Understanding and Meaning in Language Models: A Conversation with Computer Scientist Ellie Pavlick

Wednesday - May 1 2024, 05:34 UTC - 9 months ago

Computer scientist Ellie Pavlick is researching how language models understand and use language through concepts and the idea of 'grounding'. With a background in economics and music, she values complex and unsexy results in her field, and views language as encompassing everything.

In the world of artificial intelligence, language models have been a popular and fascinating area of research. But as these models become bigger and more complex, questions about their understanding and meaning arise. In this article, we will delve into the work of computer scientist Ellie Pavlick, who is examining the evidence of understanding within large language models. As a researcher at both Brown University and Google DeepMind, Pavlick's approach is characterized by a mix of precision and nuance, reflecting her background as a musician and self-described 'outsider' in the field of computer science .

We will explore her journey, her findings, and her thoughts on the concept of 'meaning' in the context of language models.