Adam Gao (Assistant Professor, Jarvis College of Computing and Digital Media) presented “Building Your Own Customized GPT Teaching Assistant for ‘Free’” at the AI in Teaching Symposium, October 18, 2024.

Professor Gao highlights the initial apprehension surrounding AI in education, acknowledging concerns about potential misuse for cheating. He argues, however, that the same technology can be harnessed to create powerful tools for supporting student learning.

Gao’s Vita, a virtual teaching assistant for his Deep Learning course, serves as a compelling example. Vita can answer a wide range of questions, covering both logistical and academic aspects of the course.

Key features of Vita:

• Course Logistics: Vita accurately responds to queries regarding exam formats, deadlines, and project guidelines, adhering to information provided in announcements and the syllabus.
• Academic Support: Vita goes beyond simple information retrieval. It can explain complex topics, summarize lecture content, and even provide guidance for debugging programming assignments.
• Maintaining Academic Integrity: Vita is programmed to identify and refuse to answer questions taken directly from quizzes or those seeking direct answers to assessment questions.
• Scope Sensitivity: Vita can determine if a question falls outside the scope of the course, such as inquiries about unrelated topics like presidential elections.

Building Your Own AI Teaching Assistant:

Gao emphasizes the accessibility of creating a personalized AI teaching assistant using platforms like ChatGPT. He outlines a simple process:

1. Naming and Description: Assign a name and provide a clear description outlining the role and purpose of your AI assistant.
2. Instructions: Develop detailed instructions to guide the AI’s responses and ensure alignment with your teaching approach and course policies.
3. Knowledge Base: Provide the AI with relevant course materials, including lecture slides, syllabus, assignments, and any other resources you want it to reference.

Gao does offer a cautionary note: crafting effective instructions is crucial for optimal performance. He highlights the need for precise instructions, such as prompting the AI to check the current date to ensure accuracy in providing time-sensitive information.

Gao encourages faculty to experiment with building their own AI teaching assistants, providing his contact information for further discussion and support. This is a valuable opportunity to explore the transformative potential of AI in enhancing the educational experience for both faculty and students at DePaul University.