Professor Pramod Viswanath, an Electrical and Computer Engineering professor, has introduced an innovative Artificial Intelligence teaching assistant named “Blockie” in his advanced engineering class. Unlike some Princeton professors who have restricted the use of AI tools, Viswanath believes in the pedagogical potential of these tools in the classroom.
Blockie is described by Viswanath as “ChatGPT on steroids,” referring to Chat Generative Pre-trained Transformer, a large language model (LLM) released last fall. ChatGPT learns from a vast dataset sourced from the internet and can generate human-like text for various purposes, such as coding, summarizing texts, and solving math problems.
The flexibility of Princeton University regarding AI tools was emphasized in a memo titled “AI & ChatGPT Guidance for Teaching” from January 2023. The memo acknowledges the concerns around ChatGPT’s potential impact on education but expresses confidence in the vitality and usefulness of Princeton’s education programs.
While the use of AI tools is at the discretion of faculty members, they are reminded to adhere to academic integrity rules and collaboration policies under the University’s Honor Code and Academic Regulations.
In response to the memo, Princeton professors shared diverse perspectives on ChatGPT. Matt Weinberg, an assistant computer science professor, described the technology as producing text that “initially seems convincing but is ultimately BS.” Some courses, like COS126: An Interdisciplinary Approach, Princeton’s largest introductory computer science course, restrict the use of AI tools to ensure students learn foundational concepts independently.
In COS126, students are encouraged to think logarithmically, and the use of AI tools is limited to discussing concepts. The collaboration policy explicitly prohibits sharing, viewing, or copying code/solutions from AI chatbots.
Sebastian Caldas, a COS126 lecturer, believes that students should learn basic concepts independently before relying on AI tools. However, he sees potential for students to contribute to improving tools like ChatGPT once they have a solid foundation in programming.
In contrast, David Malan, a professor at Harvard, has embraced AI chatbots in education. He created an AI chatbot similar to ChatGPT that functions as a teaching assistant, offering feedback on student code and answering questions.
Inspired by this approach, Viswanath and his teaching assistant, Tianle Cai, created Blockie for the ECE 470: Principles of Blockchains course. Blockie was fed with lectures and assignments from the course, providing a personalized AI teaching assistant for students.
While ChatGPT is useful for introductory computer science and programming courses, Cai points out that it may not be sophisticated enough for advanced courses like ECE 470, which teaches the Rust programming language. Blockie addresses this limitation, offering a more tailored solution for advanced coursework.
Despite initial skepticism about the limitations of AI tools, students in the ECE 470 course find Blockie valuable for efficiency and productivity. However, they emphasize the importance of having a foundational understanding of programming to assess the accuracy of Blockie’s responses.
Joy Patterson, a student in the course, sees Blockie as an exciting step in integrating AI into the classroom. She believes that as large language models (LLMs) become more powerful, they will play a larger role in computer science classrooms.
The debate around AI tools in education also extends to social sciences at Princeton. Professor Jacob N. Shapiro encourages his students to use ChatGPT in his POL386: Violent Politics course, exploring its capabilities in generating short answers for critique and editing.
Shapiro sees AI tools as an opportunity for students to develop skills in distinguishing coherent and logical responses from nonsense. He emphasizes the importance of citing ChatGPT as a source and views the University’s flexible approach as beneficial for encouraging pedagogical uses of AI.
While acknowledging concerns about academic integrity, Viswanath sees the rise of ChatGPT as an opportunity for educators to reevaluate evaluation methods. He predicts that AI assistants will become a permanent part of academia and envisions every class having an AI assistant in the future.
As one of the first AI teaching assistants at Princeton, Blockie is still evolving. Cai hopes to make Blockie more general, allowing it to be easily integrated into other classes and platforms like Canvas with a single click.
Professor Shapiro views AI as a tool that can be both a danger and a productive resource. He believes the challenge lies in harnessing technology for good purposes and sees exposure to AI tools as an essential part of preparing students for the workforce.
In conclusion, the integration of AI tools like ChatGPT and innovative creations such as Blockie in education sparks debates about their role, limitations, and potential benefits. While concerns about academic integrity persist, educators like Viswanath and Shapiro are optimistic about the pedagogical opportunities AI presents and the role it can play in shaping future classrooms. The ongoing evolution of AI in education reflects a dynamic intersection of technology and academia, offering new possibilities and challenges for students and educators alike.
