Foundational AI

Foundational AI

Our goal is to develop artificial intelligence capable of learning to accomplish complex tasks, with the amount of supervision being comparable to (or less than) what typical humans require to solve such tasks.

Research topics include

  • Unsupervised learning as evolution of complex systems
  • Diversity in evolutionary learning algorithms
  • Measures of complexity growth in various computational systems


Applications include simulated environments where agents accomplish various goals, and language-oriented applications where learning has to happen on-the-fly without any supervision or annotations, such as smart chatbots that learn while communicating with the users.

Contact Person

Tomas Mikolov