coming upon us any day, it will be many more years before we see a true level 5 self-driving car. Mark my words!
What then are the members of the club of grand convergence? The membership includes various technologies. Technologies though must be understood within a context of existence. If I invent a better mousetrap, but the mousetrap is so expensive that no one can afford it, the technology will be waylaid until it somehow reaches a point of being more affordable. Thus, the technology must also be understood within a context of the social and business factors that allow for the technology to be deployed.
Here’s my list of the members of the grand convergence that is leading us toward self-driving cars:
Size of sensors
The sensors that go onto and into a self-driving car have been getting smaller and smaller. This is significant because they are easier to place onto and into a car, they add less weight, and they don’t cause a car to become the size of a truck just to have the sensory capabilities needed to be a self-driving car. If you look at the self-driving cars of a few years ago, you can see how bulky those sensors once were. These sensors continue to be miniaturized and more readily used for self-driving cars.
Price of sensors
The sensors for self-driving cars used to be immensely expensive, meaning that if you wanted to have a self-driving car that the cost of the sensors alone would make the price of the car be astronomical. In some cases, the sensors come to a million dollars in cost. Now that sensors are getting less expensive, it becomes more realistically viable to have an affordable self-driving car.
Speed of sensors
The sensors for self-driving cars are getting faster and faster. The speed of capturing data is crucial since a car might be zooming along at 80 miles per hour and the self-driving car has to in real-time collect and process the data. We’ll continue to see the sensors speed-up.
Size of processors
If you wanted to put a vacuum tube based computer onto a car, you wouldn’t even know there was a car underneath it. In that sense, the size of computers during the last 30-50 years has made a big difference in everything that is computer-based, including for example our cell phones. Likewise, the number of processors needed for a self-driving car is quite high, and so the miniaturization of processors is helping to make them available within self-driving cars.
Price of processors
The cost of computer processing continues to drop dramatically. That’s why we see them in the Internet of Things (IoT) too. Self-driving cars need gobs of processors, and so the decreasing price of processors is making this possible.
Speed of processors
I feel the need, the need for speed. Processors inside a self-driving car are doing a lot. They need to analyze the sensory data. They need to run the AI software that allows the car to drive. All of this requires very fast processors if the self-driving car is going to contend with driving in real-world environments. Processors are getting faster and faster, fortunately.
Some self-driving cars are independent of the Internet and don’t need such interconnectivity to do what they do. On the other hand, it is more than likely that true self-driving cars will need to have some kind of interconnectivity, presumably via the Internet, but could be via some other means. Perhaps the most notable aspect of this would be the vast amount of data that a self-driving car is collecting and its ability to then share that with a centralized system, which can analyze it, and provide not only insights to the contributing self-driving car but also do likewise for a vast network of interconnected self-driving cars.
Within the field of AI, machine learning continues to provide capabilities to aid self-driving cars. The notion is that rather than trying to program explicitly whatever a self-driving car needs to know, we can have a self-driving car become “capable” by learning from data about driving. There are various techniques of machine learning which are now competing with each other to see which techniques provide the greatest benefit for advancing self-driving cars.
Neural networks are one kind of machine learning type of technique and are perhaps the most known or discussed approach. This is an effort to try and mimic somewhat how the human brain works, by simulating neurons in a network like way. It used to be that simulating neural networks was hard to do in-the-large because of the computational processing needed, but with advances in processors we’ve been able to make this more possible. This is why for “deep learning” we can have much larger neural networks, of which the results are more impressive than were the earlier smaller or more shallow ones.
You probably have heard that we are in the era of algorithms. Much of what runs our society on a computer based aspect is based on algorithms, ascertaining what actions should be taken by systems. Likewise, for self-driving cars, the advancement of algorithms and the ready availability has made them usable for self-driving cars.
To allow machine learning and also neural networks to do their thing you need data, lots of data. Advances in being able to collect, transform, process, and analyze Big Data has made self-driving cars “smarter” and more capable. See my column about data and machine learning for self-driving cars.
I know this member of the grand convergence, open source, might seem strange to those that are at the periphery of self-driving cars. What does open source have to do with self-driving cars, they ask. The open source world has brought into the fold many software developers that otherwise would never have known about or been able to contribute to the software of self-driving cars. To-date, much of the software was locked away in academic research libraries, or was held close to the vest by private self-driving car makers. This opening up of such software has led to more contributors and more advances in self-driving car than otherwise would have likely occurred. See my column about open source and self-driving cars.
Automated Driver Assist Systems (ADAS)
We have grown-up with the ability to engage our cars into cruise control. Now, we have cars that can do parallel parking and other kinds of one-trick-pony car driving aspects. The advancement of ADAS is helping humans to get accustomed to having their car take-on self-driving car tasks. Those that develop ADAS see that they can get toward a self-driving car by expanding ADAS and making it more holistic.
