Kits and AI Self-Driving Cars


By Lance Eliot, the AI Trends Insider

Suppose that you invented an innovative device for automobiles that every car owner wanted to eagerly buy so they can put it onto or into their car. Let’s call the invention the thingamabob. Congrats on inventing something that everyone wants.

You decide to price the thingamabob at one dollar. There are about 270 million cars in the United States alone, and so after you’ve sold your thingamabob to every U.S. car owner, you have a revenue of around $270 million dollars. That’s impressive.

Turns out that instead of pricing it at one dollar, you opt to price it at $100 per thingamabob. In that case, your revenue is a whopping $27,000,000,000 (that’s 27 billion dollars!).  But, wait, you realize that your invention is worth its weight in gold, so you price it at $1,000, and now your revenue is $27 trillion dollars. After further thought, you decide that it should really be priced at $100,000, and now your revenue is in the stratosphere of trillions and in the quadrillions and upwards.

That’s a lot of money.

And in fact, this is illustrative of why there is a feverish desire to develop a car kit that would turn everyday automobiles into AI self-driving cars. Imagine that if you could put together a car kit that would allow conventional cars to be converted into becoming AI self-driving cars, you’d be sought by all. Auto makers would rush to your door. Tech firms would be at your feet. You’d be beloved by all those that are craving to get to AI self-driving cars for the masses. Probably a Nobel prize would be in your future.

If you hear or see something that says the XYZ Company is developing a kit for AI self-driving cars, on the one hand it is bound to perk your interest, but on the other hand I’d suggest you take a critical eye to whatever you are told or hear in this regard. The rush toward this particular pot of gold is filled with many misleading claims, and some outright false claims, and a hefty dose of pure scams.

So, before you plunk down big bucks as an angel investor or private equity or venture capitalist, make sure to read the fine print about whatever kind of AI self-driving car kit that someone is trying to say they are inventing. I’ll be generous for the moment and say that the existing claims of those working on kits is that they are generally sincere in their pursuits, though at times it can be a bit hazy about what they are truly working on.

At the Cybernetic Self-Driving Car Institute, we are advising firms that are desirous of developing AI self-driving car kits (and, we are also making some of the elements for AI self-driving car kits).

Allow me a moment to explain what AI self-driving car kits are all about.

Defining Self-Driving Car Kits

I describe the AI self-driving car kits marketspace as consisting of three fundamental approaches. Take a look at Figure 1 and look at the tree structure that’s on the left side of the diagram.

There is the Add-on Kit approach, which consists of add-on elements that are not already native to the car itself and involves installing the kit elements onto and into the car. This add-on kit is typically provided by a third-party and not the actual auto maker. An auto maker could certainly opt to make and sell the add-on kits, and I am not ruling out the possibility, so I am just suggesting that right now these AI self-driving car kits are being crafted by third-parties, typically a tech company.  I’d predict that anyone actually successful in providing an AI self-driving car kit is a candidate to be bought up by the auto maker or bought up by some other larger tech firm or other firms interested in the auto industry. It’s a sweet spot, for sure.

For the add-on kit, the question arises as to who is going to actually install the kit onto your car. This divides into the do-it-yourself (DIY) kit versus the do-it-for-me (DIFM) kit. For AI self-driving cars, the odds of making a kit that is DIY is pretty low right now. I say this because by-and-large a kit to genuinely do any real AI self-driving car aspects is going to be complex to install and setup. Having your average consumer be able to do this is unlikely. Even having a car hobbyist do this is unlikely. The more likely scenario currently is that it would be kit that involves DIFM. You would need to take your car to some expert installers that are versed in the kit and the nature of the car that the kit is aimed to work on.

When I earlier mentioned the thingamabob, I discussed the pricing aspects of $1, $100, $1,000, and $100,000. If a kit costs $1,000, and the installation effort costs an additional $10,000, the notion of considering the kit cost alone is admittedly understated. You need to also consider the cost for installing and setting up the kit. In this simple example of a $1K kit and $10K install, the effort to do the install and setup dwarfs the actual product cost. We’ll come back to the topic of costs later on.

The second approach is the Retrofit. This is typically something done by the auto maker and involves taking an existing car line and retrofitting it with the AI self-driving car kit. The auto maker or its authorized installation partners would do this retrofitting. One advantage is that the retrofit is likely to be more emmeshed into the car, versus the add-on kit, and tends to be less obtrusive and seemingly as though it really was almost a part of the car when the car was first developed.

