Is it too premature to call this race? Many seem to think so. However, when one understands how the development of AI is operating, we can see how far ahead Tesla really is.
The reality is there are very few entrants in this race. What seems like a wide open field is not. The legacy automakers are not in the discussion so they can be ignored. This leaves basically Silicon Valley. The only one who is anywhere in the conversation is Google with their Waymo project.
Source
Before getting into why Tesla is the hands-down winner, let us look at a couple of industries we know very well.
In the smartphone market, what choices do we have? Basically, we can opt for either Apple's IPhone and the variety of devices that are run on Android.
If we look at the computer market, laptops and PCs give us a few choices. There is, again, Apple as well as Windows. We can also toss in Linux in there.
Notice we see a few different operating systems in each realm. Do we really believe that the autonomous vehicle industry will be any different? This is not likely to stray from what took place in the past.
Hence, we can conclude that Tesla will be one of the two or three operating systems that everything defaults to.
The second reason why we can make this statement is because of data. When it comes to this, there are three things that are required for AI training:
- Clean
- Diversity
- Quantity
Tesla excels in each of these areas.
Data For Autonomous Driving
The amount of data that Tesla gets is absolutely stunning. This is the quantity criteria.
There are roughly 1.5 million Teslas on the road, most with full self driving equipment. Even if the software activation is not purchased, it still is there. Also, each vehicle has 6 cameras which compiles data.
The key here is that it is always running in the background. This provided a comparison to how the software runs as compared to the human driver. When there are discrepancies, the AI system can compare them and make adjustments.
Another important point is the company keeps adding to the fleet. Each quarter they sell more cars, all with the same capabilities as just mentioned.
Here we see the lead is only getting larger. By the time anyone gets serious about trying to catch up to Tesla, that company might be over 3 or 4 million vehicles.
Clean And Diverse Data For AI
What makes Tesla unique is that it can interact with its large fleet and make requests. This provides them with the clean data it requires.
For example, if they realize the system is having difficult with tunnels, they can put out a request to record when entering and exiting bridges. This will help them to get the data they need without a bunch of other "noise" mixed in. They can target exactly what they want to improve the shortcoming in the system.
As we can guess, all of this leads to diverse data. With autonomous driving system, edge cases are vital. We can see where most driving will happen within a scope of normal behavior. The problem with driving is when we encounter things that are very rare but do happen.
With an ever-growing fleet of vehicles, the number of situations encountered just keep growing. This also happens in a variety of environments. This is much different from what Waymo is doing, which is operating in a "fenced" environment. Their autonomous vehicles operate in a select area that is mapped out. This allows the system to be strong in that area but will not do real well in areas it does not know.
Simulation
Some feel that simulation is a way to add to the system. Certainly that can help. Of course, this is dependent upon the computer it is being run on. While it would be presumed that Google has the advantage in this area, Tesla unveiled that it has the world's 5th fastest computer. Thus the company is holding its own against any competitors in the simulation arena.
Source
Also, since their fleet captures so many edge cases, simulations can be designed to train the system on some things that are rarely seen. For example, how often does a canoe fall off a pickup truck? That is something that the AI system can work on.
By this time, the amount of real world situations that the Tesla system was trained on is enormous. This gives a huge advantage since there is only so much computational power a company will have access it. Whereas other companies will have to focus upon getting the basics, Tesla can focus upon the edge cases.
If data is the new oil, Tesla is Standard Oil in the autonomous driving* race. When it comes to this arena, the company is only adding to an increasing lead.
While most look at this as something that would favor bigger names, the reality is Tesla, basically a Silicon Valley company, has taken the software development to another level.
This is why Tesla won the autonomous driving race.
If you found this article informative, please give an upvote and rehive.
gif by @doze
logo by @st8z