Dr. Lance B. Eliot, AI Insider
What did we get by landing on the moon? Smarmy answers are that we got Tang, the oh-so-delicious artificially flavored orange drink mix, and we brought back to earth about 50 pounds of rocks and dirt in the Apollo 11 mission alone.
Seriously, it is hard to imagine that anyone really would though claim that the only benefits from the moon landing consisted of a drink mix and some rocks. The world was focused on something truly inspirational.
You’d be relatively safe to argue too that the moon effort spurred advances in electronics and in computing. Galvanizing attention on an overall goal that forced along the movement of electronics and computers seems likely to have sparked and pushed forward those offshoots and spin-offs more so than if they were merely acting on everyday economic pursuits.
I don’t think we have a time machine that would allow us to somehow replay the era and pretend that there was an alternative of not going for the moon, and then see how things fared.
The overarching theme is that sometimes when you are doing one thing, there can be various offshoots and spin-offs, offering a twofer,
Let’s shift our attention right now to another kind of moonshot, namely the efforts to achieve AI self-driving cars. I’ve repeatedly stated in my writings and presentations that getting to a true AI self-driving car is very hard. Very, very hard. Tim Cook, the CEO of Apple, has been famously quoted as saying that indeed an AI self-driving car is like a moonshot. The odds of success are ambiguous, and it is not a sure thing by any stretch of the imagination.
At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars. In addition, we are identifying offshoots and spin-offs, doing so along with auto makers and other tech firms in this niche.
I’d like to first 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 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.
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.
Returning to the topic of beneficial offshoots or spinoffs, let’s consider how the AI self-driving moonshot-like efforts have a twofer built-in.
Take a look at Figure 1.
As shown, reading the diagram from left to right, there is an indication that there can be hardware related offshoots of AI self-driving car efforts, there can be software related offshoots, and there can be “transformative” offshoots.
The transformative offshoots consist of taking some innovation that originated outside of AI self-driving cars, and for which an AI self-driving maker than utilized the innovation and transformed it into something new or novel, thus there is an offshoot potential ingrained in that transformed variant which can be transplanted back into the non-AI self-driving car realm.
Let’s begin with a real-world example of an offshoot, in this case one that happens to be hardware related.
There was some eye-catching “offshoot” news recently in the AI self-driving car industry. In particular, Waymo, the Google/Alphabet autonomous vehicle entity, announced it would aim to sell or license its LIDAR sensor technology to third-parties, albeit only if those third-parties agree to not use the technology for AI self-driving car efforts.
For me, this is a loud bang of a starting gun that has gone off to highlight that the race for an AI self-driving car has also got lots of room for offshoots and spin-offs.
LIDAR is considered by many to be an essential type of sensor, combining together Light and Radar (LIDAR). You wouldn’t use LIDAR solely as the only sensor on an AI self-driving car and would instead have it act in a complimentary manner with say camera, conventional radar, ultrasonic sensors, and so on.
There are many that believe the use of LIDAR is crucial to achieving Level 5, while there are some, most notably Elon Musk and Tesla, asserting otherwise and thus Tesla’s aren’t outfitted with LIDAR. Musk though has expressed acknowledgment that he might be off-base about LIDAR and we’ll all have to wait and see whether his instincts were on-par or not.
In the case of Waymo, they are stanch believers in LIDAR. Some pundits have congratulated Waymo for taking their own path on LIDAR, allowing Waymo to presumably control and determine what the LIDAR does and how they will make use of it.
Some pundits say you’d be crazy as an AI self-driving car maker to devote your attention and precious resources to reinventing the wheel by making your own LIDAR. The number of LIDAR makers is rapidly increasing, and it seems that for each new dawn there is another LIDAR startup someplace. It’s hot.
Part of this involves historical momentum too.
In the case of Waymo, they ventured into their own LIDAR at a time when arguably the number of LIDAR options was few. You could try to make the case that they by necessity chose to take the bull by the horn. Whether they would need to do so today, well, that’s a different question.
