“How do I know I’m not a robot?”.
It seems like a simple, innocent, perhaps absurd question. But read it again and stop for a moment to think: who could be behind this phrase?
It is one of the searches that have been carried out on Google, so it could not be the work of a machine; It is formulated in the first person singular, but it would not make sense for it to be a human.
In any case, the answer to the question could be: the Turing test, or the Turing imitation game; that may sound familiar to you from the title (original, in English: The Imitation Game) of Deciphering Enigmaa 2014 film starring Benedict Cumberbatch about this brilliant English mathematician.
Decades later, already in a new century, the evolution and rise of artificial intelligence, with examples such as ChatGPT and GPT-4, among many others, has brought the famous test devised by Alan Turing back to the present day.
What is the Turing test and how does it work?
The Turing test, or imitation game, is a test created in 1950 by Alan Turing.British mathematician and computer scientist who is considered one of the fathers of computer science.
The famous Turing test is based on the following question: can machines think? Or, failing that, can they successfully imitate human thought? “If a machine behaves intelligently in all respects, then it must be intelligent,” formulated the British genius.
The premise from which it starts and with which it seeks to answer it is simple: the machine passes the test if, after a 5-minute conversation with a person (or several), it is able to convince them that it is also a human being. Thus, a human judge ‘confronts’ 2 entities: one is a person, the other is a computer or chatbot.
These are the secret websites that train ChatGPT and the like to appear smart
The test is carried out blindly, in writing and only by means of questions and answers, which may be related, for example, to knowledge of current affairs and the world, memory, common sense, humor, communication skills and understanding of the language or the ‘humanity’ of the secret entity: ‘what’s your name?’, ‘when were you born?’, ‘have you ever been to…?’, ‘what are your plans for the end of week?’, etc.
Despite having been developed in the middle of the 20th century, the Turing test is still valid today, when increasingly evolved artificial intelligence systems proliferate, such as ChatGPT, GPT-4, Bard (from Google) or Microsoft’s AI. .
Who has passed the Turing test?
Over the years, there have been many cases where the judges have erred in their verdict whether they were facing a person or a machine, but it is not clear that any machine has completely passed the Turing test and with the full support of the scientific and computing community.
In addition, it may be questioned whether the fact that a machine deceives a human necessarily means that it is intelligent and that it can think, since there are more factors at play, such as ignorance, lies, the interviewing ability of the judge, the particular conditions in every trial, misunderstandings…
“Interestingly, many of the successful systems to date could convince someone they’re talking to a person. And they wouldn’t do it with convincing human conversation, but by adding spelling or grammatical errors. The skills needed to appear human and pass the test are not necessarily linked to ‘intelligence’ and ‘thought'”, they clarify from Oxford and Google.
A case where a machine was claimed to have passed the Turing test happened in 2014, with Eugene Goostmana bot developed in Ukraine.
Goostman managed to fool a third of the 30 interviewersbut critics recall that he did not know how to answer questions such as ‘how many legs does a camel have?’, that he avoided others and that the circumstances lowered the required standards: its developers suggested that the computer imitated being a 13-year-old Ukrainian boy who did not he spoke English at a native level and lived far enough from modern society to be ignorant of various world topics, such as geography or pop culture.
Engineer fired by Google thinks his AI chatbot may have a soul, but says he’s not interested in convincing people of it
More recently, in 2022, the case for LaMDA (Language model for dialog applications).
Blake Lemoine, then an engineer at Google, claimed that the artificial intelligence chatbot had become conscious and had feelings, after which he was fired by the company. He also suggested that he could have a soul and that he would have successfully passed the Turing test, but it was later shown that this was not the case, he reports. MOF.
The next question that could be raised is whether other modern AIs —ChatGPT, ChatGPT-5, GPT-4, Bard, the next ones that will be created…— will be able to beat the Turing test, their imitation game, but for the moment not there are official studies that maintain that they have overcome it.
Demis Hassabis, CEO of DeepMind, seems to bet that it will be achieved in the future. “Philosophers have not yet established a definition of consciousness, but if we refer to self-awareness and this kind of thing… I think there’s a chance that one day AI will be.”he recently declared.
This would mean that the AI has emotions and feelings that mimic those of humans and that, perhaps, they could win the Turing game.
