Jan 16
CS Bits & Bytes: Suhas Arehalli presents "How Human is Super-human? Exploring Language Processing in Minds and Machines"
Suhas Arehalli is an Assistant Professor of Computer Science at Macalester College. His interests are at the intersection of Sentence Processing, Computational Linguistics, and Theoretical Syntax.
Abstract: Large Language Models (LLMs) like ChatGPT have received an extraordinary amount of attention in recent times, with their impressive command of language causing many to describe these models as "super-human” AI. On the other hand, some computer scientists and linguists challenge have protested this description, pointing out that these models do little more than next-word prediction, no more advanced than autocomplete. In this talk, I'll explore the tension between these two perspectives, comparing modern theories of human language processing with the mechanisms underlying modern LLMs. I'll highlight contributions to this conversation from my own research program, which investigates whether the algorithms LLMs learn to process language share similarities with those used by human brains. I'll conclude by discussing the need for interdisciplinarity in language technologies research, and argue for why collaboration between language scientists and language technology researchers can further our understanding of both human language and the limits of current models.
Join us for treats, beverages, community, and conversation.
from Computer Science