Seminar Announcement
Date: Tuesday, 7 January 2025
Time: 10:30 AM
Venue: Seminar Hall
SELP: A framework for implementing the perception-action loop in an AI agent
Bonny Banerjee
University of Memphis and the Institute for Intelligent Systems.
07-01-25

---

Abstract

In this talk, Dr. Bonny Banerjee will present a general-purpose predictive agent that interacts with its environment by relentlessly executing four functions cyclically: Surprise (compute prediction error), Explain (infer causes of surprise), Learn (update internal model using the surprise and inferred causes), and Predict the next observation, called the SELP cycle [1, 12]. To Explain, the agent can act, which includes interaction and communication with its own body (sensed via proprioception) and with its environment and other agents (sensed via perception). The SELP cycle is modular and allows experimentation with different generative models (including large language models) and different fusion methods. For the last 14 years in Dr. Banerjee's lab, AI agents based on the SELP cycle have been evaluated on multiple applications in comparison to the state-of-the-art, such as, handwriting generation and recognition, interaction generation, intent recognition, speech motor skill acquisition, speech emotion recognition, and learning when and with whom to communicate. This talk will include an overview of these applications.