In a post on X, Ethereum co-founder Vitalik Buterin said that OpenAI’s GPT-4 has passed the Turing test, citing recent preprint research from the University of California, San Diego, which shows that a production model has finally passed the Turing test.
Ethereum co-founder Vitalik Buterin claims that OpenAI’s GPT-4 generative artificial intelligence (AI) model has passed the Turing test.
A nebulous standard used by AI systems to assess how human-like a conversational model is is called the Turing test. The test was proposed in 1950 by renowned mathematician Alan Turing, who is credited with coining the name.
Turing claimed that at the time, an artificial intelligence system would exhibit “thought” if it could produce language that fooled people into believing they were speaking with another human.
The man who is mostly credited with creating the second-most popular cryptocurrency in the world, over 75 years later, believes that recent preprint research from the University of California, San Diego shows that a production model has finally passed the Turing test.
“People cannot distinguish GPT-4 from a human in a Turing test,” according to a preprint report released by researchers at the University of California, San Diego. To see if they could tell which was which, scientists put about 500 human test participants through a blind interaction between humans and AI models.
The study found that 56% of the time, people believed incorrectly that GPT-4 was a “human.” This indicates that, more frequently than not, a machine tricked people into believing it was one of them.
Vitalik Buterin’s take
Buterin claims that an AI system passes the Turing test if it can fool over 50% of the humans with whom it comes into contact.
Buterin added:
“It means people’s ability to tell if it’s a human or a bot is basically a coin flip!”
“Ok not quite, because humans get guessed as humans 66% of the time vs 54% for bots, but a 12% difference is tiny; in any real-world setting that basically counts as passing,” Buterin clarified in response to a question about his claim.
The Turing test is “by far the single most famous socially accepted milestone for ‘AI is serious shit now,'” he subsequently added in reaction to criticism of his initial cast. We should therefore keep in mind that the milestone has now been reached.
The Turing test
Despite the terms being frequently confused, artificial general intelligence (AGI) and the Turing test are not always connected. Based on his mathematical prowess, Turing devised his test and envisioned a situation in which AI could deceive people into believing it was a person through speech.
It is important to note that the Turing test is a transient construct without a solid technological foundation or benchmark. There is disagreement in science on whether computers are able to “think” in the same sense as living things and how to quantify this ability. Put simply, the scientific and technical community do not yet define or measure an artificial general intelligence’s (AGI) or AI’s capacity to “think.”
Long before token-based AI systems and generative adversarial networks—the forerunner of today’s generative AI systems—came into existence, Turing formulated his conceptual predictions.
Artificial general intelligence
To make matters more complicated, artificial general intelligence (AGI) is sometimes linked to the Turing test. A “general intelligence” is defined in scientific terms as the ability to perform any intelligence-based task. Humans are excluded from this since no one has ever demonstrated “general” ability across the board in intellectual pursuits. It follows that the cognitive capacities of an “artificial general intelligence” would be vastly superior to those of any known human.
Nevertheless, GPT-4 does not meet the criteria for genuine “general intelligence” in the rigorously scientific meaning. That hasn’t stopped members of the AI community, though, from referring to any AI system that can deceive a sizable portion of people as “AGI.”
Terms and phrases like “AGI,” “humanlike,” and “passes the Turing test” are frequently used in today’s culture to describe any AI system that generates content that is similar to that created by humans.
Recently, GreatGameIndia reported that, according to Wired, OpenAI discussed the prospect of initiating the generation of AI porn in age-appropriate contexts in a commentary note of the Model Spec document.
Hang on a minute … Buterin speaks like a robot as do most Californians so how could the people involved in the study know whether GPT-4 communicated like real human or not.