The third approach is the Fully Integrated kit. In one sense, this is not necessarily even a kit in the traditional sense. The car when designed has been developed with the AI self-driving car capabilities from the ground-up. Those elements that are there for the AI self-driving car purposes are technically considered part of the “kit” that made the car into an AI self-driving car.

AI Self-Driving Car Kit Markets

For those of you interested in AI self-driving car kits, you might consider the nature of the market for those kits. As further shown in Figure 1, there’s a pyramid that depicts the market aspects.

If you could make an AI self-driving car kit that could fit onto and into any kind of conventional car, regardless of brand or model, you’d have hit the bonanza. This would suggest that you’d be able to possibly sell this kit to any of the 270 million car owners in the United States alone. Thus, I show the “any car” is at the top of the pyramid as it has the largest market potential.

Below the “any car” is the specific brand segment. In this case, the AI self-driving car kit only works for a particular auto maker’s cars and for one or more of their specific brands. Beneath that segment would be the narrower arena of just a specific model of a car that’s within a specific brand that’s within a particular auto maker. We could also toss into this the year of the car, since it is likely going to be harder to have a kit that applies equally to older versions of a particular car. And so on.

Now, don’t be discouraged if you are aiming at an AI self-driving car kit that’s for a particular type of car. Though it means that your market size is smaller than if you aimed larger, you still have a potential of a sizable market. Think about all of those American muscle cars like the Ford Mustang that are in circulation and still being sold. It would still be a hefty sized market if you decided that was your target for your AI self-driving car kit. Plus, presumably, once successful there, you could hopefully reuse some of your kit toward making the same kind of thing for other lines of cars.

Scope of an AI Self-Driving Car Kit

One of the major confusions about AI self-driving car kits is the aspect of what they actually do. For someone to say that they are making an AI self-driving car kit, it sounds like a really ambitious effort and implies that they will magically turn a car into a fully true AI self-driving car. Not so fast!

Take a look at Figure 2.

On the left side of the figure, I show the levels of self-driving cars. This is the standard promulgated by the Society for Automotive Engineers (SAE) and has become well accepted in the self-driving car industry. There is the lowest or least level of automation that is referred to as Level 0, essentially no automation, and ranges up the highest or most level of automation, referred to as the vaunted and desirable Level 5. Pretty much, anyone that is serious about self-driving cars is seeking to reach Level 5.

It’s great to get to the other levels, such as the Level 3 and Level 4, and there is lots of money to be made by automation at those levels, but the big fish, the kind of kings, that’s the Level 5 (a self-driving car that involves AI that can drive a car as though a human could and requires no human intervention in the driving task).  I don’t want to discourage anyone from aiming at the Level 3 and Level 4. Some say that we need to get there first, before we can get to Level 5. Others say that Level 3 and Level 4 are a distraction and thus forget about those and aim at the Level 5. We’ll see how this plays out.

The reason it’s crucial to understand that there are levels of self-driving car is so that you can ferret out the claims of someone that says they are making an AI self-driving car kit. What does their kit do? Does it make a self-driving car that is at Level 5? Or at Level 4? Or at Level 3? Or what?  You cannot assume that when someone says they have a kit for turning a car into a self-driving car that it means it is turning the car into a Level 5.

Unfortunately, the mass media usually grabs ahold of announcements by an AI self-driving car kit maker and generates bold headlines that the AI self-driving car kit is nearly here. When you dig into it, you’ll likely discover it’s not at all a Level 5 oriented kit. Some of you might remember the tremendous buzz when George Hotz was pursuing his self-driving car kit in 2016 and also saying that it would cost just $1,000. This produced huge headlines. How exciting that the classic “put together in a garage by a bunch of nerds” type of story was going to leap us all forward into the Level 5 self-driving car world. It was touted by the media as a kit that would be as easy to install as assembling Ikea furniture. As those of us in the self-driving car industry knew, it wasn’t aimed at Level 5 and there were other aspects about its scope that the major media gave scant attention to.  The story did though provide inspiration for many that hadn’t thought it possible to even consider making a kit, and so in that sense the headlines spurred others accordingly.

