Understanding more about the human brain might be key to building artificial intelligence that can use language effectively, according to Amii Fellow and Canada CIFAR AI Chair Alona Fyshe thinks the human brain.
On a recent episode of the AskAI podcast, Fyshe said there have been considerable strides in natural language processing in the past few years -- the branch of AI that deals with teaching machines to use and understand language. And this area is the focus of her work at the University of Alberta, where she is an Assistant Professor in the Departments of Computing Science and Psychology.
I think understanding what the brain is doing could be used to help us build better models.
Alona Fyshe, Amii Fellow & Canada CIFAR AI Chair
Machines can now be trained to use language in ways that are nearly indistinguishable from people the majority of the time. However, machines still stumble over certain parts of speech. Specifically, they have trouble with things like inference and common sense. For example, if you were to ask a friend what they wanted to order for dinner, they might say “I haven’t had pizza in forever.” Although they didn’t directly answer your question, you’d be able to infer they’d like pizza.
Human beings can handle these tasks intuitively. Finding out how might lead to insights that could lead machines to be better conversationalists.
"I think understanding what the brain is doing could be used to help us build better models," she told host Melissa Kargiannakis.
Fyshe says that while the way machines handle information is very different from people, there are some similarities. Much of her research involves using imaging technology to study how the human brain understands language and using that information to improve artificial intelligence.
But it's not just a one-way street. Insights from building computers that can read and speak can also lead us to understand better how our own brains work.
Listen to the full episode of AskAI about how Fyshe's research into natural language processing. She will also give a keynote speech on her work at Amii's AI Week, May 24 - 27.