Rules for AI, privacy, convenience and transparency. In this course we have a few main objectives. The first is to define privacy in the context of machine learning. What does privacy really mean in the 21st century? We will explore and look in depth. Our second goal is to demonstrate privacy preserving methods in predictive modeling. What can we do to truly protect datasets from the increasingly sophisticated attacks that we see on them? And then third, we are going to analyze the concepts involved in creating transparent, explainable AI. This goes toward our goal of building ethical machine learning models, as we've covered in other courses in this specialization. So, first why machine learning privacy right now? Well, the trend that we're seeing is that big data companies require massive datasets. It's just the nature of machine learning, and predictive modeling. The more accurate the model, the more data it requires and that leads to a world in which companies scoop up and save as much data as possible, and this can lead to real privacy violations. We see that in 2016, 44 billion gigabytes, a day of data was created in already large number, and it's predicted that by 2025 that number is going to balloon to 463 billion gigabytes per day. And as more data points exist for all individuals, were going to start to see more calm caused by privacy violations. More data, more privacy violations. So our opportunity here is to work on secure anonymized models, were essentially fighting two separate battles in this case that we will explore in this course. More models are vacuuming up data in their training sets, so our opportunity is to make that data as secure as possible. Second, as more data sets get linked to individuals, and those datasets get shared as open source models, each piece of private data becomes more exploitable. And we need to make sure that that does not harm those individuals involved. So our opportunity here is to secure public data in different silos, to prevent people from being harmed by these models. Our course will be structured in the following way. We're going to have a video lesson or two, and then or reading or supplemental material. And then a practice quiz, and that knowledge check is going to make sure that you know and understand all of the things that we've covered in the lessons. And then finally, after each week we will have a graded quiz, and that is just to make sure before moving on that you have mastered all of the concepts above. And finally, will there be any coding required in this course. We know that in the world of machine learning there are many different options for coding languages. So we're going to steer clear of all of them. We know that you may know some Python or R, or Lisp or JavaScript, but to keep this course as accessible for all, we are going to work with something called pseudocode. Which is essentially in explainable model of code, that can be taught to anyone and then put into the individual syntax later. So an example of pseudocode is create variable X, input set X = 10, if X > 7, print yes, output yes. So we're going to work with this model whenever we talk about anything related to code, and that is standard throughout this specialization as well. All right, that will cover it and will see you in the course.
