Engineering professor Sam Raymond has developed a new generative artificial intelligence “teaching kit” with NVIDIA that provides university educators with resources to integrate large language models into their curricula.
The teaching kit explores various aspects of generative AI through hands-on experimentation, Raymond said. The nine modules of the kit include content about large language models, model training and multimodal learning.
“The goal of the teaching kit is to try to get students and faculty who work with the kit to see that these things work really well here, [but] they may be a little bit brittle here,” Raymond said. “You can replicate this on much simpler hardware or whatever it happens to be so that when they finish with the kit they will be able to take on any kind of problem in this field and know at least where to start.”
Raymond’s teaching kit is part of NVIDIA’s DLI Teaching Kit Program, which offers downloadable teaching materials and online courses for university educators. According to the program’s website, the kits are designed in collaboration with professors to bridge academic theory with real-world applications and equip students with computing skills.
Interim Thayer School of Engineering dean Douglas Van Citters wrote in an email statement to The Dartmouth that the school is “proud” of Raymond’s “timely production” of the teaching kit.
“NVIDIA’s collaboration with Dr. Raymond is the perfect example of how our scholars and practitioners who teach can extend Dartmouth’s impact to a global audience,” Van Citters wrote.
Discussions between Raymond and NVIDIA about the generative AI teaching kit began in early 2024, while the agreement to develop the kit was finalized midway through the year, Raymond said. The team working on the kit launched three of the nine modules in August and September 2024 to generate interest and refine logistics. The remaining modules were completed and launched on February 25.
According to Raymond, NVIDIA approaches universities and professors who are “moving in the direction that is in sync with them” to co-develop the teaching kits. Raymond had hosted a few deep learning sessions on campus as an ambassador for the DLI before NVIDIA approached him with the invitation to co-develop the kit.
“We were getting really good feedback from the students and the faculty and staff that attended that my contacts at NVIDIA were curious to go further with this and build this generative AI teaching kit,” Raymond said.
Computer science professor Soroush Vosoughi — who was not involved with developing the kit — said he believes the kit was designed for “democratizing education of generative AI.” He highlighted the kit’s modular design, which features standalone units and “some topics” that build on others, because a single kit “can’t cover everything.”
Although Vosoughi will not implement the kit into his courses, which are tailored to students who already know basic concepts needed to create new models from scratch, he will implement the kit’s resources — including cloud-based computing and NVIDIA computer resources — into his courses.
“I would say that at least in my own teaching, I probably will not be using these materials much because I think at least at the level we are talking about, it's a little bit more surface level than you would like,” Vosoughi said. “But that's why it was designed.”
Raymond stressed the importance of preparing students amid the rapid evolution of generative AI.
“Because generative AI is brand new but making very swift changes to things… it’s the duty of a university or a college to really do two things: to push the frontier of knowledge in general, but also to prepare the next generation,” Raymond said.
