AI is getting smarter every day —with thought processes already eerily similar to humans: study
Artificial intelligence doesn't just look and act human — it supposedly thinks like us as well.
Chinese researchers found the first-ever evidence that AI models like ChatGPT process information similarly to the human mind, detailing the dystopian-seeming discovery in the journal 'Nature Machine Intelligence.'
Advertisement
'This provides compelling evidence that the object representations in LLMs (large language models), although not identical to human ones, share fundamental similarities that reflect key aspects of human conceptual knowledge,' wrote the team behind the study, which was a collaboration between the Chinese Academy of Sciences and the South China University of Technology, the Independent reported.
3 The AI models displayed remarkable powers of categorization.
Malambo C/peopleimages.com – stock.adobe.com
The team reportedly wanted to see if LLM models can 'develop humanlike object representations from linguistic and multimodal data (data in different forms such as text, audio, etc).'
Advertisement
3 Prior to the study, many tech experts simply assumed that language models like ChatGPT mimicked human responses through pattern recognition.
AlexPhotoStock – stock.adobe.com
To discover whether the AI 'bot process' mirrors our cognition, researchers had OpenAI's ChatGPT-3.5 and Google's Gemini Pro Vision perform a series of 'odd-one-out' trials, in which they were given three items and tasked with selecting the one that doesn't fit, the South China Morning Post reported.
3 Researchers remain unsure if AI understands the significance or emotional value of cats and other entities on a human level.
AntonKhrupinArt – stock.adobe.com
Remarkably, the AI created 66 conceptual dimensions to sort the objects.
Advertisement
After comparing this cybernetic object sorting to human analysis of the same objects, they found striking similarities between the models' 'perception' and human cognition — particularly when it came to language grouping.
From this, researchers deduced that our psychological doppelgangers 'develop human-like conceptual representations of objects.'
'Further analysis showed strong alignment between model embeddings and neural activity patterns' in the region of the brain associated with memory and scene recognition.
Advertisement
Researchers noted that the language-based LLMs were a bit lacking with regard to categorizing visual aspects such as shape or spatial properties.
Meanwhile, research has shown that AI struggles with tasks that require deeper levels of human cognition, such as analogical thinking — drawing comparisons between different things to conclude — while it's unclear if they comprehend certain objects' significance or emotional value.
'Current AI can distinguish between cat and dog pictures, but the essential difference between this 'recognition' and human 'understanding' of cats and dogs remains to be revealed,' said He Huiguang, a professor at the Chinese Academy of Sciences' (CAS) Institute of Automation.
Nonetheless, the scientists hope that these findings will allow them to develop 'more human-like artificial cognitive systems' that can collaborate better with their flesh-and-blood brethren.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Bloomberg
a few seconds ago
- Bloomberg
Musk Tried to Enlist Zuckerberg to Help Finance Bid for OpenAI
Elon Musk tried to enlist Mark Zuckerberg for his unsolicited bid to buy OpenAI this year, the startup said in a court filing. OpenAI said Musk identified Zuckerberg, the chief executive officer of Meta Platforms Inc., as one of the people with whom he had communicated about potentially financing a deal to purchase the ChatGPT maker.
