Matt Rattley, Lecturer, Department of Biochemistry
In the Department of Biochemistry AI is a useful tool for helping students apply their knowledge to a range of complex biological systems. Leveraging AI's ability to generate related questions and examples, within a framework laid out for it, gives students an almost endless supply of practice with minimal additional work to us as educators.
For a workshop on the type of reaction that's commonplace in biological systems, where molecules become temporarily bound to an enzyme, I created a custom GPT which delivers an example of an enzyme where this happens. It gives enough information for students to work out what's going on, and the student is then tasked with fully outlining the processes involved.
The GPT has been developed to conform to a particular structure with examples of good and bad answers for it to check against. It has been instructed to not offer up complete answers until a student has given them an answer, and can then, if prompted, ask follow-up questions.
While it takes some work to get GPT behaving correctly, this is offset by the work needed to manually build or mark these problems. The tool can also be used by students for revision, providing essentially limitless examples to practice on, giving students more opportunities to hone their skills.
For people new to AI, the best way to find out what it can do is to give it a shot and see how it goes. I've found that it can, very quickly, do a reasonable job of many tasks - and you won't know the ways it fails until you see how it responds to particular prompts. Taking a tool from "reasonable" to "very good" does take some work, and importantly, rigorous testing. Try to break it and then aim to mitigate where it struggles, and casting human eyes over any outputs is always a good idea.