In this guide I'm going to walk through a few real-world examples of how machine learning can be used. And the reason I like adding this guide into our course is that I think it's really important to see how wide ranging machine learning has really become.

The first and probably the most well-known way of us using machine learning is for the spam filter. There's a few different ways of doing it, but generally they work by relating keywords or phrases to spam and non-spam emails. So then when a new email comes in, the algorithm calculates the probability of that email being spam or not spam. And typically if an email has words like ship in, available, fingertips or online, the likelihood of that email being filtered goes way up. And later on in this course we're actually going to spend an entire guide building our very own spam filter.

The next example I'd like to talk about is something I think that we're all pretty familiar with and that's the recommendation engine. I'm not exactly sure on the year, but I believe it was around 2014 when Amazon fully committed to integrating machine learning into their eCommerce platform. And how they did it was by adding in the ability for them to track and record the purchase history for all of their customers. From that, they actually had the purchase history dictate which items should be recommended to their customers. Then as the algorithms got better, they were able to dramatically increase their revenue. I'm not sure if it's still the case, but at one point over a third of Amazon's revenue was generated through the recommendation engine alone.

To stay with the same theme, I thought we should talk about targeted advertising and how machine learning leverages consumer behavior to help with marketing strategies. To improve efficiency, a lot of companies have started to use digital platforms to manage all of their data. And the platforms work by creating user profiles for all of their customers and then each profile is personalized through data collection by way of cookies or purchased from a second party. And once enough information is gathered, the behavioral data is applied to a series of algorithms that place similar profiles into target groups. Then based on group demographics, companies can utilize predictive modeling to actually generate ads that align with the specific consumer preference.

The next example we'll go through is machine learning in sports. And I personally think this is a great example because sports are driven by stats. So right now in baseball a lot of teams are using machine learning algorithms to determine player value. And then with that information, it dictates which players they should sign, how much money they should sign them for and who they should draft. Then other algorithms are being used for pitch prediction, defensive alignments and maybe one day, an automated strike zone.

Basketball is kind of similar, but it's a little bit more complex because it's more of a team game. So what a few teams have done is develop algorithms that actually calculate shot value for every potential shot in their offense and then throughout the course of the game or over the course of a year, they make modifications in their play calling to maximize the number of times the high-valued shots are available. And like baseball, the trades, draft picks and free agents also focus on optimizing the algorithm.

All of this isn't limited to real sports either. In fact, most of the high level fantasy experts have been using their own machine learning algorithms for years.

Even though this one isn't fully operational yet, it's easily one of the coolest applications for machine learning and that's the self-driving car. As you could imagine, this is an insanely complicated project and the role of machine learning is easily the most important. The way I'm going to explain this is obviously way over-simplified, but basically it works by having the car constantly use different sensory inputs like cameras, radar or lasers to check for changes in the environment. Then based on all of that sensory information, a bunch of algorithms work together to generate their own digital map that can predict what objects are and how they should respond to them.

This last example is a little bit creepy, but it also has the potential to be really fun. So there's been an ongoing idea of how we can use facial recognition in the gaming industry and what's great about this idea is that a lot of the algorithms that we would need are already developed for image classification in the self-driving car. So the basic idea is that we can use changes in facial landmarks to actually measure a player's engagement and machine learning algorithms would be used to recognize all of the emotional variations for that player. Then over time your console would actually learn what combinations of sensory stimuli elicits your strongest emotional response and from that, it gives the system the ability to promote or inhibit your emotions based on your desired experience. So if you love to be scared, one day your console will know exactly what it needs to do to absolutely freak you out.

Now these are obviously just a couple examples, but there's really an endless number of ways that we can use machine learning. As we move through this course, I would really like you to keep this guide in mind because if there's something you're really passionate about, there's a good chance machine learning can offer a solution.