Harvard Will Teach Students Using an AI Instructor Next Semester

Getting what you paid for?

Hey TeAIch

Starting this fall, a popular intro-level coding course at Harvard University, CS50, will be taught by an AI instructor. You know, because Harvard probably can't afford to pay human teachers.

"Our own hope is that, through AI, we can eventually approximate a 1:1 teacher:student ratio for every student in CS50, as by providing them with software-based tools that, 24/7, can support their learning at a pace and in a style that works best for them individually," CS50 professor David Malan told the university's student newspaper, The Harvard Crimson.

Per the Crimson, Malan further clarified to the paper that course staff are "currently experimenting with both GPT 3.5 and GPT 4 models," and said the course has always incorporated new software, meaning that employing an AI teacher is just an "evolution of that tradition."

All fine and good, and a healthy student-teacher ratio is certainly important. But considering that neither GPT-3.5 nor GPT-4 is terribly reliable at spitting out structurally-sound code each and every time, the decision to provide students with an AI-powered coding prof is more than experimental — and paying students, it seems, will be the experimentees.

Bit Iffy

As the Crimson points out, CS50 is notably one of the most popular classes featured on edX, an online learning platform that was built in collaboration between MIT and Harvard and sold for a cool $800 million last year.

Speaking to the student paper, Malan did admit that "early incarnations" of AI programs are likely to "occasionally underperform or even err." Still, he reiterated his hope that the AI will help alleviate some course staff's busy work, thus freeing up time to interact with students one-on-one.

“[A]ssessing, more qualitatively, the design of students' code has remained human-intensive," Malan told the paper. "Through AI, we hope to reduce that time spent, so as to reallocate [teaching fellows'] time toward more meaningful, interpersonal time with their students, akin to an apprenticeship model."

Again, though, we can't stress enough how experimental this is, and while a pinch of healthy skepticism rarely hurts, it's not every day that you hear a professor basically tell you to take everything a different instructor says with a grain of salt.

"We'll make clear to students that they should always think critically when taking in information as input," Malan told the Crimson, "be it from humans or software."

More on AI and education: Bill Gates Says AI Will Be Teaching Kids to Read in 18 Months