RE Log - Fall 2023

22 Ransom Everglades LOG FALL 2023 factors: RE’s location on Biscayne Bay, the access to the water facilitated by RE Dockmaster Dave Sanderson and also, of course, the fact that RE has two card- carrying shark scientists on its faculty. “RE is just in such a unique position to offer this type of course,” said Marshall. “I think it’s a chance for our students to delve a little bit deeper in a way they haven’t had access to before.” Predicting the future, one algorithm at a time How does Netflix know that you’re probably going to watch a gratuitous true-crime documentary? How does your iPhone know that you took a photo of Aunt Cecilia? How exactly did ChatGPT figure out how to write a song about cake in the style of Rufus Wainwright? The answer is machine learning – a form of computation that essentially runs the world these days, for better or worse. Four years ago, STEM teacher Luis Felipe created the Applied Data Science course at RE, a yearlong STEM elective in which students use programming tools to sort, parse and make sense of large datasets. Advanced Machine Learning is the new follow-up, now in its second year, for students who have already taken that course: an intense, project-based foray into the predictive powers of modern algorithms. “Predictive” is the key word. In his other role as RE’s new Institutional Research Coordinator, responsible for analyzing a huge variety of datasets for the school, Felipe doesn’t actually use machine learning at all. Data science involves parsing the data we already have. Machine learning involves training an algorithm to predict new data points – “unknown” data points that nevertheless follow recognizable patterns. The course starts with foundational concepts and some classic machine learning problems. One example: How might we use a machine learning algorithm to predict if someone survived the Titanic , based on data points like class, gender and age? Students then pursue projects in any field that strikes their interest. Last year, students created algorithms that could predict heart disease, create “more comprehensive college rankings,” and even lend some order to the chaos of the stock market. Elliott Gross ’24 created an algorithm that can predict the stability of super-heavy elements – an impactful project at the intersection of machine learning and physics that he is planning to present at conferences. No more quizzes and tests: The projects are big, open-ended, and entirely student- driven. “I teach you the tools, you pick the area where you want to research,” Felipe said. “I think that it’s probably more similar to an English course or history course than to a physics course in that regard. I can’t say to them, ‘Hey, you have a week for this.’ These are complex projects.” The course demands that they think beyond the boundaries of the classroom. Five percent of the grade asks them to put their projects into a portfolio that they can use for future opportunities: jobs, fellowships, competitive internships, college projects. Students in Advanced Machine Learning get another five percent for “extracurricular engagement”: demonstrating that they’ve used the tools elsewhere. “You want an A+ in this class? You will not get it without applying machine learning to another class that you are taking, or collaborating with a student, or using it in an entrepreneurial competition or your Bowden Fellowship,” Felipe said. “I want you to grow in the direction that you want to.” Telling stories in digital worlds It all starts with a cube: a six-sided cube with intricate designs on each side. Look at it through a phone camera, though, and each side turns into something else: a 3D model vignette, crafted to represent a scene or a moment from a person’s life. As you spin the cube around and move from one vignette to another, each side becomes a piece of the larger story