The Turing Test proposed by Alan Turing in 1950 aims to evaluate machine intelligence by making it difficult for humans to distinguish between machines and humans through textual communication. Recently, scientists at the University of California, San Diego, explored the ability of AI and humans to differentiate each other when reading conversation logs rather than engaging in direct communication. They designed the Inverted Turing Test and the Shifted Turing Test. The results showed that the accuracy of AI judges was lower than that of humans who participated in direct communication, and in some cases, the AI even mistakenly identified another AI as a human. This indicates that in the absence of direct communication...

In the development of artificial intelligence, the Turing Test has always been an important milestone. Recently, researchers from the Department of Cognitive Science at the University of California, San Diego, conducted a replication of the Turing Test for GPT-4, and the results were remarkable.

A study from the University of California, San Diego, found that GPT-4 has a success rate of over 41% in the Turing Test, surpassing ELIZA. ELIZA achieved a 27% success rate in simulating humans, while GPT-3.5 only managed 14%, leading to an awkward evaluation of modern AI technology. The research indicates that ChatGPT was not specifically designed to excel in the Turing Test, but GPT-4 performed significantly better in the evaluation. The reflections on the design and performance of modern AI chatbots have not undergone peer review, garnering attention from the industry.

A study by the University of California found that GPT-4 did not pass the Turing Test and performed worse than human intelligence. The GPT-3.5 model lost in conversations with the 60-year-old AI chatbot ELIZA. The research used the Turing Test to explore the human-like characteristics of AI models, sparking academic interest. Investigators' judgments were primarily based on language style and social-emotional traits, with no significant prediction based on educational level.

A team from UCSD has confirmed that GPT-4 has not passed the Turing Test, with the classic AI ELIZA defeating ChatGPT. Humans had a win rate of only 63% in the test, raising discussions about the effectiveness of the Turing Test. The research found that ELIZA achieved a 27% success rate in convincing humans that it was human. The study examined the performances of GPT-4, GPT-3.5, ELIZA, and humans through various design changes, including altering interaction prompts.

GPT-4 achieves a success rate of 41% in the Turing Test, surpassing traditional models ELIZA and GPT-3.5. Research by Jones and Bergen shows that GPT-4 demonstrates a significant advantage in the field of artificial intelligence. The study indicates that GPT-4 is a paid version, which performs better than the free version GPT-3.5 in the tests. Despite the progress made, there is still room for improvement in artificial intelligence in the Turing Test, reflecting the potential limitations of the technology.