George Konidaris

Postdoctoral Associate
Learning and Intelligent Systems Group
MIT Computer Science and Artificial Intelligence Laboratory

gdk at csail dot mit dot edu

Hello there

Welcome to my home page. I'm a postdoc at the Learning and Intelligent Systems group at MIT CSAIL, working with Professors Leslie Pack Kaelbling and Tomas Lozano-Perez. I recently completed my PhD at the Autonomous Learning Laboratory at the University of Massachusetts Amherst under the supervision of Professor Andy Barto.

My research focuses on building intelligent, autonomous agents that display open-ended learning and broad competence. To this end I work on methods for creating agents that can acquire general purpose skill hierarchies through interaction with their environments, and thereby become able to solve harder and harder control problems over time. More details are available on my research, publications, and thesis pages.

Recent News:

• I am very excited to be collaborating with the newly-formed AIML Group at the University of the Witwatersrand.
  
  
• I am co-organizing the AAAI Spring Symposium on Designing Intelligent Robots: Reintegrating AI at Stanford University, March 26th-28th 2012.
  
  
• Ulrich Paquet, Dylan Shell and I recently launched za-cs.net: the South African Computer Science network as a way to bring together South African CS researchers, wherever they happen to currently live. The site includes a forum and a wiki-fied version of my old guide for South African students interested in going overseas for a postgraduate degree in Computer Science.
  
  
• Scott Kuindersma and I won Best Student Video at the 2011 AAAI Video Competition for our video Autonomous Robot Skill Acquisition.
  
  
• I have completed my thesis (!), graduated (!!), and moved to the Learning and Intelligent Systems Group at MIT CSAIL as a postdoc, as of January 2011.
  
  
• I took some photos of the talks and attendees at the Multidisciplinary Symposium on Reinforcement Learning that was held after ICML in June 2009 in Montreal, Canada.
  
  
• Özgür Şimşek and I organized a workshop on abstraction in reinforcement learning at ICML/UAI/COLT 2009 to encourage research that combines both state and temporal abstractions (skills).