Inception

The
Origins

Brown University, 1994. A young woman chasing a lego robot down the 5th floor of CIT and a Master's thesis.

"I like to think of flexible systems — be they carbon-based or silicon-based — that have the potential of evolving into independent thinking beings as Unconventional Intelligences or UIs. A quest for UIs inspires me to play with learning systems and robots. This thesis constitutes my first endeavor towards the quest for UIs."

— Thesis Prologue, Brown University, 1994

Early Days · circa 1992

Backpropagation — C++ code on VAX Mainframes and a Robot.

An immigrant on a scholarship.

Ronny Ashar, circa 1992, working on backpropagation and robotics

The Thesis

"Hierarchical Learning in Stochastic Domains"

Teaching autonomous and adaptive intelligent systems to accomplish dynamically changing goals in stochastic domains. Imagine programming a robot to pick up rock samples from the surface of Mars — adaptive, functioning in complex unknown terrain. Or teaching a child to collect her toys scattered across the carpet.

Not hand-coded rules. Not a babysitter hovering over the machine. Something that learns.

Grid world simulation — Reinforcement Learning at Brown

Reinforcement Learning at Brown

Grid world simulation on Sun Sparc Workstations. Built on Richard Sutton and Andrew Barto's Q Learning work.

Sutton and Barto received the Turing Award in 2024.