Creative Problem Solving

Abstract

Although current state-of-the-art machine learning techniques have led to significant advances in AI systems, they are still far from demonstrating fundamental problem-solving capabilities possessed by human toddlers and even some animals. After decades of research and millennia of scientific and philosophical thought, the central goals of AI -- to explain and replicate human intelligence and creativity -- still remain unmet. I argue for instilling in AI systems the ability to continually "make sense'' of their changing world and challenge their own assumptions, laying the foundation for creative thinking. Future AI systems will face problems their designers did not anticipate and will need to find a way to improvise. In this talk, I will first briefly share my dissertation work in building foundational aspects of cognition needed to be able to do creative problem-solving. I will then discuss a proposed computational model for creative problem solving which requires the agent to actively engage with its environment and form flexible, symbolic mental representations. Symbolic representations allow an agent to reason beyond statistical patterns, verify the veracity of their knowledge, recognize gaps in their understanding, raise questions, construct abstractions and explore the world to seek out answers, possibly even novel ones. Moreover, such representations also allow artificial agents to provide us, humans, explanations of their behaviors, allow us to better interpret and understand their actions, and ensure that they comply with our social and moral norms.

To join Zoom meeting:

https://tufts.zoom.us/j/97054272422

See colloquium email for password.