Why AI Should Be a Required Course for Every Graduate Program

“Joining hands with AI, management science and operations research can aspire to tackle every kind of problem-solving and decision-making task the human mind confronts.” – Herbert Simon, plenary address at the TIMS/ORSA Joint National Meeting, Miami, Florida, October 1986 [1]

In 1956, a small workshop held at Dartmouth College gave birth to the field of artificial intelligence (AI). Fast-forwarding to the present day, AI is so pervasive that many people forget its dominance was only recently attained.

As recent as 2007, when I began my studies in robotics at Carnegie Mellon University, AI was certainly not a hot subject. At best, AI was just another academic field like philosophy, psychology or economics. Never in a million years did I think that AI would be a staple of any business curriculum.

That is exactly what has recently happened at Johns Hopkins Carey Business School, where I am a professor. From 2021 onward, all full-time MBA students at Johns Hopkins are required to take the AI course as a prerequisite to graduation.

How Did It Happen?

Around 2012, the power of AI in classifying images (e.g., cats vs. dogs) sparked the “deep learning revolution.” By 2018, deep learning had taken center stage in AI, with applications in almost every industry and sector. Academic institutions are no exception. Today, one would be hard-pressed to think of a discipline that has not been shaped by AI.

Since 2013 – the year I graduated from Carnegie Mellon and joined the Johns Hopkins faculty – I have not ceased to be amazed by the rapid advancement of AI. When Johns Hopkins started revising its MBA curriculum in 2018, I jumped at the chance to advocate for AI to be included as a required course. I didn’t expect my proposal to be taken seriously; after all, none of the business schools I knew of required AI. I was surprised to see that my forward-thinking colleagues appeared willing to take a chance on me.

My proposal was accepted and I was privileged to be asked to design and teach the course starting in 2021, in the midst of the COVID-19 pandemic. It was a daunting task, not least because few students had prior computer programming experience and many did not understand the basics of machine learning.

Against all odds, the AI course has been a huge success, becoming one of Johns Hopkins’ top-rated MBA courses. Drawing on my own experience, I’d like to share some tips with those who are hesitant to teach AI.

First, determine the scope of your course based on student interest and backgrounds. I chose to focus on deep learning (not “shallow” machine learning techniques), particularly convolutional neural networks (CNNs), because students can almost always find applications in their work or personal lives. Some of my students are medical or public health students, and they can quickly identify scenarios in which medical images can be translated into concrete insights.

The key concepts of deep learning are also, surprisingly, much easier to grasp when compared with those of “shallow” machine learning techniques. This is because deep learning is built on neural networks, which were inspired by brain science and have very intuitive representations. As an added bonus, the math that governs how different neurons interact with one another is simple to explain to students, even for those who are not familiar with linear algebra.

Second, select an AI development tool with which your students are comfortable. Because few of our MBA students were familiar with the Python or R programming languages, I chose Keras for teaching much of the deep learning. As the most popular AI development tool in the world [2], Keras requires a basic understanding of Python, but nothing more. Overall, my students were quite comfortable reading Keras code in Python and modifying it for their development tasks.

Third, create an environment in which students start developing AI tools in their first class. I begin my AI course by demonstrating a simple image classification task, such as the famous cats vs. dogs problem, using a tiny data set. My demonstration consists of 10 dog and 10 cat images randomly found on the internet. We train a model that can tell whether a new image is of a dog or a cat with surprising accuracy using a simple CNN. Students could appreciate the power of AI through hands-on learning in a matter of minutes. Following class, they are asked to experiment with a different image classification task.

Lastly and perhaps most importantly, when students are challenged to create a novel AI system, they learn the most. The “AI Lab” module of my AI course – arguably the most important segment – is when students work in groups to identify a business or societal setting in which AI can play a role in transforming structured and unstructured data into tools with the potential to generate business and human value. Each student team comes up with a novel idea, builds a high-performing AI system, and develops a concrete plan to scale up the impact of their AI solutions.

As an example, one of our student teams created a “smart tampon” system that helps detect cervical cancer [3]. The “smart tampon” is inserted like a regular tampon, but not during menstruation and only for the duration of data collection. Images of the cervix would be captured by a camera at the top of the “tampon” and sent to an AI model for analysis. The outcome is linked to an app that is accessible to both the patient and their provider. A simple red light would notify the provider of the possibility of an anomaly and prompt the patient (user) to seek medical attention. A green light means that cervical health is normal.

Why It Should Happen

Now to reiterate why AI should be a required course for every graduate program. The answer is evident, according to a McKinsey study [4]: “40% of all the potential value that can be created by analytics today comes from the AI techniques that fall under the umbrella ‘deep learning,’ which could account for between $3.5 trillion and $5.8 trillion in annual value.”

By not offering an AI course, students are missing out on 40% of the value of analytics. Whether a graduate student plans to work in academia or industry after completing their program, a working knowledge of AI is not only desirable but also required.

So, the real question should instead be, why hasn’t AI become required in most graduate programs? The answer, I believe, isn’t because AI is not a necessary skill but because of a widely held myth that teaching AI to non-computer science students is impossible. That myth has been thoroughly debunked in my experience.

Once we have overcome the impossible myth, we will realize that AI can and should be a required course for all graduate students, particularly those in OR/MS programs. And an exciting journey follows.

References

1. Simon, H.A., 1987, “Two Heads Are Better than One: The Collaboration between AI and OR,” Interfaces, Vol. 17, No. 4, pp. 8-15.
2. https://keras.io/
3. “Could Artificial Intelligence in a ‘Smart Tampon’ Detect Cervical Cancer?,” 2022, Johns Hopkins Carey Business School, Sept. 29, https://carey.jhu.edu/articles/artificial-intelligence-smart-tampon-detect-cervical-cancer.
4. Chui, M., Henke, N. and Miremadi, M., 2019, “Most of AI’s Business Uses Will be in Two Areas,” Harvard Business Review, March 7, https://www.mckinsey.com/capabilities/quantumblack/our-insights/most-of-ais-business-uses-will-be-in-two-areas.