Models & Mechanisms for Motivating Machines

Author(s): Prithwis Mukerjee

Abstract

We describe a mechanism that may allow digital machines to display intelligent behaviour without the overt intervention of human programmers. While most studies focus on how systems may be built to demonstrate intelligence, our focus is on how to motivate machines to demonstrate intelligence in new areas when they are not explicitly programmed to do so. Domain generalisation is an attempt in this direction but these attempts seek to enlarge capability in existing domains without moving into new domains of expertise. In this paper, we look at a set of existing software constructs, namely (i) the universally used TCP/IP protocol that connects machines, (ii) Docker containers that create portable programs, (iii) the Ramanujan machine that generates novel conjectures about number theory, (iv) blockchain technology that forms the basis of decentralised autonomous organisations (DAO) and (v) generative adversarial networks (GAN) that use pairs of generator-discriminator neural networks to create original content that is computationally indistinguishable from naturally occurring content. We show how these seemingly disparate dots can be connected to reveal a pattern that delineates the contours of a novel mechanism that may be used to define different levels of motivation for machines to demonstrate intelligent behaviour in new areas in an autonomous manner. We also speculate on how this may lead to a political colour in the behaviour that emerges through this process.

The Chartered Data Scientist Designation

Achieve the highest distinction in the data science profession.

Explore more from Association of Data Scientists

Become ADaSci Chapter Lead

As a chapter lead, you will have the opportunity to connect with fellow data professionals in your area, share knowledge and resources, and work together to advance the field of data science.