The Laboratory for Interactive Artificial Intelligence
The Laboratory for Interactive Artificial Intelligence is dedicated to the idea that a key component of human-level intelligence is the drive to be interactive; that is, intelligent agents seek out opportunities to interact and communicate with other intelligent agents. Further, this drive supports intelligence and encourages learning.

Thus, moving forward, many of the interesting and fundamental research issues in Artificial Intelligence lie in domains where the purpose of an intelligent and adaptive system is to interact with humans on their own terms: humans are not only in the loop, they are the loop. The IAI group is interested in building real and robust systems that demonstrate the usefulness of this way of thinking, and in developing fundamental algorithms and technology that support such systems. We are concerned with intelligent systems that are:

  • Immersive
systems that aren't just agents but a natural part of the environment, or are even the environment itself.
  • Expressive
systems that have a character and personality, emotion, and perhaps even a life beyond the user
  • Socially-aware
systems that care about the social relationships between others
  • Knowledge-driven
systems that exploit domain knowledge without being trapped by domain assumptions
  • Grounded
systems that build models and representations that are driven by actual data
  • Adaptive
systems that change their behavior over time based upon feedback from the environment
  • Robust
systems that deal with novel situations gracefully
  • Interactive
systems that seek out other intelligences and communicate with them: the dialogue is the fundamental data structure

Much of our focus over the last few years centers on modeling human beings and their interactions using statistical machine learning techniques to perform activity discovery, modeling and recognition.

A natural consequence of our domain is the need to ensure scalability. As such, a second focus of this group is to find both engineering and algorithmic ways to provide scalable machine learning solutions.

There are lots of folks who care about this, including pfunk, my research group.