14.2.3 Measuring progress: the Turing test
One of the most fundamental requirements necessary to accomplish any task is a method for detecting when the task has been completed. Absent such a test, a person will be unable to determine whether progress is being made on the task, or even recognize that the goal has been achieved.
In 1950, Alan Turing published a paper entitled “Computing Machinery and Intelligence” in which he described a test that he believed could be used to measure progress towards creating intelligent machines. The goal of the Turing test, as it has come to be known, is to determine whether a machine acts in an intelligent manner – whether weak AI has been achieved.
Here is one version of the Turing test. A human interrogator enters into a conversation with an entity (which may either be another human or a machine) via a computer terminal. The human and the entity are not allowed to see one another or communicate in any way, other than by typed messages.[5] The goal of the human interrogator is to determine whether he or she is, in fact, communicating with another human or with a machine. If the human cannot reliably determine the difference between the two, we conclude that the machine is exhibiting intelligent behavior.
Figure 14.3: A depiction of a Turing Test.
This test is designed to limit bias against the machine. Passing the test does not require an ability to see, to hear, or to speak. It does not require the ability to perform physical tasks, such as riding a bike or playing a musical instrument. What is required is that the machine be capable of reasoning in a manner similar to humans and of communicating in a human language, such as English. While the fictional computer HAL of “2001: A Space Odyssey” could easily pass such a test, no real computer has ever even come close.
Figure 14.4: A demonstration of Eliza the first and most famous conversation program.
Footnotes
[5] The modern equivalent would be texting from your smart phone.