Which group do you think is larger: professional computer programmers or computer users who engage in some programming? It's the latter. There are millions of so-called end-user programmers. Many of these individuals are not pursuing careers as professional programmers or computer scientists; instead, they’re entering fields like business, teaching, law, and various other professions where a little programming boosts their efficiency. The era of programmers being limited to software development companies is over. If you have created formulas in Excel, filtered emails using rules, modified a game, scripted actions in Photoshop, utilized R for data analysis, or automated repetitive tasks, you qualify as an end-user programmer. As educators who instruct programming, we aim to assist students in non-computer science fields in reaching their objectives. However, mastering programming sufficiently to produce complete programs can be daunting to achieve in a single course, given the vast knowledge required about the programming language itself. This is where artificial intelligence can be beneficial. Going Down the Rabbit Hole Many students find that learning the syntax of a programming language, such as proper placement of colons and indentation, consumes a significant amount of time. For those who simply wish to use coding as a tool for problem-solving rather than as a skill to master, this focus on syntax can feel unproductive. Consequently, we believe our current courses have not adequately supported these students. Some students struggle even to write small functions, distinct snippets of code, much less complete programs that could enhance their lives. Tools utilizing large language models, such as GitHub Copilot, have the potential to change these outcomes.

These tools have already transformed how professionals code, and we believe they can similarly assist future end-user programmers in crafting software that is relevant to them. These AI tools typically generate syntactically correct code and can often create small functions based on plain English prompts. By allowing students to manage lower-level programming details with these tools, they are able to concentrate on broader, more conceptual issues related to software development. Many universities are now incorporating Copilot into their programming courses. At the University of California, San Diego, we’ve developed an introductory programming course primarily aimed at non-computer science students that integrates Copilot. In this course, students learn programming with Copilot assisting them through our curriculum. The course emphasizes high-level concepts, such as breaking down large tasks into manageable segments, ensuring code correctness through testing, and troubleshooting faulty code. Empowered to Tackle Problems In this course, we assign students significant, open-ended projects, and the outcomes have exceeded our expectations. For instance, one student in a project focused on finding and analyzing online datasets—a neuroscience major—developed a data visualization tool to showcase how age and other variables impact stroke risk. In another project, students integrated their personal artwork into a collage, applying filters they designed using Python. Such projects push the limits of what we could have previously asked of students before the emergence of large language model AIs. Amid discussions around AI negatively impacting education by doing students' homework or writing essays, it may be surprising for you to hear educators like us discussing its advantages. AI, like any tool created by humans, can be beneficial in certain contexts while potentially unhelpful in others. In our introductory programming course, which primarily targets students who are not computer science majors, we witness firsthand how AI can empower students in valuable ways, with the promise of significantly increasing the number of end-user programmers.