Recent Alumnus Lectures on Advances in AI

News subtitle

Jason Wei ’20 talks about large language models with Machine Learning class.

Image
Image
Jason Wei '20
Jason Wei ’20 is a researcher in artificial intelligence. (Photo courtesy of Jason Wei ’20)
Body

An artificial intelligence researcher who formerly worked at Google Brain this month is teaching about some of the latest developments in the field with a graduate class in quantitative biomedical sciences.

Jason Wei ’20, who is now a researcher at OpenAI, and his former colleague from Google Research, Hyung Won Chung are delivering a total of three lectures between May 18 and May 25 on state-of-the-art large language models that power chatbots and generative AI applications. The AI systems are trained to process and understand human language.

Wei’s first lecture focused on transformers, a deep learning architecture that underpins large language models, his area of research at Google. Transformers identify relationships between words in a sentence, no matter how far distant they are from each other. They revolutionized natural language processing in recent years, making it possible for AI models to understand and generate text, translate with accuracy, and answer questions conversationally.

“My hope for the students is that they understand the motivation behind why language models have been successful and see the trends moving forward. I hope they feel the excitement about the potential for AI to impact society,” says Wei.

Machine Learning, a graduate course taught by Saeed Hassanpour, associate professor of biomedical data science, computer science, and epidemiology, is designed to help students gain hands-on experience in using machine learning techniques to solve real-world complex problems in areas such as the biomedical domain.

“The future of AI in health care is promising and exciting; there are new developments and breakthroughs almost every day,” says Hassanpour, who is keen on introducing students to the latest techniques and novel approaches that are being used to train language models.

Soroush Vosoughi, assistant professor of computer science, has taught Wei and worked with him on research projects when he was an undergraduate at Dartmouth. Wei is enthusiastic about teaching and being connected with the Dartmouth community, says Vosoughi.

“It is very exciting for students to learn from someone who is in the thick of things,” he says.