Can chatGPT think?
50+ years ago, Alan Turing was already thinking about it.
Recently, the Queen of England passed away, and on the day of her passing, a rainbow appeared near Buckingham Palace. Some believed it was good vibes for the Queen, but one guy tweeted that it was Alan Turing mocking the death of “that bitch”.
It was joking about the fact that Turing was celebrating the Queen’s death, since he was persecuted through all his life by the royals for being gay.
This tweet made me want to relive an article that Turing published in the Oxford journal of philosophy and psychology, in which he opened up the question “can machines think?”.
Turing wanted to do an experiment to answer this question, but at the time they thought that kind of research was totally out of place. Of course, machines couldn’t think like humans, after all, this was a machine:
In the quest to have his research accepted and be able to carry out his experiments on the consciousness of machines, Turing leaves the technical issues of computer science a little aside and addresses existentialist issues in an article, exposing the reader’s prejudices and making him question his own humanity.
The Imitation game
Humanity flirts with the idea that another being, devoid of a soul, can perform difficult tasks for it. Before, it was black-skinned people subjected to repetitive and tiring work, with the pretext that they didn’t have a soul or feelings.
Now we shape the machines to do the heavy work, including the cognitive one, because those machines don’t have a soul… do they? To unravel this mystery, Mr. Turing proposed the development of a guessing game.
This game has an interrogator who, through questions and answers, aims to distinguish the differences between the two other participants to discover which of the two is a man and the other a woman (Is he taking advantage of the article to make a provocative joke about the gender theme?) but that he would like to develop this same game to discover which of the two is a person and the other a machine.
For Turing, it is so difficult to understand where the knowledge of a human and a machine comes from, because physically, both pass their information via electrical impulses (us, through the nervous system).
The learning process is also a mystery, there is no way to attest if people create their skills from scratch or imitate and improve, just as machines do, through their algorithms. There are people with different intelligences (we call them cognitive capacity), just as there are machines with different computational capacities.
For these reasons and others, Turing believes that the question “Can machines think?” is very relevant and uses the article to refute the most common arguments, here are some (try reading with your own intelligence and humanity in mind):
Humans are umpredictable
The argument about the informality of behavior
The human being can act in different ways in the same situation, so it seems impossible to write rules of conduct for the machine to follow that imitate the unpredictability of a human being.
Turing refutes saying that one day he configured a program using only 1,000 storage units and the machine surprised him by answering, in just 2 seconds, a sixteen-digit number.
For him, this means that machine learning ability is much faster than ours, and he still challenges: Can you predict what the machine will respond? Therefore, for our ability to understand, the machine’s response time would be like an informal and random event in life. It’s so fast that the machine’s response seems random to us — even more so in a GPT chat.
Thinking comes from the soul
a theological objection
Thinking is a function of man’s immortal soul, machines and animals do not have the same capacity.
-Turing mocks and says that he is even too lazy to refute the church’s argument, but he says that it is only a matter of time before his questioning is accepted and becomes commonplace, just as Galileo’s discoveries were.
There is a limit to what machines can do
Argument from various shortcomings
Machines can do a lot of things, but they can’t do X — like being kind, creative, beautiful, friendly, initiative, sense of humor, knowing right and wrong, making mistakes, falling in love, liking strawberries and cream, making someone fall in love with it or learn from the experience.
-Although we know that machines cannot do X today, we shouldn’t limit ourselves to what we know. The unknown is always supposedly limited in some respect. Why not explore what we don’t already know? Why not venture into new possibilities? If we didn’t go beyond what we know, we wouldn’t go to space.
We sometimes use this type of affront against people too, usually in a racialized way. We think that a poor person, black, Arab, woman, etc., is not capable of this or that.
We also use this argument for more abstract things, such as subjugating other cultures: an English child, in his arrogance, may think that learning French is nonsense, as this skill is very little useful in his reality. However, as you grow and experience, you may find that the French language can also do X or Y.
