Towards a Stronger Unified Core for AI:  Reasoning and Learning

Benjamin Grosof, PhD
Kyndi, Inc.
http://benjamingrosof.com

IBM T.J. Watson Research Center, Yorktown, NY, USA
March 6, 2019
(hosted by Alex Gray, head of IBM Research AI Science)

AI does not live by machine learning (ML) alone.  
Adding knowledge representation & reasoning (KRR) is critical, including
for natural language understanding and conversation (NLUC).
We discuss some recent approaches and promising directions for deeply combining
KRR with ML.  These include extended forms of 
knowledge graphs and logic programming.  We discuss how to combine high
KRR expressiveness with strong scalability of reasoning.
We also discuss how to deeply and strongly combine ML with KRR, including roles
for probabilistic uncertainty, neural nets, NL-based knowledge acquisition,
and explainability.