Title: Programmable Smart Machines

Authors: Jonathan Appavoo, Amos Waterland (Harvard University), and Dan Schatzberg

Date: April 15, 2012

Abstract:

In this paper we conjecture that a system can be constructed that
exploits the general ability to learn through the counting,
correlating, and memorizing of occurrences of events to fast-forward
a programmable computer.  In particular, we propose a signal based
interpretation of a computer's execution that can be used to implement
a form of system state memoization using a predictive associative
memory.  Such an approach may some day lead to a system that can
utilize both traditional logic and neuromorphic or other biologically
inspired mechanisms to be both programmable and smart.