In Figure 2, on the right side of the diagram, I show that the AI self-driving car kits tend to make a base assumption about the existing level of the car, and then by installing and using the kit that you can turn that car into a higher level. Let’s use a notation of stating that the assumed base level of the car as a L0, L1, L2, L3, L4, or L5, and the target is also one of those designations, and we’ll use a left-to-right reading protocol of stating the base first, followed by a colon, ending with the target level.

As an example, consider this: L2:L4.  This means that a particular AI self-driving car kit assumes that the existing car is at a Level 2, and that by installing and using the kit that the car will act like a Level 4. Another example would be L3:L5. That’s a kit that assumes a Level 3 car and then upon installing and using the kit we’d have a Level 5 car.

This is a handy nomenclature. Whenever anyone starts claiming they are making an AI self-driving car kit, you should ask them what is the assumed base level and what is the targeted level. These are radically different in that taking a Level 2 car to a Level 5 (L2:L5) is going to be generally a much harder feat, in comparison to say taking a Level 4 to a Level 5

(L4:L5). One of the main reasons why the feat is much harder for L2:L5 than a L4:L5 is that the higher the level the car already is, the more that the kit can leverage about what is already there. If the kit has to take on more of the core aspects of a self-driving car, it is going to be usually be much more complex and elaborate. Easier to just build on top of what is already there, and thus the further along the car is to start with, the better it can be for the kit.

Mixture of Car & Kit

Part of the reason that an AI self-driving car kit is going to be more particular and less of a universal type of kit involves the aspect that the kit needs to integrate with the car. The specifics of the car controls on a car made by one auto maker are going to be different than the car controls of another car maker. The nature of the electronics systems is going to be different. And so on. It’s a mess in that there’s not really just one standard way that all cars are designed and made.

Imagine if you were making an add-on for a smartphone. You might create the add-on for the Apple iPhone. But, it’s unlikely it would then work for an Android phone. Trying to make something universal for things that aren’t universal in their design and approach is a tough act. For now, expect that most of the AI self-driving car kits are going to be targeted at a specific make of car. That’s reality.

Take a look at Figure 3.

We define three zones of the mixing of the car and the kit capabilities. There’s Kit Zone A, which consists of a “wide kit” that has to do a lot and the car itself is not bringing much to the table for doing the AI self-driving car aspects. There’s the Kit Zone B consisting of a narrower kit due to the aspect that the car already has some amount of AI self-driving car capabilities and the kit leverages those. Then there’s the Kit Zone C that involves a car that has a lot of AI self-driving car capabilities and so the kit can do less. All of this is a relative notion, meaning that without respect to the levels of the car, this is just saying that there are more or less elements needed in the kit as dependent upon what the car provides.

The types of AI self-driving car kit elements can be divided into hardware and software, as shown on the right side of Figure 3.

The hardware typically involves this:

  •         Sensor Devices
  •         Communications Devices
  •         Computer Processors
  •         Power Devices
  •         Car Augmentation Parts
  •         Attachment Devices
  •         Etc.

The software typically involves this:

  •         Sensor Software
  •         Communications Software
  •         Specialized Operating System
  •         Power Management Software
  •         Sensor Fusion Software
  •         AI System Software
  •         Car Controls Software
  •         Security Software
  •         Etc.

There might be additional hardware and additional software elements beyond what I’ve listed. There is also the chance that there might be less elements involved. A kit might be only taking on a smaller piece of the automation and so it is aimed at just say the sensors.

This brings up the other notion that there are the all-in-one kits and there are the subcomponent kits. You might have someone that makes a kit to put LIDAR onto your AI self-driving car. Someone else makes a kit to do the car controls. These various kits might be made and sold individually. There is also the possibility that someone combines those together into one all-in-one kit.

Keep in mind too that some of the kit elements might be intended as being a supplement to what is already on the car, while other kit elements are intended to replace the original equipment parts. You might for example put a supplemental element that attaches to the accelerator pedal and will push down on it to cause acceleration, and other such supplemental element that attaches to the brakes and one that attaches to the steering wheel. Or, it could be that someone makes a replacement accelerator pedal, a replacement brake pedal, and a replacement steering wheel, which then replaces those already in the car and you make use of these new elements instead.

