By: Erin Vollick
9 Oct, 2023
Alona Fyshe, Amii: AI that learns the way babies do
October 9, 2023
Alona Fyshe, Canada CIFAR AI Chair, Amii
Humans are far more efficient at learning languages than their AI counterparts, says Canada CIFAR AI Chair Alona Fyshe. The question is, why. “Why are these extremely powerful models still lagging behind when it comes to picking up language?” asks Fyshe — and that line of inquiry lies at the heart of her research.
“The Canada CIFAR AI Chair is a remarkable opportunity because it’s a set of funding that’s very open-ended. It allows me a lot of flexibility to change direction quickly, which, when it comes to AI, is extremely important.”
Alona Fyshe, Canada CIFAR AI Chair, Amii
In one project, in collaboration with researchers at University of British Columbia, Fyshe scans the brains of infants to map how language evolves and becomes nuanced as it is acquired. By better understanding human language models, Fyshe hopes we can learn more about AI language models and how to improve them.
Fyshe’s research also looks at training large AI language models to generate questions, work that has already led to a partnership with education software company Shoelace Learning for an adaptive learning game, Dreamscape.
Fyshe is also a Fellow with the CIFAR Learning in Machines and Brains Program, and a Fellow at Amii, where she develops communications to educate the public on the impact of AI. For Fyshe, this kind of communication is vital — and was the motivation behind her recent TedTalk, which has garnered more than 1.2 million views since its February 2023 release.
“The Canada CIFAR AI Chair is a remarkable opportunity because it’s a set of funding that’s very open-ended. It allows me a lot of flexibility to change direction quickly, which, when it comes to AI, is extremely important.”