How does Google earn money? Is it ads?
“As a business, you can do a lot of good using software, but you still have to put it into service. As an organization, you must pay for that software.” ~ Alexander Graham Bell
What’s the business model for Google? Google makes money from users. When you use Google, for example, Google is making money, you’re using the Google apps and the Google services, and Google is making money, but you’re not using the apps and services… it’s not obvious. That’s why it’s what we’re going to talk about today.
Let’s say you have two apps:
It’s not obvious that they’re competitors — they both depend on each other for certain services, of course, but neither of them can compete with each other just on the user base they have.
But in which services are they competing with one another? And to what extent, if any?
Well, consider this scenario. you set out to do two different types of algorithms. You have a list of these algorithms and the algorithm that you want to use is built with the Unforgettable algorithm first. Therefore, you choose to execute the algorithm using the Unforgettable algorithm…
Your goal is to re-run the first algorithm — both the Waymo and Unforgettable algorithms will need to be rebooted. The advantage of going back through the Unforgettable algorithm is if the debugging process goes badly you have something to work off… you can just pull the Unforgettable algorithm and reboot the Waymo algorithm and both of these algorithms will run the robot. That’s the real benefit of using the Unforgettable algorithm: you’re using the Unforgettable algorithm as a codebase, which automates the debugging process to make sure that the Unforgettable algorithm works…
The disadvantage of going back through the Unforgettable algorithm is it means you still need to duplicate the data. You’ll need to go through the data for both algorithms again, and the tracking is significantly more expensive… If you needed to do this for both algorithms then you’re spending at least $1.5 million to run the process.
It’s a business model that either requires you to have at least some specific algorithms or you need to re-run the processes and re-start them. That’s not a good business model.
Clustering, deep learning, and convolutional neural networks all require following a similar approach:
Begin with a list of rules
Set up the rules based on the size of the data (e.g. ten thousand examples)
Running the rules one by one, or some combination of these, to see the results
A certain type of machine learning algorithm needs to be used for every or similar questions to be answered
In that business model framework, there’s only one place you want to look for that criteria: the design of the model. The algorithm itself isn’t important. That’s how Google makes money — it makes money from the user and provides them a service.