‘Chatbot Eugene Goostman passes Turing test’ – update

Chatbot Eugene Goostman reportedly passed the Turing test. That reports the University of Reading, where the test was held last weekend. According to the university, the bot convinced a third of the participants that it was dealing with a human being.

The Turing test comes from the 1950s and is named after the British computer scientist Alan Turing. He stated that a machine shows human intelligence if it can convince thirty percent of the participants that it is human during an experiment.

In London, the University of Reading, supervised by the European consortium RoboLaw, held a competition in which scientists subjected several chatbots to the Turing test. In addition to Eugene Goostman, four other bots also participated, including Cleverbot and Albot.

According to the University of Reading, Eugene Goostman convinced 33 percent of the human participants that the program was human during the 2014 Turing competition. That means the chatbot is the first to pass the Turing test in this form. There have been previous variants and it has therefore already been claimed that a program can pass the Turing test. “It has never happened before that so many tests were performed simultaneously, which were independently verified. It is also crucial that the conversations were not restricted; in a real Turing test, no questions or topics are predetermined,” the university said. know in a statement.

Eugene Goostman is a computer program that a team, led by Russian computer scientist Vladimir Veselov, has been developing since 2001. The program simulates the character of Eugene, a 13-year-old boy from Odessa, Ukraine. The boy has some characteristic features. He talks about his guinea pig and his father, who is a gynaecologist. “Our idea was that Eugene can claim to know everything, but his age makes it perfectly logical that this is not the case,” the scientists say.

The chatbot breaks the sentences that the program receives into pieces, so-called tokens. Those tokens are given a certain rank and based on that, the bot gives a ‘relevant’ answer, in which the information is retrieved from a database. According to the programmers, Eugene can also feed back information related to previous sentences in a conversation. The bot also distinguishes itself with a personality, which not all competing bots like Cleverbot have.

The computer program is not an unknown name, by the way. In a similar contest at the University of Reading two years ago, Eugene convinced 29 percent of the subjects that the program is human, which just barely passed the Turing test. In addition, Veselov’s team previously made it to the finals of the annual Loebner competition, a renowned competition in the field of artificial intelligence. That competition does have the limitation that a conversation in advance must be about a certain topic. The Loebner competition was won by Mitsuku last year.

Professor Kevin Warwick of the University of Reading praises Veselov’s achievement, which he believes has implications for today’s society. “A computer program that can pretend to be someone we can trust is a wake-up call to cybercrime. The Turing test is an important tool to counter this. It’s important to know how online- and real-time communication can affect a person.”

There are, however, some caveats to be made about the experiment. For example, the University of Reading does not say how many participants were involved in the study. In addition, it is unclear whether they were aware of Eugene’s personality in advance, by which they could recognize the bot. So it is possible that personality influenced the interpretation, but this cannot be said with certainty as the research data on this is still lacking. In any case, the outcome of the experiment does not immediately mean that computers can now be ‘smarter’ and ‘more human’ than before.

The Turing 2014 competition was held in the UK last Saturday. There were multiple chat sessions per bot and for each session there were five judges. One of the judges was actor Robert Llewellyn, who plays the robot Kryten in the British science fiction comedy Red Dwarf. Another was Lord Sharkey, who campaigned for the pardon given to Turing last year.

Turing played an important role in cracking the codes of German Enigma rotor machines during World War II and is regarded as a brilliant mathematician. He became famous for his research into computability, which he published under the title On Computable Numbers, with an Application to the Entscheidungsproblem. This article gave rise to the thought experiment of the Turing machine, which is known as a milestone in the history of computer science. In 1950, in an article called Computing Machinery and Intelligence, Turing described the famous Turing test. In 1954, two years after his conviction and mandatory chemical castration, he committed suicide. That was Saturday, on the day of the competition, exactly sixty years ago.

Update, June 11 – Comments added thanks to feedback.