On NSF “open questions,” some external properties of the brain as a learning system and an architecture for autonomous learning

Abstract

The 2007 NSF workshop report titled “Future Challenges for the Science and Engineering of Learning” (http://www.cnl.salk.edu/Media/NSFWorkshopReport.v4.pdf) raises lots of questions about how the brain works and learns and they have important implications for the development of autonomous adaptive systems. The report also defines some general characteristics of biological learners that, in essence, impose constraints on any kind of learning systems that we call brain-like. This paper examines these general characteristics of biological learners, as defined in the NSF report, and relates them to a set of properties of brain-like learning defined as early as 1994. The paper also shows how a control theoretic architecture for autonomous learning systems mitigates or resolves many of the “open questions” posed by the NSF report. It also provides some recent evidence from neuroscience on the nature of learning in biological systems that support the notion that the brain has a control theoretic architecture.