Si from Openai officially reached a person's level

< img src = "/uploads/blogs/ec/a9/ib-free2jdi_db2585858e.png" Alt = "Artificial Intelligence from Openai officially reached a person's level"/> ~ < P > California University Researchers at San Diego Cameron Jones and Benjamin Bergen on March 31 & nbsp; nbsp; nbsp; Nbsp; Results Trigoing Test. In it GPT-4.5 in 73% of cases was perceived by subjects as a person who is the first similar achievement among the SI-systems.

< p > during the experiment, the participants simultaneously communicated with the person and one of the models, after which they tried to determine which of the interlocutors were. For other systems, the results were lower: LLAMA-3.1 & ndash; 56%, Eliza & ndash; 23%, GPT-4O & ndash; 21%.

< P > The standard thuring test assumes that if a person cannot distinguish a car from another person in correspondence, then AI successfully imitates intelligence. In 2024 a similar test showed that Chatgpt was mistaken for a person in 54% of cases.

< p > for the first time GPT-4.5 was represented by Openai in February 2025, its key improvement was the & ldquo; emotional intelligence & Rdquo;. The new experiment confirmed that the model has advanced further into the ability to reproduce human communications.