BERT is Shallow

Tim Niven joins us this week to discuss his work exploring the limits of what BERT can do on certain natural language tasks such as adversarial attacks, compositional learning, and systematic learning.

Guest

Tim Niven: Tim Niven is originally a philosophy graduate, from New Zealand, where he studied with Max Creswell among other philosophers. He moved to Taiwan in 2014 to learn Mandarin and have a very different cultural experience whilst he learned about the exciting successes of machine learning and became fascinated with artificial intelligence. He is currently pursuing a PhD in the computer science department of National Cheng Kung University, in Tainan, the beautiful south of the country. Coming from philosophy, he naturally gravitated toward argumentation mining, which remains an important stream of his research. As he has become increasingly unsatisfied with overly empirical approaches to artificial intelligence, Tim begun to spend significant time studying linguistics and cognitive science as well.