Intelligent Machines: A Historical Perspective on Turing's Test
Alan Turing is often considered one of the first people who theorized the possibility of artificial intelligence. In fact, in 1935, Turing conceptualized an abstract machine for computing that was equipped with limitless memory and could process data to perform tasks. Intrinsic to this machine was the idea that the machine kept on improving itself by modifying its program over time. This is very close to techniques of machine learning that artificial intelligence systems incorporate today.
Often, one of the central questions associated with artificial intelligence is the fact if machines can ever really be capable of the abilities that we associate with the word “intelligence”. Towards this end, Turing introduced a practical test to gauge the measure of computer intelligence. For Turing, it is rather meaningless to try and determine if machines can think, what is far more interesting and fruitful to consider is the ways in which machines process intelligence.
As such, the Turing test involves three participants, a computer, a human interrogator, as well as a human foil. The human interrogator is supposed to ask questions to the other two participants to determine which of them is the computer. As such, all communication takes place via an online medium. The interrogator is free to ask as far-reaching or as penetrating a question as they may like, and the computer is allowed to do everything possible to evade correct identification, including by replying in the negative for a question that directly asks it if it is the computer. On the other hand, it is the job of the human foil to aid the interrogator to deduce a correct identification. According to Turing, this process is repeated over and over again with different people in different roles to collect a robust sample size. Then, if a sufficient number of people are incapable of correctly distinguishing between a computer and a human, the computer can be considered as an intelligent entity according to the Turing test.
Significantly, the Turing test has exercised enduring control over the imagination of several people interested in machines and artificial intelligence. Thus, in 1991, an American benefactor pledged a 100,000 payout to the first computer that would be able to successfully transcend the Turing test. However, as of now, no AI program has come even close to successfully evading the Turing test. Even as there are emerging conversations about whether ChatGPT would be able to clear the Turing test, experts claim that technically ChatGPT is a large language model, which means that the content it generates is directly based on the data it is trained on. Therefore, as we continue to foster newer developments in AI territory, the Turing test serves as a gold standard of what machines need to do in their quest to emulate human intelligence.
Sources:
https://plato.stanford.edu/entries/turing-test/
https://www.britannica.com/technology/artificial-intelligence/The-Turing-test