“A simple way to understand machine learning it is to think that with traditional programming we provide data and write a set of rules to get an answer. With machine learning, you provide the data and the answer, and find ways for your machine to generate the underlying rules, making for a wider, more flexible range of applications,” explains Charles Lee, founder and CEO of CoderSchool.

From translation technology, to pharmaceutics, architecture, computer vision or online shopping, machine learning, a sub-field of artificial intelligence research, recently attracted worldwide attention as well as massive investments, catalyzing an ongoing revolution in data science and computer technology. Fulbright students were introduced to this groundbreaking field by participating in an 8-week intensive “Introduction to Machine Learning” with CoderSchool, one of the foremost programming schools in the country.

Vietnam is still catching up to recent advances in the field, but for Charles Lee, there is a good chance for the country to “leapfrog above some of its western counterparts.” Initiatives have begun at every scale from individuals to big companies.  ChoTot, a classified ads platform and former CoderSchool alumni uses computer vision to parse photographs, sorting car models and flagging fake classifieds. Meanwhile, VinGroup, the largest business conglomerate in the country, recently established their very own artificial intelligence laboratory in Hanoi.

 Machine learning is not just the latest tech trend. It represents the most recent advance in data analysis technology, a field that will only become more relevant in the future, as more and more jobs become related to data. Some say 90% of all data was created in the last two years. Even a small school like ours gathers huge amounts of information on the number of students, attendance, arrival times, homework and more. The challenge is how to make us of it. It’s crazy exponential growth, and it is already spurring a variety of new careers, from data analysts to AI, business intelligence, business analysts, data engineers, and many more.

Lifelong appreciation for AI

Students who come to CoderSchool usually undergo training so they can become machine learning professionals, a different profile to students pursuing a liberal arts college education. The programming school therefore had to adapt its course for our young Fulbrighters. Their approach was simple: CoderSchool’s introductory course aims to inspire a lifelong appreciation for artificial intelligence.

The course builds a strong theoretical foundation, diving deep into the mathematical concepts underpinning machine learning. Students were provided with solid foundations, ranging from linear regression, to logistic regression, complex algorithms and convolutional neural networks, all important aspects of building predictive models and computing tools capable of extrapolating rules. For Minh Do, Head of Academics at CoderSchool and main educator for the course, “although the programming language can change in 5 years, and the AI models will change with advances in the field, the math is here to stay, whether students want to go on to become professionals that utilize it, or just to pull the curtain and understand this relatively intimidating piece of technology.” 

Charles Lee (L) and Minh Do

Despite its strong theoretical axis, the course also emphasises experiential learning. In this, CoderSchool’s educational values mirror Fulbright’s own: opening a space for students to explore is the fastest way to grow and learn. Each lesson was designed to offer students the opportunity to dive deep into core programming languages, such as Unix commands and Python, directly connecting theory to practice through self-directed projects.

Students therefore imagined, coded then implemented their own tools, from a predictive recommendation software for clothes shopping, to a tool that sorts pictures of cats from pictures of dogs, or an app that recognizes handwritten numbers. “30% of the class time is for lab sessions, where we implemented real models and built functional apps. Those are based on the theories introduced in class and ready for production use. I really love to learn by creating and so this was perfect for me,” reflects Khoi after participating in the course. Each of these projects requires a good command of both mathematical theory and industry standard coding languages. More fundamentally, it requires a true understanding to fully utilize the possibilities offered by the technology.

An app made by Fulbright students that identify cat and dog photos

Building models from scratch was also Nguyen Cao Nghi’s favorite aspect of the course. “As a tool, machine learning has never been more accessible: there are tons of ready-made models whereby anyone with some coding experience can solve a machine learning problem in just a couple of hours. But building them gave me deeper insight into machine learning as a discipline. From designing models based on mathematical theories to implementing its nuts and bolts, I could better appreciate the brilliant minds behind the field’s advances as well as understand its strengths and limitations.” 

Multifaceted growth

As an intensive course, the class involved advanced mathematics and programming. This means that ideally, students would have already achieved a certain competence in both, prior to taking the class. But as students from a liberal arts college, Fulbrighters who chose to participate did not all have a similar approach or prior understanding. This plurality of backgrounds, rather than a detriment, served to demonstrate the power of students who take ownership of their learning. “Some of the students were already focused on their own projects and knew exactly what they wanted. Others asked me a lot of questions, all the time. And that is what I really found impressive: I knew these students were starting from zero and had absolutely no idea about this stuff before taking the course. But they try hard and they are dedicated to overcoming the challenge. They have a strong idea of their goal,” reflects Minh Do.

But why would students not specializing in computer science choose to attend a machine learning course? Although it might seem at first glance a very specialized subject, such a course cultivates many competencies in students. Most visibly, participants come out of the course with the ability to build a real application to machine learning as their final project. But the experience was formative in other aspects. For Charles Lee, computer science exposure cultivates a unique way of thinking, broken down in a step by step, logical way. “Whether you become an author or an artist, this kind of exposure really helps establish systems of thought that are rigorous, goal-oriented, and constructive.” 

From CoderSchool’s experience, a major barrier to learning about AI is the fear of the unknown, thinking it is too complex, reserved to “geniuses”. Thus, another important goal was to demystify “Artificial Intelligence”. CoderSchool educators expect all participants to be more confident in their ability by outlining basic principles, and step by gradual step open the field wider for exploration.  After participating in the course, Chi reiterates that “Machine learning is for everyone; there are a lot of ways to apply it and make our life easier, and it is not too hard for people to learn and use it in their career.”

 Fulbright students continue to find inspirations to fuel their own growth, taking full advantage of their opportunities. Some of our students are already making plans to incorporate what they learned to their own interests. Nguyen Cao Nghi explains his interest in fields tangential to machine learning, saying that “aside from real-world applications, this course has fed many of my personal interrogations, such as how intelligence emerges from complexity, or how the scientific definition of knowledge will change in accordance with the ubiquity of machine learning and AI.”

Our university always strives to prepare them for an uncertain future, to always stay adaptable so they may forge their own path. Data, on the contrary, can help make this world more predictable instead. As Charles concluded: “One of the first assignments we worked on in the course was actually a coronavirus prediction model. The data today shows us what most experts were telling us a few months ago. As unpredictable as events such as the coronavirus crisis seem to be, I would make the argument that what we are going through is in fact not that unpredictable. If we understand and can trust data, if we know how to build these rules and more sophisticated models, we can not only prepare for an unpredictable future, but also make it somewhat more predictable.”

Antoine Touch