ChatGPT has become the first artificial intelligence (AI) chatbot to pass the Turing test based on human-like intelligence, scientists have claimed.
This test was proposed by computer expert Alan Turing, according to which an AI technology can be considered truly intelligent when people are unable to tell whether they are talking to a human or a machine.
Scientists at the University of California claimed that chat GPT4 is very difficult for humans to recognize through written communication. This test was introduced in 1950 by British scientist Alan Turing. He predicted that one day the computer’s capabilities will be equal to human intelligence. Therefore, he proposed a test to identify whether a computer is capable of human intelligence or not. Under this test, a person conducts a written conversation with a human and a computer to determine who is the machine and who is the computer. If the person fails to determine, it means that the computer has passed the test.
Keeping the same test in mind, American scientists recruited 500 volunteers and had them interact with 4 different agents out of which 3 were AI bots and one was human. AI chatbots included ChatGPT 4, ChatGPT 3.5, and the 1960s chat program Eliza.
Volunteers were made part of an online chat room and instructed to determine whether the person speaking to them was a human or an AI chatbot.
The individuals chatted for 5 minutes and then left the chat room, after which they were asked if they could tell whether they were talking to a human or a computer. The results showed that the Eliza program was able to fool 22 percent of people at a time, with ChatGPT 3.5 having a success rate of 50 percent. But Chat GPT 4 was more successful in this regard, with 54 percent of people calling it human.
This means that Chat GPT4 is capable of passing the Turing test more successfully.
Source: In a first, ChatGPT 4 passes human-like intelligence test – Jasarat