Therefore, this type of induction is not scientific and especially not suitable for human inventions. We always have to explore space-time more — And how many times have we not been wrong throughout history?
Machines can never create something new
Lady Lovelace’s Objection
The analytical engine does not pretend to originate anything; she just follows orders. Another variant of this argument is that a machine can never do something new; she can only reproduce.
-This is the same as assuming that as soon as the facts are presented, all the consequences of that fact arise in the mind along with it. However, this does not happen. We need time and contact with anything or anyone to understand its possibilities.
Machines have no feelings
The argument from conscience
One professor raised the argument that machines cannot feel.
- Turing accuses it of being a solipsistic point of view, that is, the idea that only we can be sure that “we” feel and think, there is no way to be sure that nothing or no one else feels and thinks too.
He considers this theory immoral and believes that a kinder thought is to think that everyone feels and that there is no way to prove whether people create or imitate.
It is clear that there is a mystery surrounding the consciousness of each being, but there is a paradox when trying to solve this question, so this question does not need to be answered definitively before the one he proposed: “Do machines think?”
Machines will always be inconsistent at some point
The mathematical objection
There are questions that some machines cannot answer (there are statements about the logical system that cannot be proved or disproved within the system). Therefore, it would be inconsistent to create a game that would fall into this problem.
-Turing raises some theorems that could be used as a basis for the program that will play the imitation game, for example, if Gödel’s theorem were used, we would need to have ways of describing logical systems in terms of machines and machines in terms of logical systems .
He understands that usually when there are inconsistencies and the machine answers something “wrong”, we feel superior to them, without remembering that we also make mistakes often.
So, everyone has some inconsistency, but that doesn’t invalidate the fact that this game could improve understanding of the issue and that we could create many games, after all, there is no way to beat all machines (past, present and future).
Machine system is linear and less complex than human
Argument from continuity in the nervous system
The nervous system is not a discrete state machine, as causes and consequences do not happen in the same proportion. Therefore, it is not possible to mimic its behavior with a discrete state machine.
-However, within the rules of the game, the interrogator will not gain any advantage from this difference. For example, if we use a differential analysis machine, the digital computer will not be able to get the answers exactly right, but it will be able to give approximate correct answers. Thus, it would be very difficult for the interrogator to distinguish which system is the nervous system and which is the machine’s.
The “heads in the sand” objection
The consequences of thinking machines would be so dangerous that it’s better not to think about it.
- He thinks this argument doesn’t have much substance (yes, in many moments the article has a jocular tone, that’s great).
Turing’s text continues with further rebuttals and considerations on how to assemble a machine to play the guessing game.
The game could take decades to build. Compared to other issues, considering that it took us hundreds of years to legally abolish slavery and that racism is still an active agenda, the time frame does not seem out of the human scale of historical discoveries.
More than 50 years have passed since Mr. Turing’s article, machines have evolved a lot and the question of whether they think is increasingly topical.
At that time, he believed that there would be a computer capable of playing the imitation game with such performance that, in just 5 minutes of interaction, the interrogator would have less than a 70% chance of identifying who is the machine and who is the human.
This would make the question “Can machines think?” insignificant.
Turing’s death
In England, less than a hundred years ago, homosexuality was prohibited. So Alan and thousands of other men were subjected to experimental (they were guinea pigs) and disturbing hormone treatments to cure themselves of this “disease”.
Unfortunately, the legalization of homosexuality only came in 1967, too late for Turing, who committed suicide in 1954, most likely by poisoning himself.
In August 2009, British programmer John Graham-Cumming started a petition urging the British government to apologize for its accusation of Turing as a homosexual, a petition recognized by Prime Minister Gordon Brown who apologized and described Turing’s treatment as “appalling”.
It is quite sad to think that Turing was persecuted for being homosexual and, therefore, his contribution to science was so reduced: he wrote this article when he was only 41 years old — 4 years before his death.
- link to original article by Mr Turing:
