Its developers and organizers of the competition in London: we passed the Turing test, in an experiment where there were many independent conversations
The British University of Reading reports a breakthrough in the field of artificial intelligence, when a computer won the Turing test, the same test proposed by Alan Turing, one of the fathers of computer science. They used a supercomputer to create a character named Eugene Gottesman, a 13-year-old boy, who managed to convince 33% of human judges that he was human.
The test was held under the auspices of the Royal Society in London on Sunday this week.
Eugene, a computer program that simulates a 13-year-old boy was developed in St. Petersburg, Russia. The development team included Eugene's creator, Vladimir Vaslov, who was born in Russia and now lives in the US, and Yevgeny Demkachenko, a Ukrainian who now lives in Russia.
The Turing test is based on the question and answer game "Can machines think?" Written by Alan Turing. The experiment investigated whether humans can recognize whether they are talking to a mechanic or to another person. The event took place on the 60th anniversary of Turing's suicide and half a year after he was retrospectively pardoned for being accused of homosexuality, which was a criminal offense at the time.
According to the rules of the game, if the human participants are wrong and think of the computer's answers as if they were provided by a human more than 30% of the time in a series of chats, this is considered a victory for the computer in the Turing test. So far no computer has achieved this figure so far. Eugene managed to convince 33% of the human judges that he was human.
Prof. Kevin Warwick, Visiting Professor at the University of Reading and Vice-President for Research at Coventry University, said: "There are no more iconic and controversial milestones in the field of artificial intelligence than the Turing Test, when a computer convinces a sufficient number of people that it is human. But it is fitting that such a competition will take place under the auspices of the Royal Society in London, the home of British science and the place where many breakthroughs in human knowledge have occurred over the centuries.
"Some will argue that the test has already been passed a long time ago. Similar competitions, which bore the name of the Tyring test, were held in several places in the world, but this test included the largest simultaneous comparison capability ever, when parallel and independent calls were made. A real Turing test is not a series of pre-planned questions or topics, therefore we are proud to announce that this is the first time a computer has officially passed the Turing test."
He adds that the big fear is that computers that are used by people, could be used to commit computer crimes. "It is important to study how real-time online communication can influence people in such a way that they think something is real, when in fact it is not."
And Vaslov adds: "I want to thank all the team members who worked on the development of Eugene Gutesman. We are thrilled with the results, this is an impressive achievement and we hope it sparks interest in royal intelligence and chatbots. Eugene was 'born' in 2001. The main idea is that he can claim to know everything, but his age suggests that he doesn't really know everything. We spent a lot of time developing his character in a believable way. This year, for example, we improved the 'Dialogue Controller' which makes the conversation more like a conversation between people compared to software that only answers questions. We continue with the plan to test Eugene and will continue to work and improve what we call 'conversational logic'."
And adds Warwick, "Shortly before his death on June 7, 1954, Turing, himself a Royal Fellow, predicted when a computer would pass the test, but he certainly could not have imagined what today's computers would look like, and the networks connecting them."
More of the topic in Hayadan:
8 תגובות
Life
Thanks! I read the article. In general, I think we agree, but not in details. For example - you said that the animal is allowed from the computer. I think this is not a fair comparison. If you want to compare - either disconnect the brain and place it in a suitable jar, or set up the computer inside a suitable robot.
Another point: a computer consists of two parts - hardware and software. Our brain consists of parts of one type - nerves (I know this is far from true, but I want to simplify). Each nerve is a small computer, including input, processing, memory and output. In my opinion, this difference is extremely fundamental, and is equivalent to the difference between a finite state machine and a Turing machine. In particular - I think that the human mind is stronger than a Turing machine and therefore it is impossible to simulate it on a computer.
One more short point (I have a lot of points....): Our brains evolved for a certain purpose. Computers are designed by people, for different purposes. I think this is also a fundamental difference between the 2 things.
Miracles
I am attaching a link to my article on the problematic nature of artificial intelligence. This article was first published in 1987 in the magazine "People and Computers". In the mathematical section, there was a disruption in the transition from software to software. I hope you understand what this is about.
https://www.hayadan.org.il/ai-iluusion-or-reality-290512/
Life
The programs have convinced quite a few people that they are human. Don't get me wrong - I don't believe that a computer will ever think like a human being. My intention was that this test, in my opinion, is not a real test of intelligence.
Miracles
Two comments
A. The fact that she did not ask for payment already indicates the problem. Such a large amount can show a sense of humor. Is the computer able to go? Will he give an invoice or not, and if he doesn't give it, will it be possible to accuse him of tax evasion? Describe to you a situation in which the Lahav 433 unit would ask him for a capital statement.
The problem with the ELIZA software was that it takes words from what the speaker says and constructs a question from that. In the experiments made with this software, the subjects realized within a few minutes that they were having a love affair with a computer... and the very repetition of the sentence "Why are you asking me that?" can raise suspicions in the user, since a person answering questions will phrase his question differently each time and does not recycle his question, not to mention his intonations. Imagine a situation where you enter into a discussion with another person and in every sentence you repeat the same question endlessly. I'm more than convinced that you would go out of your mind and break his bones too.
Life
It was invented a long time ago. There was software in the early 70s called Parry. This software conducted a conversation like a paranoid. Every question she didn't know how to answer - she answered - "Why are you asking me that!"
And Parry was not the first show - in the mid-sixties there was a famous show called ELIZA that pretended to be a psychiatrist - if you told her "everyone hates me" - then she answered "why do you think everyone hates you" and so on. The easy way to find out that it's software and not a person - she didn't ask for 1200 shekels for a half hour call....
Elad Stern
Moreover. If you let the computer perform some action, is it able to say I don't want or I don't feel like it?
Although they invented a computer that can claim to know everything, and fool the brain that it is human. Will they succeed in inventing a computer that claims it does not know everything, without falling into the human impression?
Obviously if I see a sub level of chat I will believe it is a boy. Give this test to pedophiles, they are the ones who spend days hanging out with children.
In my opinion, it is quite clear that the Turing test should not be with the image of a child or with the image of an autistic person, but with the image of a mature and healthy person.