If a robot traveling to a destination has just two possible paths, it needs only to compare the routes’ travel time and probability of success. But if the robot is traversing a complex environment with many possible paths, choosing the best route amid so much uncertainty can quickly become an intractable problem.
In 2012, the best language models were small recurrent networks that struggled to form coherent sentences. Fast forward to today, and large language models like GPT-4 outperform most students on the SAT. How has this rapid progress been possible?
To teach an AI agent a new task, like how to open a kitchen cabinet, researchers often use reinforcement learning — a trial-and-error process where the agent is rewarded for taking actions that get it closer to the goal.
It isn’t easy for a robot to find its way out of a maze. Picture these machines trying to traverse a kid’s playroom to reach the kitchen, with miscellaneous toys scattered across the floor and furniture blocking some potential paths. This messy labyrinth requires the robot to calculate the most optimal journey to its destination, without crashing into any obstacles. What is the bot to do?
For more than 60 years, MIT has been an undisputed pioneer in developing computing technologies that have transformed the world. The Institute’s largest research lab, the Computer Science and Artificial Intelligence Laboratory (CSAIL), has had a hand in everything from time-sharing and computer graphics to data encryption and early versions of the Internet.
