ChatGPT passes the Turing test: Could artificial intelligence be equivalent to human intelligence?
First proposed in 1950, passing the “Turing test” is considered one of the highest goals in artificial intelligence. According to a new claim, the chatbot ChatGPT has become the first AI to achieve this milestone.
ChatGPT: The First AI to Pass the Turing Test?
Researchers claim that ChatGPT, a chatbot developed by OpenAI, is the first artificial intelligence to pass the Turing test.
According to the claim, most people cannot distinguish the chatbot’s responses from those given by a human.
Proposed as the “imitation game” by computer pioneer Alan Turing, the Turing test argues that if the responses given by an AI cannot be understood as machine-generated by a human, then the AI should be considered truly intelligent.
Scientists suggest that ChatGPT-4 can fool more than half of human test subjects.
Study Reveals GPT-4’s Human-Like Responses
In a study involving 500 participants, subjects were made to converse with four different responders: a human, the 1960s AI program ELIZA, GPT-3.5, and the latest version of ChatGPT, GPT-4.
After a five-minute conversation, participants were asked whether the responder they conversed with was a human or an AI. According to the published study, 54% of the participants evaluated GPT-4 as human.
ELIZA, which contains pre-programmed responses, was evaluated as human by 22% of participants, while GPT-3.5 was seen as human by 50%, and the human responder was correctly identified as human by 67% of the participants.
Researchers consider this study a step toward future interactions between humans and machines. Additionally, the study highlights how AI has evolved in the GPT era.
Page Contents
Toggle