A Summary of Alan Turing’s Computing Machinery and Intelligence
A summary of computer scientist Alan Turing’s Computing Machinery and Intelligence in 1950.
In Computing Machinery and Intelligence, computer scientist and philosopher Alan M. Turing proposes to consider the question “Can machines think?” and argues that there is no convincing argument that machines cannot think intelligently like humans, and different approaches can be undertaken in hopes of advancing machine intelligence.
The question “Can machines think?” can be formulated as the “Imitation Game” (p.1), where an interrogator attempts to distinguish a machine from a human to draw a difference in thinking ability. Machines involved in the game are defined as “digital computers” (p.3) that execute fixed instructions without any discrepancy. Some assumptions about the digital computers hold: digital computers have limitless storage capacity and are universal machines that can replicate any machine with discrete, non-continuous states. Such assumptions are theoretically significant as they imply that only one digital computer is required to be programmed for every context.
Turing predicts that in half a century’s time, human interrogators will not correctly differentiate machines from humans with more than 70 percent chance in the Imitation Game, and considers disagreeing opinions to the prediction. A theological objection that machines cannot think as God has not given a soul to them does not hold, because an omnipotent God would face no difficulty conferring a soul to a machine. A mathematical argument that discrete state machines are limited in answering questions can be dismissed by the lack of proof that humans are not restricted by the same limitations.
Given the undeniable possibility that machines could “learn” (p.14) based on a “conditioned reflex” (p.14), the ability of machines to think can be framed as the ability of machines to learn. A human mind can be replicated by programming a child’s brain that has relatively few processes, and then influencing it with “education and other experiences” (p.19). The learning process to improve a machine can be made faster than the process of evolution through intelligent teaching methods. A carrot-and-stick approach can be undertaken to guide the machine toward improvements.
A machine that learns may seem paradoxical because a conventional machine’s rules do not change. However, the rules that change over time in the learning process are temporary. The learning process is a tendency to act favourably toward specified objectives. The hope is that in time, machines will be able to rival humans in all intellectual fields, but for now, there is still much exploration and experimentation to be done. (398)
Turing, Alan M. “Computing machinery and intelligence.” Parsing the turing test. Springer, Dordrecht, 2009. 23–65.
