Some have requested a deep dive into what the flywheel means and how it truly works. For this reason, I am going to spend a few articles to exemplify how this can occur.
This is a topic we heard discussed on a number of the LEO AMAs. It is part of the strategic path being pursued. When it comes to digital platforms, this is a sound approach.
While Amazon is the king of the flywheel, something that will lead a future article, we will focus upon Tesla to illustrate what can happen. Specifically, we will concentrate the discussion on data.
It is no secret that Tesla is pursuing autonomous driving. Like most things in the artificial intelligence world, data is at the heart of everything. We also have compute as a critical factor.
Which brings us to Tesla. Here is where there are a series of steps forward which put the flywheel in motion.
Before getting to that, the flywheel is the concept that there is no breakthrough moment. all progress is the result of a multitude of steps. There is no singular moment. Instead of focusing upon the single "game-changer", the flywheel seeks to set a number of things in motion and have them feed upon each other.
With Tesla's pursuit of autonomous driving, data and compute are major factors.
Once we combine all the pieces, we can see the impact this can have.
To start with the potential end, a Morgan Stanley analyst made this forecast.
Tesla recently announced that its cars have driven one billion miles with the help of its full self-driving (FSD) software. Jonas sees this increasing to 400 billion miles per year or over 1 billion miles per day by 2030, offering Tesla a monumental dataset for machine learning and subsequently improving its attempts at autonomous driving. Source
The total lifetime miles of FSD is 1 billion. According to this projection, that will be the daily total in just 6 years.
How does this occur? The answer is the flywheel.
With Tesla, the first announcement was the Switch to end-to-end neural network. This altered how the training occurred.
The next component is compute. The company recently announced that it was no longer compute constrained. This means there is enough processing to train on the data it is receiving (and expects).
A final variable is the number of cars. Each quarter more are being sold. Tesla is also looking at enhancing the take rate of the FSD subscriptions by offering a free trial along with reducing the price by 50%.
The quest is for more data.
@taskmaster4450le/re-leothreads-fjsdexrf
We start with the outcome of more data. In this case, we are dealing with miles driven using the FSD software.
It doesn't matter which variable we focus upon. Any change, i.e. improvement, has an impact on the rest.
Improvements to the neural networks leads to shorter update times. This means the software gets better at a faster rate.
As this happens, even if no new vehicles are added to the fleet, the miles will increase as people use it more. Adding in more subscriptions enhances this as does more sales.
We see the result in more data being generated. This leads to Tesla adding in more compute which to run the NN. Here is another component that impacts the amount of data generated.
Start the process anywhere on the circle and the result is the same.
The entire premise is tied to Newton's Law:
An object at rest stays at rest and an object in motion stays in motion with the same speed and in the same direction unless acted upon by an unbalanced force.
The hardest part is getting started. One the process starts, the additions to each component causes an increase in speed on the object. Here is where we are talking about the "flywheel".
When applying this to Leo, the key is to keep the circular nature of things in mind. Chart the different components that Leo has. It can be applied to data, revenues, and interactions. Actually, each of these are the components form a larger flywheel.
This is how digital platforms can become extremely powerful. They do not operate in a linear pattern.
