According to the statement of the co-founder of Ethereum, Vitalik Buterin, the model of artificial intelligence GPT-4 from Openai successfully passed the Turing test. Test Tering, proposed in 1950 by the famous mathematician Alan Turing, is the method of assessing the degree of “humanity” of the artificial intelligence system in the conversation.
Researchers from the University of California at San Diego published A scientific article that states that people cannot distinguish GPT-4 from Man in the Turing test. The experiment was attended by about 500 people who were to determine in a blind test whether they communicate with a person or with the AI model.
According to the results of the study, people mistakenly considered the GPT-4 person in 56% of cases. Buterin believes that the ability of AI to mislead more than half of the people with whom he interacts indicates that the model has passed the Turing test. He noted that now the ability to distinguish a person from a bot is comparable to tossing a coin, while people guess as people by 66%, and bots – by 54%.
Buterin emphasized the significance of this achievement, indicating that overcoming the stage of the Turing test is an important milestone confirming the seriousness of achievements in the field of AI.