Terrifying study reveals AI robots have passed ‘Turing test' — and are now indistinguishable from humans, scientists say
The dystopian lessons in every sci-fi movie from 'Terminator' to 'Ex Machina' appear to be coming true. Artificial intelligence has become so sophisticated that bots are no longer discernable from their human counterparts, per a concerning preprint study conducted by scientists at the University of California in San Diego.
'People were no better than chance at distinguishing humans from GPT-4.5 and LLaMa (a multi-lingual language model released by Meta AI),' concluded head author Cameron Jones, a researcher at UC San Diego's Language and Cognition Lab, in an X post.
The researchers set out to see if AI bots, which are programmed to parrot human speech patterns, could pass the iconic Turing Test.
Developed by British WWII code breaker and computer scientist Alan Turing, this tech-istential exam gauges the intelligence of machines by determining if their digital discourse could be differentiated from that of a human — if they can't tell the difference, the machine has passed.
Researchers tested four large language models (LLMs) — GPT-4o, LLaMa-3, and GPT-4.5 and Eliza (a 1960s era chat program) — in two randomized and controlled trials, Mashable reported.
To administer said Turing Test, they enlisted 126 undergraduate students from the University of California San Diego and 158 people from online data pool Prolific, the Daily Mail reported.
These participants were instructed to have five-minute simultaneous online exchanges with a robot and a human to see if they could spot the Decepticon — with the kicker being that they didn't know which was which.
Meanwhile, the human and AI respondents were tasked with convincing their interrogator that they were human.
Researchers found that, when'prompted to adopt a humanlike persona, GPT-4.5 was judged to be the human 73% of the time.'
This was 'significantly more often than interrogators selected the real human participant,' and of course, more than enough to pass the test, per the study.
Meanwhile, 'LLaMa-3.1, with the same prompt, was judged to be the human 56% of the time.' While not significantly less than its human brethren, this still achieved a passing grade.
Finally, the baseline models (ELIZA and GPT-4o) failed after deceiving the witnesses just 23% and 21% of the time respectively.
Researchers found that it was paramount to have the synthetic mimics adopt a human persona because, when administered a second test sans said prompt, they performed significantly worse.
Caveats aside, these 'results constitute the first empirical evidence that any artificial system passes a standard three-party Turing test,' researchers concluded.
Does this mean that AI-infused learning language models are intelligent?
'I think that's a very complicated question that's hard to address in a paper (or a tweet),' said Jones on X. 'But broadly I think this should be evaluated as one among many other pieces of evidence for the kind of intelligence LLMs display.'
Interestingly, the experts at Psychology Today concluded that the bots had beaten the Turing Test, not through smarts, but by being a 'better' human than the actual humans.
'While the Turing Test was supposed to measure machine intelligence, it has inadvertently revealed something far more unsettling: our growing vulnerability to emotional mimicry,' wrote John Nosta, founder of the innovation think tank Nosta Lab, while describing this man-squerade. 'This wasn't a failure of AI detection. It was a triumph of artificial empathy.'
Nosta based his analysis on the fact that participants rarely asked logical questions, instead prioritizing 'emotional tone, slang, and flow,' and basing their selections on which 'one had more of a human vibe.'
He concluded, 'In other words, this wasn't a Turing Test. It was a social chemistry test—Match.GPT—not a measure of intelligence, but of emotional fluency. And the AI aced it.'
This isn't the first time AI has demonstrated an uncanny ability to pull the wool over our eyes.
In 2023, OpenAI's GPT-4 tricked a human into thinking it was blind to cheat the online CAPTCHA test that determines if users are human.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
New York Times
42 minutes ago
- New York Times
Trump Administration Violated Order on U.C.L.A. Grant Terminations, Judge Says
A federal judge in California ordered the National Science Foundation to reinstate millions of dollars in grants awarded to the University of California, Los Angeles, finding that the agency had tried to circumvent a ruling in June requiring restoration of the funds. In a pointed order on Tuesday evening, Judge Rita F. Lin wrote that the Trump administration had misleadingly framed its latest attempt to cancel the grants as suspensions. 'N.S.F. claims that it could simply turn around the day after the preliminary injunction' and freeze 'funding on every grant that had been ordered reinstated, so long as that action was labeled as a 'suspension' rather than a 'termination,'' she wrote. 'This is not a reasonable interpretation.' Judge Lin, a Biden appointee, noted in the order that the University of California system had lost around $324 million in grant funding earlier this year as the Trump administration began culling science funding for projects it considered out of step with the president's agenda. In the previous ruling in June, Judge Lin informed the Trump administration that it could issue cancellations of individual grants for coherent reasons, but not blanket terminations. But beginning on July 30, the administration sent out a round of letters announcing what Judge Lin described as 'en masse, form letter funding cuts,' targeting U.C.L.A. specifically, freezing more than $300 million in research funds. That sum appeared to include around $81 million in funding awarded by the N.S.F. The judge said that letters on the cuts echoed familiar grievances about the university's handling of diversity in admissions practices, alleged antisemitism on campus and policies surrounding transgender athletes — the same grounds on which the administration has tried to extract enormous settlements from Harvard and other universities in recent weeks. Judge Lin said that the Trump administration's freezing of university grants appeared designed more to suspend research the Trump administration has associated with liberal causes than to sincerely address concerns about racism or antisemitism. In a related case focused on grants from the National Institutes of Health, a federal judge in Massachusetts described the cancellations of those grants as discriminatory toward racial and sexual minorities and driven by animus toward vulnerable groups. He similarly ordered that funding restored in an impassioned ruling from the bench in June. In Judge Lin's ruling, she directed the government to return next Tuesday to update the court on its progress in complying with the order.
