As I’d promised in the previous article : Should you be worried about AI?, this blog post is going to be about the Turing Test or “the imitation game” as he called it. Most of the information I got for this article was from the paper that Turing had written called “Computing Machinery and Intelligence”.
If you like what you read, please leave a like. You can comment with your opinions and suggestions below even if you’re not on WordPress. Also, please follow by leaving your email address in the right hand side of the page to receive updates (no spam guaranteed) when I post something new. There is no other way I can know if my readers like what I write. Thank you so much!
Before we begin understanding the Turing Test (if you’re in a hurry you can skip to the next section) it would probably help to know who the man behind the idea was. Alan Turing was a British logician, mathematician and computer scientist who is famous today for breaking the ciphers on the German Enigma machine. Basically, Turing was able to break the intercepted coded messages that the Germans were beaming and thus put the British at an advantage. It is believed that this reduced the length of WWII by nearly four years. However, being the ungrateful country that Britain was at the time, Turing was prosecuted for homosexual acts in 1952 – when they were sure that they didn’t need him. Refusing the option of going to prison, Turing chose chemical castration and eventually died of cyanide poisoning (probably suicide). 57 years later, the British Prime Minister decided to apologize “for the appalling way” in which was treated. And the Queen apparently “pardoned” him. We don’t need to be hardcore liberals here to understand how stupid the whole apology thing was-but it is what it is. Talk about crying over spilt milk.
Can Machines Think?
The initial problem that Turing poses in the paper is that of whether machines can think. However, if we took the meanings of the words “machine” and “think” from their common use, it would be easy to conclude that the posed question can be sought after in a statistical public survey. It shouldn’t be hard to imagine why this is a fallacious conclusion especially from a scientific perspective. This question is replaced by another in the form of a game. The rules of the game are as follows:
- Three players: Tom, Dick and Harry play the game.
- Harry plays the role of the interrogator.
- Tom is a human being and Dick is a computer.
- Harry doesn’t know which one is the computer and which one is the human.
- Harry’s objective is to guess accurately by posing questions at the two.
If the interrogator is fooled by the computer and is unable to guess the identities of Tom and Dick accurately, then the computer has passed the “Turing Test”. One of the advantages of using the Turing Test is that it draws a very sharp line between the physical and mental capabilities of a human. This means that the interrogator cannot ask for practical demonstrations from the test subjects. Here’s an example question-answer session in a Turing Test:
Now, we must understand that the objective of the machine is to provide answers that would normally be given by a man. This means that the machine must try to mimic the behaviour of humans while playing the game. Hence the term: “imitation game”.
There are various refutations to the idea that machines can really think. They are as follows:
- The Theological Objection
Thinking is a function of man’s immortal soul. God has given an immortal soul to every man but not to machines. Hence machines cannot think. However, theology is usually wrong on things that can be explained with science and logic. Turing brings up the example of Galileo’s time when mere statements in religious texts were adequate refutations of the Copernican theory of the Earth revolving around the Sun. This theological objection thus is not solid and can thus be discarded as there exists no concrete proof of its validity.
- The Consciousness Argument
This is a common belief held by philosophers and artists that a machine cannot think unless it can write a poem or wax lyrical about its emotions and thoughts. The real question, and a very valid one, is whether a machine is truly “conscious” of its existence. An extreme way of thinking about this is that unless we “become” the machine and “feel” the thoughts, we cannot say that a machine can think. This is similar to the solipsist argument. In order to refute this, I ask you as the reader, how you truly trust that another human being can think? This question must be thought about at the most fundamental level and you’ll find that indeed there is no way that you can say for sure that anything other than your own thoughts are real. This is a very interesting concept to explore but needless to say holds no practical implication. Most people who have thought about this go on with their daily lives without being affected by it too much. Thus for all practical intents and purposes this argument need not be given much importance.
I strongly believe that computers in the future will be able to write poems and sonnets since we’re making concerted steps to replicating the human brain and the entire nervous system. People who believe in the “life force” and the intangible “soul” will say that however close you come to replicating the human brain, you will never be able to add that magic ingredient. I guess it’s just a matter of time until we find out. But my concern is that these religious people will then brand computers “Satanic” and then they’ll abandon any inkling of reason or logic when talking about this…
Sorry about the tangential rant.
Have computers ever passed the Turing Test?
No! A computer has managed to fool a judge once in a competition before but that was mainly due to the human contestant (Tom) pretending to be a chatbot. It is sad because Turing, in his paper, had predicted that by 2000 a computer would be be programmed such that after 5 minutes of questioning, the interrogator would have not more than 70% accuracy in predicting the identity of the subjects.
The validity of the Turing Test
- It is argued that the Turing test only tests whether the computer behaves like a human being, it doesn’t actually test whether the human being is intelligent. This is based on the fact that some of human behaviour is unintelligent. This includes the human tendency to lie or even typing errors. Thus it means that if the machine wants to appear human, it must deceive the interrogator by not appearing too intelligent.
- Another argument posed against the Turing Test is this: the test is only concerned with how the subjects act. This means that the test is based on a third person view of the subjects. This argument is similar to the “Consciousness Argument” posed above for which I have already provided some refutation.
- The anthropomorphic fallacy is an argument that states that humans have an inherent tendency to consider inanimate objects as human. For example: giving human names to cars and talking to them, considering the sun to be a human-like God. However, I think that this is a rather weak argument because when we do these aforementioned things, we still recognise the difference between truly human and merely human-like characteristics. We would have to be mentally unstable to truly believe that cars and humans having no differences between them.
- The argument that actually makes the most sense is this: is it really helpful for robots to behave like humans? It might be a fun and interesting idea to think about whether computers can really think but in the end, we want them to solve problems that we have.
The philosophy of AI is unlikely to have any more effect on the practice of AI research than philosophy of science generally has on the practice of science.
The question of the validity of the Turing Test is open for debate. Does the Turing Test actually determine whether a machine can think? Is it of any significant importance to the advancement of AI? Leave your opinions in the comment section down below and don’t forget to follow!