The reason I’ve brought up the Waymo LIDAR and the announcement is due to the offshoot or spin-off notion. You might have noticed that I mentioned earlier that Waymo is restricting who can potentially purchase or license the LIDAR from them. They are excluding uses of the LIDAR for AI self-driving cars.
Why, you might be wondering? Simply stated, they aren’t willing to handover their own “secret sauce” to other AI self-driving car makers.
Okay, so we have a major AI self-driving car making entity that is willing to provide as an offshoot or spin-off their own proprietary LIDAR, as long as it is used for anything but AI self-driving cars (there might be other restrictions they’ll eventually land on, which depends on what third parties approach them about the possible usage).
Though selling or licensing their LIDAR wouldn’t likely in the near-term bring much dough, it would nonetheless showcase the inherent value of the technology and IP that Waymo has been developing.Doing so can help offset the cash burn being consumed by Waymo and provide marketplace support that Waymo is a worthwhile bet, plus aid in suggesting the kind of valuation that Waymo embodies.
There’s another angle on this too. It could be that while floating out the innovation to the marketplace, you end-up getting feedback that otherwise you would have been unlikely to get on your own.
I’m shifting away now herein about the Waymo announcement and want to cover other facets overall about offshoots and spin-offs, along with identifying other kinds of such aspects that might occur in the AI self-driving car arena.
AI Self-Driving Car Sensors Most Ripe for an Offshoot
The sensors aspects are the ripest for an offshoot. If you are making a sensor that you devised specifically for AI self-driving cars, the odds are high that such a sensor can be used in other ways and other means. The most obvious would be in other kinds of Autonomous Vehicles (AV), such as using your sensor in an autonomous drone or an autonomous submersible vehicle.
Using your sensor in another family-related AV’s is not much of a stretch, admittedly. Presumably, those should be ways that already jump out at you.
A more pronounced stretch would be to consider using your sensors in something other than a vehicle. Move your mindset away from vehicles and consider how else might the sensor be used. Could it be an Internet of Things (IoT) device that might be used in the workplace? Or maybe in the home? There is no doubt that the IoT marketplace is enormous and growing, so perhaps you can re-apply your sensor into that space.
One of the difficulties often times about brainstorming about other uses of your own internally developed innovation is that you might fall into a groupthink trap. If everyone on your team was brought to the table to develop a sensor for purposes of Y, they are likely steeped in the matter of Y. It’s all they think about it. It’s what they know best.
Trying to get them to go outside the box of Y is not usually readily done. In fact, sometimes they can be forceful about staying inside the box.
They might be right, they might be wrong.
You need to ferret out whether in fact trying to use the innovation for other purposes might be inappropriate, or whether it is just a hesitation based on an anchoring to what the team already knows. This can be difficult to discern. Trying to shoehorn an innovation into other uses might not be productive, and worse still might be untoward.
I’ve worked with some top tech leaders that were constantly coming up with new (and often wild) ideas about how they could repurpose their innovation. They’d be eating a meal and come up with another idea. They’d be on the phone and suddenly come up with an idea. They were like miniature idea generating factories.
At times this was handy and provided opportunity for adapting the innovation to some other notable use. In other cases, it was as though the innovation was a swiss army knife that could be used in a thousand ways, when the reality was that it was simply a toothpick and did not have any of the other tools, lacking a can opener, a knife, a screwdriver, and so on. I’m not saying that they could not have ultimately adapted the innovation, only that the distance was greater than was in the mind of the top leaders.
Sometimes bringing an innovation to the marketplace can be a fresh dose of reality to a top leader. Within the firm, perhaps it is hard for the staff to pushback on wild ideas. They don’t want to be pigeonholed as a naysayer. By allowing the innovation to touch into the market, it will be the marketplace that provides the needed feedback. This can get top leaders to listen and pay attention when they otherwise might have been hesitant to do so.