Important Safety and Liability Concerns

There is already a huge conventional automotive aftermarket aimed at augmenting cars and also aimed at simply replacing the wear-and-tear parts of conventional cars. There has been an increasing rise in the distance driven per conventional car and the average age of cars is around 11.6 years in the United States, trending toward people hanging onto their cars longer. Cars being kept longer and being driven more is a good thing for the car aftermarket. Some statistics say that we spent $300 billion per year just a few years ago and that by the year 2024 it will jump to $680 billion.

Just imagine how high the auto aftermarket could go if we start to see viable AI self-driving car kits being sold.

Speaking of which, I promised earlier that I would poke further into the costs of buying an AI self-driving car kit.

Realistically, a kit for $1,000 is going to be a marginal kit, meaning that it will not do much in terms of making a car into a true AI self-driving car. Some of these low-end kits are aimed at making a conventional car become a Level 2, so it purports to do this: L1:L2. Well, you might not be very happy with this kind of AI self-driving car, especially when you look closely and realize that there are a ton of limitations as to what the kit provides – for example, the kit assumes that there are lane markings on the road, and so the moment that you are driving along and the road markings aren’t there, you as the human driver need to right away take over the controls.

The odds are that a true AI self-driving car kit is going to more akin to costing in the tens of thousands of dollars. Let’s pretend for the moment that a kit comes to the market at a cost of $100,000 to purchase it (ignoring the installation costs for the moment). You might say that few people could afford the $100,000 and so this is not going to sell very well. That’s partially true. Keep in mind that presumably a true AI self-driving car could become a ridesharing vehicle that generates revenue for you. You might use your now transformed self-driving car to take you to work, and then the rest of the day it is being used by others that are paying to use your self-driving car. It might therefore make financial sense to go ahead and buy that $100,000 kit, doing so via some kind of loan or other financial approach, knowing that it will ultimately “pay for itself” possibly.

What we haven’t yet covered in all of this discussion about AI self-driving car kits is the bigger questions about safety and liability.

Would you be willing to trust that your car when using a $1,000 self-driving car kit is going to properly work on our roadways and you’ll be perfectly safe in using it?

Right now, people are reluctant to get into a self-driving car that has a gigantic auto maker or tech firm backing it, which they’ve invested millions upon millions in developing, and so imagine that you are going to use a self-driving car that was augmented by a kit from an unknown start-up. I am not saying that you should be saying no, and only bringing up that we would want to know that somehow this thingamabob has been sufficiently tested for safety purposes.

Where does an AI self-driving car kit fit into the federal regulations about cars, and the state and local regulations about cars? Who would certify that the use of the kit will produce a safe self-driving car? What does safe mean in this context? There is a myriad of such questions.

The zillion dollar question involves liability.

If the kit is put onto your car, and your now transformed self-driving car gets into an accident, who has the liability for the accident if the kit played a part in the cause of the accident?  You could argue too that the kit might only be 1% of the cause of the accident, or maybe it’s 100%, but either way the kit is going to certainly be considered a factor in whatever accident happens.

Are you as the owner solely responsible for whatever the kit did? The odds are that everyone else in the food chain is going to get some liability attached. The maker of the kit. The installer of the kit. A mechanic that maybe worked on your car and somehow touched the kit. A previous owner that maybe put the kit into the car and you now are claiming that you didn’t know that the kit would do this or that. The list of culpable parties is pretty long.  An astute lawyer is bound to go after any of those in the food chain, regardless of whether there is substantive proof per se that the kit was a factor or not.

In short, the path to viable AI self-driving car kits is going to be a gauntlet that involves navigating the safety issues and the liability issues. I suppose you could say that with the prospects of big bucks comes the dangers of big risks.

Will we soon see AI self-driving car kits for sell on Amazon and eBay Motors (those are considered the two largest sellers of aftermarket auto parts and accessories in the United States)? Well, if you do, make sure to consider what the kit actually does, look at the fine print, and consider the safety and liability aspects.

If we could wave a magic wand and have AI self-driving car kits, it would be a tremendous boon towards getting us toward a world of AI self-driving cars. There is pent up demand for this. The auto makers and tech firms realize this too. Many of them are making investments in firms that are doing some kind of kits, whether an all-in-one or piecemeal, whether an add-on versus retrofit, etc. These firms all realize that there is a pot of gold to be found. Let’s get to the pot of gold with a mindset of ensuring that the sought riches come with a need to be safe and sound.

This content is originally posted on AI Trends.