This paper examines John Koza's "invention machine," a computer system that uses genetic programming to autonomously develop solutions to complex problems. Drawing on Jonathon Keats's Popular Science article, the paper traces the historical anxiety surrounding automation — from the Industrial Revolution to modern artificial intelligence — before explaining how Koza's system mimics Darwinian natural selection by generating, evaluating, and recombining candidate solutions until an optimal result emerges. The paper also distinguishes Koza's open-ended genetic programming from earlier, more constrained genetic algorithms, and concludes by assessing the machine's current limitations and its broader implications for human creativity and invention.
The Industrial Revolution had profound effects on the human psyche and civilization as a whole. Machines began to do many of the things that human artisans and craftspeople used to do, and they did them faster and, arguably, better. Along with this advent of machinery came concerns and philosophical reflections about humanity's place in the world. Machines seemed to be taking over, rendering many human workers obsolete. The jobs that did exist changed in many cases from creative and engaging work to running and maintaining these machines, almost seeming to make humans subservient to the very devices they had created. Many felt that lower-class workers were becoming nothing more than components of the machines themselves — meaningless as individuals and useful only for how consistently and quickly they could accomplish repetitive tasks over long stretches of time without wearing out.
Just as the effects of the Industrial Revolution called humanity's purpose and place in the world into question, the invention and implementation of computers did the same thing, but on an even more profound level. Computers are machines that generally do not perform the physical tasks humans once did, but they do — or at least appear to do — something even more innately human: they think. As computers developed further, concerns grew over which new class of human workers would be replaced by these remarkable machines. The idea of artificial intelligence was seen as a frightening concept by many; it seemed to suggest that the machines we had built could eventually begin to make decisions and plans for themselves. Many science fiction writers, social commentators, and intellectuals believed that such a development was not entirely impossible. As computers grew ever more advanced and began to be incorporated into other machines and entered new fields, these fears only grew.
In his article "John Koza Has Built an Invention Machine," Jonathon Keats details how one man developed and built a computer with the kind of creative intelligence that has long been feared — though, as yet, without any of the predicted disastrous results. There is one very important thing to note about Koza's creation: though brilliant, this computer does not have any will of its own. Like any other computer, it needs to be given an initial set of instructions. After that, however, things change. The "invention machine" does not actually invent things itself, but it is able to tackle very complex problems and develop its own set of solutions without any further input: "What Koza has done is to automate the creative process" (Keats, 2).
Koza himself calls the system his computer uses genetic programming. In large part, his machine works along the same principles that have been applied to genetics. After generating a huge number of random solutions to a given problem, the invention machine determines how effective each solution is, then discards most unsuccessful solutions and begins making random alterations to others, combining aspects of the most effective solutions to create a new generation of candidates. This process continues until an optimal solution is reached — meaning all other solutions modeled by the computer perform less efficiently. As Keats describes it, "over and over, bits of computer code are, essentially, procreating," mirroring "Darwinian evolution, the process of natural selection" (Keats, 1). The computer does not tackle problems in exactly the same way as humans in every situation, though trial and error are parts of many scientific discoveries and engineering projects. But although its method is limited to one basic system, it performs that system far faster than any human brain.
"Koza's advance beyond constrained genetic algorithms"
Koza's invention machine is far from taking over the world. As of now, it cannot choose which problems to work with, but must still be given an opening set of instructions. When computers develop a will of their own in addition to the intelligence that Koza's invention machine displays, it might be time to worry. Until then, we can continue to admire and wonder at the capabilities of our artificial machines.
You’re 71% through this paper. Sign up to read the remaining 1 section.
Sign Up Now — Instant Access Already a member? Log inAlways verify citation format against your institution’s current style guide requirements.