Anyone interested in digital computers and artificial intelligence (AI) has probably heard of the Turing Test. When British founder Alan M. Turing (1912-1954) discussed the question of whether machines could think in his article “Machine Computing and Intelligence” in 1950, he proposed the famous thought experiment, later named after him—because real computers did not exist at the time. .
According to Turing’s “Imitation Game”, a person must communicate in writing with two interlocutors hiding behind a curtain – a person and a machine – and find out in a dialogue which is the device that only imitates the behavior of human communication. If the examiner cannot make a decision even after a long conversation, the machine must be given intelligence similar to that of a human.
Much is known about the Turing Test, but few have actually read the Era-Making article. For me, too, a belated reading of the edition, now published by Reclam in both languages and commentary, was a surprise. Not only does the text have classic clarity, but it essentially predicts the entire debate about artificial intelligence that has been going on ever since. This is all the more surprising because Turing himself was only laying the theoretical foundation for computer science, without there being such a thing as semiconductor transistors, chips, and digital computers — not to mention neural networks and machine learning.
It is my intellectual pleasure how Turing lists the arguments that can be made against thinking machines—which have been tirelessly circulated ever since—and discusses their pros and cons in short sentences. The “theological objection” states that only man with his immortal soul can think; On the other hand, Turing objects that it would be a limitation of God’s omnipotence to prevent him from enabling AI. Turing cannot accept the ostrich’s objection that AI is so dangerous that it should be banned immediately. With the “mathematical objection” Turing anticipates an argument which has meanwhile been put forward in several thick books by the mathematical physicist Roger Penrose (Nobel Prize 2020): from the incompleteness theorems derived in 1931 by the Austrian logician Kurt Gödel – as well as from the Thuringian problem – It turns out that there are in principle non-determinable statements for mathematical systems and therefore for computers as well; So AI machines are limited. Turing admits this with a shrug. Such peculiar problems would not play a decisive role in the Turing test.