The other side of that coin is that sometimes the internal AI developers are so burned out that they cannot imagine taking on something new with their innovation. If you are pouring your heart and soul into a sensor for an AI self-driving car, and you are exhausted in doing so, even if there is a glimmer of promise for the sensor in some other ways, you cannot cope with the added effort that will undoubtedly fall onto your shoulders. Thus, you might subliminally nix the new use, somewhat due to basic survival instincts.
Besides sensors, there is a slew of other hardware that has the potential for being used beyond the realm of AI self-driving cars. There are specialized processors, GPU’s, FGPA’s, and the like, all of which can be applied to other fields of endeavor.
I realize that many of those hardware advances were already being done for other fields, and then were re-applied into the AI self-driving car niche. I’m not suggesting they were made necessarily initially for AI self-driving cars. In some cases, something that was made for another purpose has been brought into the realm of AI self-driving cars.
My description about the hardware aspects can be readily applicable to the software aspects.
If you develop a simulation for AI self-driving cars, based on crafting a new way of doing simulations, it could be that you can re-apply that capability to other areas.
Think about the entire software stack associated with AI self-driving cars.
You’d need to decide whether or not you want other competing AI self-driving car makers to be able to use your new-to-the-market software. Is it something that provides you with a competitive edge? Would it reveal too much about your secret sauce?
We’ve of course seen some of the AI self-driving makers that have opted to not only bring an offshoot into the marketplace but even make it available as open source.
Uber’s Autonomous Visualization System (AVS) Released as Open Source
For example, at the Autonomous Vehicle (AV) 2019 conference, I had a chance to chat with Hugh Reynolds, Head of Simulation for the Advanced Technologies Group (ATG) of Uber. After having used a number of simulation packages, they developed an internal capability that they decided recently to share with the industry.
He and his team have released as open source version of their Autonomous Visualization System (AVS). It consists of an element known as XVIZ, which is a spec that deals with the managing of generated AI self-driving car data, and includes their streetscape.gl, which provides a means to build web apps that leverage the data that’s based on the XVIZ formats. You can find these tools on GitHub (https://uber.github.io/).
There are already other AI self-driving car makers that have indicated they’ll likely be making use of the capability. Since it is open source, this reduces the qualms by those other AI self-driving car makers about necessarily getting locked into something that another maker might otherwise control. Making it open source might seem odd to some, but there is not just some kind of altruism in doing so, the odds are that this will ultimately also help Uber by spurring an ecosystem around the simulation and benefit the simulation by boosting it in ways that Uber itself might not have the time or have considered doing.
In the combination of both software and hardware, we’ve seen that the Machine Learning and Deep Learning aspects are also spurring offshoots. For AI self-driving cars, one of the most significant elements is the use of deep artificial neural networks, especially in the analysis and interpretation of sensor data. There are software tools and hardware capabilities of Machine Learning and Deep Learning that have been forged within the AI self-driving car space that are gradually coming onto the market for use in other domains.
Suppose that while the engineers and scientists were working on developing the needed innovations and high-tech to get to the moon that they opted to right away do offshoots or spin-offs?
I ask the question because it brings up an important consideration about offshoots and spinoffs. What is the right timing for having an offshoot or spinoff?
Timing the Offshoot or Spinoff
Imagine the high-tech moonshot workers in the 1960s that rather than focusing on how to control the space capsule to land on the moon, instead they became attentive to making microwave ovens that could be used in the home. Maybe we would not have gotten to the moon. Or, maybe we would have taken ten more years to get there.
The point being that if you begin to take on the aspect of doing an offshoot or spin-off, there is a chance you are risking keeping to your knitting. You are maybe taking on more than you can chew. The problem could become one of the core getting second fiddle to the offshoot, which might not have been your plan, yet you fell into it, slowly, inexorably, like quicksand.