Belief in Possibility
I would assert that a belief in being able to achieve self-driving cars is a crucial aspect underlying the advances in technology. If no researchers or developers thought it was a possibility, they would focus their energies elsewhere. By having a belief that it is possible, and that it is possible within a reasonable time frame, they are willing to devote their attention to this realm.
Though you can invent technologies without caring whether consumers will have any interest in it, when you have a potential for widespread consumer interest, it helps to raise the stakes and get the attention of a myriad of inventors and developers. Consumers are primed to accept self-driving cars. They’ve seen the ADAS advances, and they want more.
Money makes the world go around. Academic researchers that have toiled in self-driving cars have been scrapping together NSF grants and DoD grants for years have finally seen the spigot start to flow. Now, venture capitalists and other investors are agog about self-driving cars. The money is flowing. This drives the technologies and the technologists, since they can get the money needed to explore and experiment, plus they are certainly attracted to the potential for personal wealth.
Self-driving cars cannot be invented by one person alone. The number of aspects of a self-driving car means that you need lots and lots of inventors and developers to each be contributing this or that piece of the larger puzzle. Fortunately, with a herd mentality that has taken place toward the advent of self-driving cars, you have enough of a widespread army of inventors and developers that each of the pieces is coming together.
Foundational or basic research about self-driving cars has been augmented by applied research. Besides universities and colleges, we’ve seen a cottage industry of entrepreneurs that have now staked out space in the applied research toward self-driving cars.
You’ve got to give credit to Google for wanting to take-on the moonshot of self-driving cars, having been the first highly visible tech company to really take this seriously. That being said, they weren’t being altruistic, since their efforts have obviously gotten them a tremendous amount of publicity over the years, and likely attracted other top high-tech talent to them, regardless whether that talent was immersed into any of the self-driving cars aspects. For Google, it was a good bet toward high-tech that might someday pay-off and that at the same time paid-off handsomely in publicity and attracting talent. They have now also spawned many former Googlers into helping to push along self-driving cars by having made the leap to other firms or going forth with their own new ventures.
Elon Musk is quite the character and his charm and publicity-getting attention has made the self-driving car craze what it is today. Tesla is a pioneer that has pushed forward on ADAS in a manner that we would likely not have had any of the conventional car makers try to do. Musk is a visionary that has taken a chance at spending on something earlier than most thought would make sense to do. The jury is still out whether Tesla will make the leap toward a true self-driving car, and some think maybe they’ll get eclipsed. In any case, the Tesla remains a formidable challenger and something that has opened the eyes of the public, the investors, and the inventers about the strong possibilities of self-driving cars. Go, Elon, go.
Self-driving car developers and inventors have pretty much been working in the shadows for many years. Not much interest by the media. A little bit of an advance here or there sometimes caught the attention of the media, for a moment. Without media coverage, it is hard to get a large preponderance of interest toward these advances and the technologies and technologies that make them happen. The media now seems to be in a frenzy to cover any and all advances of self-driving cars. I’ve though criticized some of that media coverage as being fake news, see my column on that topic.
What does ride sharing have to do with self-driving cars? You betcha it has a lot do with self-driving cars. Ride sharing has brought to the forefront of our attention that we depend upon our cars and that there is a need for convenience of car access. With the billions of dollars that have flowed to ride sharing, it has coincided with the realization that if we had self-driving cars it would profoundly impact ride sharing. Ride sharing is helping to drive attention toward self-driving cars. See my article on the topic of blockchain and self-driving cars.
I have now laid out my list of the members of the grand convergence. Do you have other members you’d like to add to it? Be my guest. Do you believe that some of the member don’t belong on the list? I think I’ve made a pretty strong case for each one, and keep in mind that I am not saying one is the most significant, I am saying they all had and have a contributory role toward the “sudden” appearance of self-driving cars.
Welcome to the bandwagon for self-driving cars. You can contribute too. No need to be a propeller head, since as you can see from my list, there are lots of ways in which the self-driving car realm is being formulated and grown. I hope that mainly I have dispelled the prevailing myth that there has been one factor alone that has led to the advent and now popularity of the self-driving car topic. There isn’t a magic appearance of say penicillin or the breakthrough of the invention of fire that has been the silver bullet to make self-driving cars either possible or of interest. Furthermore, the grand convergence is so strong now that I doubt we’ll see a fading of interest or advances. The appetite by everyone involved, inventors, investors, developers, media, consumers, regulators, and so on, has become so pronounced that we are now on our way to the moon, whether we want to get there or not, and we’ll continue to see advances in self-driving cars, even if it takes longer than some of the pundits have been predicting. This train is underway and it ain’t going to stop. Drive safely out there.
Originally published on AI Trends.