Some Thoughts on NETL, 15 Years Later

1. Overview NETL [Fahlman 79] is a knowledge-representation system combining two key ideas: first, the use of a semantic network representation, with built-in capabilities for inheritance, inference, and simple search; second, the use of massively parallel hardware to perform most of these built-in operations in near-constant time, without the need for complex, hand-crafted indexes or application-specific search procedures. NETL was developed as my Ph.D. thesis project in the period from 1974 to 1977, and was published in book form in 1979. In the years since then, the NETL ideas have been influential in a number of ways, though NETL has never really become a part of the mainstream in AI. In this short paper I will present my own personal view of the evolution of NETL, its strengths and weaknesses, and the lessons it might still hold for AI.