Bayesian Program Learning and Concept Induction
Abstract
A long-standing dream in computing has been to build machines that learn like a child Turing (1950) that grow into all the kinds and forms of knowledge that human adults do, starting from much less. At a minimum, any such learning system must be able to acquire many different kinds of expertise. Every child becomes an expert in natural language, motor control, intuitive physics and social interaction, and many will grow into adults with specialized expertise in cooking, calculus, tennis, drawing pictures, or writing software. Despite great advances, artificial intelligence (AI) is still far from acquiring human-like expertise in any of these domains let alone all of them, as a person can.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 21, 2019
- Accession Number
- AD1096503
Entities
People
- Joshua B. Tenenbaum
Organizations
- Massachusetts Institute of Technology