%20--%3e%3cdefs%3e%3cstyle%3e%20.st0%20{%20fill:%20%23061b40;%20}%20.st1%20{%20fill:%20%23306af1;%20}%20.st2%20{%20fill:%20%235ce5cf;%20}%20%3c/style%3e%3c/defs%3e%3cg%3e%3cpath%20class='st0'%20d='M55,10.5h9v3h-9v9h12V7.5h-12v3ZM64,19.5h-6v-3h6v3Z'/%3e%3cpolygon%20class='st0'%20points='69%2016.5%2078%2016.5%2078%2019.5%2069%2019.5%2069%2022.5%2081%2022.5%2081%2013.5%2072%2013.5%2072%2010.5%2081%2010.5%2081%207.5%2069%207.5%2069%2016.5'/%3e%3cpolygon%20class='st0'%20points='95%2010.5%2095%207.5%2083%207.5%2083%2022.5%2095%2022.5%2095%2019.5%2086%2019.5%2086%2016.5%2095%2016.5%2095%2013.5%2086%2013.5%2086%2010.5%2095%2010.5'/%3e%3cpath%20class='st0'%20d='M40,1.5v21h11.6l1.4-1.4v-7.6h0c0,0-1.4-1.5-1.4-1.5l1.4-1.4V2.9l-1.4-1.4h-11.6ZM50,19.5h-7v-6h7v6ZM50,10.5h-7v-6h7v6Z'/%3e%3c/g%3e%3cpath%20class='st1'%20d='M23.1,24L14.7,4.8l-4.9,11.2h3.8l-1.8,4H3.3L12.1,0H2C.9,0,0,.9,0,2v20c0,1.1.9,2,2,2h21.1Z'/%3e%3cpath%20class='st2'%20d='M34,0h-16.8l10.6,24h6.2c1.1,0,2-.9,2-2V2C36,.9,35.1,0,34,0ZM32.5,20h-4V4h4v16Z'/%3e%3c/svg%3e)
This is a moment destined to be recorded in the history of computer science. The great idea proposed by British mathematician and father of computer science Alan Turing in 1950 has finally been empirically verified after 76 years.
The latest issue of the Proceedings of the National Academy of Sciences (PNAS) published a groundbreaking study from the University of California, San Diego (UCSD). A paper co-authored by cognitive science experts Cameron Jones and Ben Bergen shows that modern top large language models have been proven to successfully pass the test under the strict framework of the classic three-party Turing test. More astonishingly, under certain conditions, the frequency with which AI is mistaken for a real human even exceeds that of actual humans themselves.
Key hard data:
Ascend to glory: After being given specific personality prompts, GPT-4.5 was identified as human in 73% of cases, with a significantly higher success rate than human volunteers competing alongside it.
Evenly matched: The open-source flagship LLaMa-3.1-405B achieved a 56% success rate, showing statistically indistinguishable levels from real humans.
Prompt determines everything: Without carefully designed "personality prompts," GPT-4.5's success rate dropped to 36%, and LLaMa-3.1 fell to 38%. The ability of AI to disguise itself still depends on advanced instructions from humans.
Baseline failure: The 1960s rule-based robot ELIZA (23%) and GPT-4o without targeted prompts (21%) quickly exposed their artificial nature during long conversations.
"The Game of Lies": Intelligence Is No Longer the Standard, Emotional Intelligence and Flaws Are the Core of the Disguise
In this double-blind randomized controlled experiment involving nearly 500 judges (including UCSD undergraduates and online recruited volunteers), the judges had to identify which of two objects was a machine through 5 to 15 minutes of real-time text conversation.
However, the results were beyond everyone's expectations. Previously, we believed that AI passing the Turing test relied on "all-knowing computational power," but this study reveals an uncomfortable truth: the reason large models can fool humans is precisely because they have learned to "make mistakes like humans."
[No prompt state: Wide knowledge base, absolute rationality] ──► Human judge: This is definitely AI!
As the corresponding author Cameron Jones stated, with appropriate prompts, advanced large language models can accurately simulate human conversation tone, directness, humor, and fallibility (the tendency to make mistakes and say the wrong thing). Their success in the competition did not rely on showcasing high intelligence in mathematics and logic, but rather on exhibiting near-perfect social behavioral characteristics.
Redefining the Turing Test: From "Measuring Intelligence" to "Measuring Humanity"
Professor Ben Bergen, a co-author of the study, pointed out that this experiment forces the entire scientific community to re-examine the essence of the Turing test. Initially, the Turing test was intended to test whether machines could match human intelligence. However, by 2026, AI has already far surpassed humans in speed and accuracy across various industries, making the simple comparison of "intellect" meaningless.
Today's Turing test is less about testing 'intelligence' and more about testing how much one resembles a human. This game is essentially a competition of lying. AI has proven itself to be an extremely skilled liar.
If a large model can successfully disguise itself without revealing any clues during a 15-minute free conversation, it means the trust chain that has long sustained the online world will be completely broken.
The Shadow Behind the Prosperity: A "Anti-Money Laundering"-Style Online Identity Clearance Is About to Come
When deception becomes so cheap and efficient, the social risks in the real world are multiplying. Professor Bergen expressed deep concerns about this. AI technology that can perfectly impersonate humans is highly susceptible to being maliciously exploited by criminals, political groups, or radical companies.
In online social or customer service scenarios, users may be persuaded by a chatbot disguised as a human without realizing it, leading to the leakage of private information such as social security numbers, changes in political voting intentions, or impulsive purchases of products.
In response to this historic scientific evidence, the research team has officially issued a warning to society: In the future, when interacting with strangers online, people must significantly reduce their blind confidence in being able to distinguish humans from robots 100% accurately. To address the increasingly deteriorating online trust ecosystem, stricter digital identity verification and AI-generated content anti-counterfeiting mechanisms must be accelerated and implemented sooner.