The benefits

Janapa Reddi believes that faculty should assume students have access to these utilities and recognize that, for many students, familiarity with these platforms will be a requirement for future workplace success. For example, ChatGPT is capable of producing code needed to design functional computer processing chips; students entering engineering fields will likely use these tools in the workplace because of their code-writing speed. Faculty can be at the forefront of training students to responsibly engage with these new tools. 

“It’s critical that we get students to understand what Generative AI is and how it’s important...The tech has evolved to cut through a range of courses. It’s no longer, ‘I want to study machine learning.’ Now, it’s a tool in the box across all engineering disciplines.”

The challenges

Training in generative AI is not without its risks, and when engineering students lack the foundational knowledge of how various code is created and the ability to understand “what is going on under the hood,” Janapa Reddi warns that there could be catastrophic consequences. Thus, it is crucial for faculty in STEM fields to reformulate their courses to train students in how to use these tools effectively and to know their limitations. 

Takeaways and best practices

• Reinforce foundational training.
  Faculty may benefit from familiarizing themselves with both the strengths and weaknesses of generative AI tools that their students may be utilizing. Understanding where there are gaps in the technology’s capabilities will help faculty to understand what critical foundational skills and knowledge students need to acquire in order to critically assess the output of a tool like ChatGPT, whether they are asked to do so as part of a Harvard course or in a future professional assignment when the stakes are high.
• Take your time with assessments.
  Given generative AI’s evolving capabilities, faculty should take their time to research how the platform handles their current assessments and be intentional about redesigning questions and assessment formats to ensure they are resistant to generative AI platforms. In his own courses, Janapa Reddi took a step back when ChatGPT became available last year, returning to in-class assessments rather than open-note take-home exams. When he tested his exam questions, ChatGPT generated correct answers!
• Consider how to leverage AI platforms.
  Despite the challenges ChatGPT poses for assessment purposes, faculty may want to think creatively about how students might be using generative AI platforms beyond their time at Harvard and incorporate activities that highlight the potential of those tools.  For instance, he is considering if students could use chatGPT to design a microprocessor as a capstone project after they have learned all the fundamentals of microprocessor design and “Verilog” programming.

Bottom line

Generative artificial intelligence will surely have an impact on education for years to come. Instructors have a unique opportunity now to learn about this emerging technology, prepare students to leverage it in meaningful and forward-thinking ways, and facilitate creative experimentation with its use in educational settings.