Tom's Guide
9 hours ago
- Tom's Guide
Zuckerberg reveals Meta's AI superintelligence breakthrough — and why you won't be using it anytime soon
Meta is shifting gears in the AI race, claiming its systems are beginning to improve themselves; a potential early step toward artificial superintelligence (ASI). But in the same breath, CEO Mark Zuckerberg says the company will no longer release its most advanced AI models to the public, citing safety concerns. In a newly published policy paper, Zuckerberg revealed that Meta's AI has started refining its own abilities without human input. While the pace is 'slow for now, but undeniable,' he framed the breakthrough as a foundational moment on the path to ASI — AI systems that not only outperform humans in nearly every domain but can also evolve on their own. Researchers often describe ASI as the next rung above artificial general intelligence (AGI), which matches human adaptability. AGI is considered the key milestone before an 'intelligence explosion,' where AI could rapidly improve beyond human control. For years, Meta has touted its open-source approach to AI, making large language models like Llama freely available to researchers and developers. Now, that policy is changing. Zuckerberg says the company will continue releasing competitive models, but the most advanced systems will stay internal to prevent potential misuse. ASI, he warned, introduces 'novel safety concerns' that demand tighter controls, even at the expense of openness. For those unfamiliar with what open source means, it's software that's built on the principle that its source code (the instructions that make it work) is freely available for anyone to view, use and modify. A good example of a completely open source chatbot is DeepSeek. Get instant access to breaking news, the hottest reviews, great deals and helpful tips. The openness allows developers worldwide to collaborate on improvements, fix security flaws, and adapt the software for specific needs. It also promotes transparency, since anyone can inspect the code to understand how it works and ensure it's trustworthy. However, it also comes with safety concerns like the ones Zuckerberg is referencing including guardrails can be removed, it's harder to know who is using the software for harm because there is little to no gatekeeping. Meta's superintelligence ambitions are now housed under a new division: Meta Superintelligence Labs. Launched in June 2025, the group is based in Menlo Park, California, and reportedly oversees development of the ultra-secret 'Behemoth' model. Tech figures Alexandr Wang and Nat Friedman are said to be leading parts of the initiative. Meta's decision puts it at odds with rivals like OpenAI, which still provides limited access to its flagship models through public platforms. The move raises bigger questions for the AI industry: Meta is planting its flag in the superintelligence conversation and it's willing to change its long-standing approach to do it. It's hard to know if this is a necessary safeguard or a bid for exclusive AI dominance, but either way, it will likely define the next chapter of the company's role in the AI race. Follow Tom's Guide on Google News to get our up-to-date news, how-tos, and reviews in your feeds. Make sure to click the Follow button.
Yahoo
10 hours ago
- Yahoo
AI companion apps on track to pull in $120M in 2025
Demand for AI 'companion' applications outside of bigger names, like ChatGPT and Grok, is growing. Of the 337 active and revenue-generating AI companion apps available worldwide, 128 were released in 2025 so far, according to new data provided to TechCrunch by app intelligence firm Appfigures. This subsection of the AI market on mobile has now generated $82 million during the first half of the year and is on track to pull in over $120 million by year-end, the firm's analysis indicates. Unlike general-purpose chatbots, AI companion apps anthropomorphize AI interactions by allowing users to converse with custom characters, including friends, lovers, girlfriends or boyfriends, fantasy characters, and more. Appfigures defined the market segment in the same way, describing companion apps as those in which the user can interact with either premade or user-generated synthetic characters meant to embody an actual personality. Popular apps in this space include Replika, PolyBuzz, Chai, and others. As of July 2025, AI companion apps across the Apple App Store and Google Play have been downloaded 220 million times globally. During the first half of 2025, downloads were up 88% year-over-year, reaching 60 million. Appfigures crunched the numbers and found that, as of July 2025, AI companion apps have driven $221 million in consumer spending worldwide. So far this year, these apps have generated 64% more revenue than during the same period in 2024. The top 10% of all AI companion apps generate 89% of the revenue in the category, the data shows. In addition, around 10% (or 33) of the apps have exceeded $1 million in lifetime consumer spending. Revenue per download is also up $0.66 from $0.52 in 2024 to $1.18 for the category so far in 2025. While dedicated AI companion apps are fairly popular, bigger companies like xAI are also moving into the market. In July, xAI's Grok launched AI companions, including an anime girl and guy, as well as a snarky 3D fox. Meanwhile, ChatGPT's recent upgrade to GPT-5 brought to light the fact that many of its users felt a kinship with the older model, as they mourned the loss of their AI companion, whom they had come to depend upon. To address these and other concerns about GPT-5's performance, OpenAI CEO Sam Altman brought back the 4o model for the time being. Google last year tapped into the market, too, when it hired away founder, Noam Shazeer. The app lives on and still has tens of millions of monthly active users. According to Appfigures' data, the most popular AI companion apps are those used by people looking for an AI girlfriend. Of the active apps on the market today, 17% have an app name that includes the word 'girlfriend,' compared with 4% that say 'boyfriend' or 'fantasy.' Terms like anime, soulmate, and lover, among others, are less frequently mentioned. The firm notes there were likely a number of other AI companion apps that launched on the app stores since 2022, but were later removed after failing to gain traction in terms of revenue or downloads. Those weren't factored into its analysis, however.
