McGinnis, John O. --- "Accelerating AI" [2018] ELECD 1404; in Barfield, Woodrow; Pagallo, Ugo (eds), "Research Handbook on the Law of Artificial Intelligence" (Edward Elgar Publishing, 2018) 40

Book Title: Research Handbook on the Law of Artificial Intelligence

Editor(s): Barfield, Woodrow; Pagallo, Ugo

Publisher: Edward Elgar Publishing

ISBN: 9781786439048

Section: Chapter 2

Section Title: Accelerating AI

Author(s): McGinnis, John O.

Number of pages: 15

Abstract/Description:

This chapter argues for government support for “friendly AI”—a kind of artificial intelligence that will not endanger humans. AI is not, contrary to some theorists, necessarily a threat to humans, because AI will not necessarily be anthropomorphic and possess an all-too-human will to power. Government support for friendly AI makes it more like that friendly AI will emerge before other more dangerous kinds. Moreover, such support is also justified because the acceleration of AI will aid in the analysis of the risks and benefits created by the many other kinds of technology that are also accelerating—from nanotechnology to biotechnology. Only AI of all the many accelerating technologies of our present day helps perform this crucial function. Alternative approaches to AI—relinquishment or regulation of kinds of AI that might be deemed harmful—are infeasible. Given that AI is so central to modern military power, relinquishing or inhibiting AI would empower the worst nations on earth.