Language-guided machine learning

Title: Language-guided machine learning

Abstract: In modern machine learning, language has predominantly been treated as just another modality for applying statistical techniques. The question of ‘what can natural language provide for machine learning’ remains relatively underexplored. Humans have used language for centuries as a means of encoding and communicating task and domain knowledge – how can we similarly leverage such knowledge to train machines? In this talk, I will describe some of our recent attempts at language-guided machine learning for the paradigms of both supervised and reinforcement learning (RL). Specifically, I will show how language can be used to provide more semantic supervision to improve generalization and to help RL agents learn efficiently and safely. 

Bio: Karthik Narasimhan is an assistant professor in the Computer Science department at Princeton University. His research spans the areas of natural language processing and reinforcement learning, with the goal of building intelligent agents that learn to operate in the world through both their own experience (”doing things”) and leveraging existing human knowledge (”reading about things”). Karthik received his PhD from MIT in 2017, and spent a year as a visiting research scientist at OpenAI, contributing to the development of the GPT language model, prior to joining Princeton in 2018. He is the recipient of a Google Research Scholar Award (2022), an Amazon research award (2019) and best paper awards/nominations at EMNLP (2015, 2016).

To find out more information about this event or to obtain the Zoom link, please see the event announcements from MLFL on the college email lists or contact wenlongzhao [at] cs.umass.edu (subject: MLFL%20Zoom%20Link) (Wenlong Zhao).