It is easy to do. Sometimes the offshoot gets all the glory. The core use is already well-accepted within the firm. Most take it for granted. The excitement about seeing your hardware or software applied to a new domain is rather intoxicating. Top leaders can readily get caught up in the allure and begin to inadvertently drain resources and attention away from the core use.
Advances for the core use begin to get pushed aside or delayed. Maybe the quality of the updates or revisions start to lessen. The other use of the core saps the energy and willpower that got the core to where it is. Sure, the other use might be promising, meanwhile sacrifices to the core can undermine the core overall.
I caution top leaders to make sure they have their ducks aligned when they make the decision to forge some kind of offshoot or spin-off. Are they ready to do so? How much of their existing resources will get pulled away to it? Will they provide as much attention to the core as they are to the offshoot, or will they subconsciously starve the core? These are all important matters to be discussed.
The timing question is a tough one to balance. You want to bring out the offshoot while the core is still considered new and worthy. If you wait too long and the core is now already eclipsed by other substitutes in the market, you missed your window of opportunity. The timing needs to be the vaunted Goldilocks mode, not too early, not too late, just the right temperature, as they say.
Another consideration is whether the innovation if created by an internally focused team is ready to deal with becoming a business within a business. When selling or licensing your innovation to other firms, you suddenly have a whole new enchilada to deal with, meaning that you need to provide service to that customer or set of customers. Is your internal team prepared to deal with external entities that want support or otherwise require a services aspect that your team was not having to do before?
There are some doom-and-gloom pundits that say we will never achieve true AI self-driving cars. We are all on a fool’s errand, they contend. Though I disagree with their assessment, I like to point out to them that even if they are right, which I doubt, but even if they are right, the push toward AI self-driving cars is creating numerous benefits that otherwise I assert would not likely exist.
In essence, I am claiming that the race toward true AI self-driving cars has other benefits beyond whether we actually are able to achieve true AI self-driving cars.
One obvious benefit is that conventional cars are getting more automation. As much as that seems good, I’ve also cautioned that we need to be leery of automating non-self-driving cars to the degree that humans get lulled or fooled into believing the AI can do more than it really can.
AI self-driving cars are an exciting notion that has energized the field of AI. It has helped move AI out of the backrooms of university labs and into the sunshine. As a former university professor, I still maintain my roots at numerous universities, and I’ve seen first-hand how AI self-driving cars are “driving” faculty and students into areas of AI that I believe would not have gotten as much attention otherwise.
Society as a whole has been energized into discussing topics about transportation that I believe would not have been as active or headline catching, were it not for the AI self-driving car efforts. Regulators are considering the advent of AI self-driving cars, which also brings up the topic of mobility and how can our society do more for increasing mobility.
In short, similar to the real moonshot, I’d argue that the advent of AI self-driving cars has become a motivator. It has inspired attention to not just AI self-driving cars, but encompasses far more, including societal, business, economic, and regulatory aspects. This inspiration sparks innovators, dreamers, engineers, scientists, economists, and all of the myriad of stakeholders that AI self-driving cars touch upon.
Whether you will grant me that the race toward AI self-driving cars has produced those aspects or not, at least perhaps we can agree that the advances made in AI, along with hardware and software, are having and will likely continue to have a profound spillover effect. The number of offshoots and spinoffs will gradually increase, and I predict you’ll see that the AI self-driving car pursuit produces more than you might have anticipated.
I don’t think we’ll look back and say that all we got was Tang, and instead we’ll be saying that without the AI self-driving car pursuit we wouldn’t have amazing advances that we’ll be relishing in the future. Admittedly, there won’t be moon rocks to look at, but it will still be good, mark my words.
For free podcast of this story, visit: http://ai-selfdriving-cars.libsyn.com/website
The podcasts are also available on Spotify, iTunes, iHeartRadio, etc.
More info about AI self-driving cars, see: www.ai-selfdriving-cars.guru
Copyright 2019 Dr. Lance Eliot