Free Copilot for verified educators

GitHub has extended its free tier for educators: teachers verified on GitHub Global Campus now get GitHub Copilot at no cost. The move builds on GitHub's existing education benefits, which already included free access to Codespaces for teachers. Verification is handled through GitHub Global Campus; educators unsure whether they're verified can check their status there.

GitHub Copilot is an AI pair programmer powered by OpenAI Codex that suggests code and whole functions in real time inside the editor. It became generally available for individual developers in June 2022, after a year-long technical preview involving more than 1.2 million developers. GitHub's position is that the tool frees developers to spend more time on larger problems and better software.

What Copilot means for your classroom

GitHub acknowledges that folding Copilot into teaching is not straightforward and says it is working with its teacher community to understand the implications for curricula. So far, research points to several ways Copilot intersects with education.

It performs well on intro assignments

Academic studies show Copilot is capable of solving many introductory programming tasks. One investigation found that its solutions placed it among the top 25% of students in a cohort. Paul Denny, associate professor at the University of Auckland and coauthor of that work, notes that the technology arrived quickly and will strain existing teaching methods, but also opens avenues for new course design and teaching styles.

This technology should be of great interest to all computing educators. What we are dealing with is a freely-available program that can take casually defined English language problem specifications, much like typical exam questions, and return often-correct, well-structured code that could pass as human-written.

- Finnie-Ansley et. al, The Robots Are Coming: Exploring the Implications of OpenAI Codex on Introductory Programming, https://doi.org/10.1145/3511861.3511863

Testing Copilot on a classic "Rainfall problem" yields a plausible, but not flawless, solution. That is expected, and results may vary as the model changes over time. GitHub encourages teachers to try their own assignments against Copilot to gauge its performance.

It can explain code

Denny and colleagues at Aalto University have also used the model to generate explanations of code examples for students. According to Denny, many of these explanations—from line-by-line walkthroughs to high-level summaries—are accurate, often textbook-like, and students find them helpful. In tests, Copilot solved progressively harder problems and finished by explaining its own solution. Given a modular exponentiation task, it described the code, the underlying principle, and the time complexity.

It can draft assignments

The same research group demonstrated that Copilot can create new programming assignments. GitHub tried this approach and got usable results, though not every example works as well. The point, GitHub argues, is that such tools could help students generate practice material and sharpen their ability to evaluate the quality of what's produced.

Joining the effort

Verified teachers can share feedback and ideas in the GitHub Global Campus teachers' discussion. Those not yet verified can apply to join GitHub Global Campus for teachers. GitHub credits the authors of the referenced studies, including James Finnie-Ansley, Paul Denny, Brett Becker, Andrew Luxton-Reilly, James Prather, Sami Sarsa, Arto Hellas, and Juho Leinonen.