Christopher Collison
I love both AI and Chemistry, and I have a deep passion for uncovering patterns in data, particularly in how theory connects to practical outcomes. During my PhD, I would spend months analyzing vast datasets collected over a week on a shared laser system, but now, thanks to AI, I can derive insights in just an evening. I’m fascinated by how AI can reveal connections that we might overlook and how the algorithms of learning are reflected in AI processes.
1. How do you use AI in your teaching, and what are your favorite resources?
I use an agent-based system with an upbeat 1970s hippy persona, along with retrieval-augmented generation (RAG) from an open-source textbook, to replicate the experience of my office hours. Through this TutorBot, students describe the chemistry problems they’re working on and build up solutions by applying their foundational knowledge. By guiding them to describe these chemical concepts, they better understand the path forward. We’re piloting the TutorBot now, and students love the 24/7 access it provides. My favorite resources include OpenAI's Assistants tool and, of course, our very own TutorBot built around it.
2. Can you share or describe an example or two of an AI-related assignment?
One AI-related assignment I use involves applying an "Explainable AI" package (Andrew White's group) to a dataset of the student’s choice. For this capstone project, students modify the code and provide commentary/pseudocode, including a description of how the chemistry of molecules might be numericized. They must use a surrogate model and counterfactuals to demonstrate that the black-box neural network model’s output is reasonable, all while connecting this back to their foundational chemical knowledge. It’s a beautiful piece of work where students critically evaluate AI’s output based on their understanding.Link to GitHub for the package: https://github.com/ur-whitelab/exmol
3. What do you tell students about using AI?
I encourage my students to recognize that they are adults responsible for their learning, with tools like generative AI to help them along the way. I encourage them to use AI as a tool to enhance their understanding, not as a substitute for it. To ensure accountability, I provide closed-book exams where AI reliance isn’t an option. I also encourage them to use my TutorBot, which we’re currently piloting, to guide their problem-solving and deepen their learning. What challenges, if any, have you had with AI in your courses?
4. What challenges, if any, have you had with AI in your courses?
The main challenge is that we don't have enough time to discuss the role of AI in my Advanced General Chemistry class, given that this class is necessary to build the foundation of chemical knowledge that is so powerful for their subsequent chemistry education. The other challenge with my AI in Chemistry course is to convince Chemistry students that they will be able to succeed in an AI course!
5. How do you think AI has or will impact your domain?
AI is poised to revolutionize chemistry by allowing chemists to explore the nearly infinite chemical space more efficiently. The days of spending countless hours in a "wet-bench" lab may be numbered, with AI enabling chemists to focus on solving larger global challenges like energy production, climate change, and drug discovery. The increased productivity from AI will be key to exploring vast chemical spaces, and by integrating AI with automated synthesis, purification, and testing, chemists can employ high-throughput, AI-driven decision-making systems, or "autonomous agents." This synergy will dramatically change how we design, predict, and synthesize new compounds