Yahoo
37 minutes ago
- Yahoo
China's DeepSeek quietly releases an open-source rival to GPT-5—optimized for Chinese chips and priced to undercut OpenAI
Chinese AI startup DeepSeek shocked the world in January with an AI model, called R1, that rivaled OpenAI's and Anthropic's top LLMs. It was built at a fraction of the cost of those other models, using far fewer Nvidia chips, and was released for free. Now, just two weeks after OpenAI debuted its latest model, GPT-5, DeepSeek is back with an update to its flagship V3 model that experts say matches GPT-5 on some benchmarks—and is strategically priced to undercut it. DeepSeek's new V3.1 model was quietly released in a message to one of its groups on WeChat, China's all-in-one messaging and social app, as well as on the Hugging Face platform. Its debut touches several of today's biggest AI narratives at once. DeepSeek is a core part of China's broader push to develop, deploy, and control advanced AI systems without relying on foreign technology. (And in fact, DeepSeek's new V3 model is specifically tuned to perform well on Chinese-made chips.)While U.S. companies have been hesitant to embrace DeepSeek's models, they've been widely adopted in China and increasingly in other parts of the world. Even some American firms have built applications on DeepSeek's R1 reasoning model. At the same time, researchers warn that the models' outputs often hew closely to Chinese Communist Party–approved narratives—raising questions about their neutrality and trustworthiness. China's AI push goes beyond DeepSeek: Its industry also includes models including Alibaba's Qwen, Moonshot AI's Kimi, and Baidu's Ernie. DeepSeek's new release, however, coming just after OpenAI's GPT-5—a rollout that fell short of industry watchers' high expectations—underscores Beijing's determination to keep pace with, or even leapfrog, top U.S. labs. OpenAI is concerned about China and DeepSeek DeepSeek's efforts are certainly keeping U.S. labs on their toes. In a recent dinner with reporters, OpenAI CEO Sam Altman said that rising competition from Chinese open-source models, including DeepSeek, influenced his company's decision to release its own open-weight models two weeks ago. 'It was clear that if we didn't do it, the world was gonna be mostly built on Chinese open-source models,' Altman said. 'That was a factor in our decision, for sure. Wasn't the only one, but that loomed large.' In addition, last week the U.S. granted Nvidia and AMD licenses to export China-specific AI chips—including Nvidia's H20—but only if they agree to hand over 15% of revenue from those sales to Washington. Beijing quickly pushed back, moving to restrict purchases of Nvidia chips after Commerce Secretary Howard Lutnick told CNBC on July 15: 'We don't sell them our best stuff, not our second-best stuff, not even our third-best.' By optimizing DeepSeek for Chinese-made chips, the company is signaling resilience against U.S. export controls and a drive to reduce reliance on Nvidia. In DeepSeek's WeChat post, it noted that the new model format is optimized for 'soon-to-be-released next-generation domestic chips.' Altman, at that same dinner, warned that the U.S. may be underestimating the complexity and seriousness of China's progress in AI—and said export controls alone likely aren't a reliable solution. 'I'm worried about China,' he said. Less of a leap, but still striking incremental advances Technically, what makes the new DeepSeek model notable is how it was built, with a few advances that would be invisible to consumers. But for developers, these innovations make V3.1 cheaper to run and more versatile than many closed and more expensive rival models. For instance, V3.1 is huge—685 billion parameters, which is on the level of many top 'frontier' models. But its 'mixture-of-experts' design means only a fraction of the model activates when answering any query, keeping computing costs lower for developers. And unlike earlier DeepSeek models that split tasks that could be answered instantly based on the model's pretraining from those that required step-by-step reasoning, V3.1 combines both fast answers and reasoning in one as well as the most recent models from Anthropic and Google, have a similar ability. But few open-weight models have been able to do this so far. V3.1's hybrid architecture is 'the biggest feature by far,' Ben Dickson, a tech analyst and founder of the TechTalks blog, told Fortune. Others point out that while this DeepSeek model is less of a leap than the company's R1 model—which was a reasoning model distilled down from the original V3 that shocked the world in January, the new V3.1 is still striking. 'It is pretty impressive that they continue making non-marginal improvements,' said William Falcon, founder and CEO of AI developer platform Lightning AI. But he added that he would expect OpenAI to respond if its own open-source model 'starts to meaningfully lag,' and pointed out that the DeepSeek model is harder for developers to get into production, while OpenAI's version is fairly easy to deploy. For all the technical details, though, DeepSeek's latest release highlights the fact that AI is increasingly seen as part of a simmering technological cold war between the U.S. and China. With that in mind, if Chinese companies can build better AI models for what they claim is a fraction of the cost, U.S. competitors have reason to worry about staying ahead. This story was originally featured on
CNBC
an hour ago
- CNBC
Household robots are about to get a big price cut — if China's top 'robovac' player has its way
BEIJING — Household robots for cleaning are about to quickly become an affordable reality. At least that's what Quan Gang, president of Beijing-based robot vacuum cleaner company Roborock, has in mind as he strategizes for the next five years. The company ranks first among smart vacuums by global market share, according to IDC Research. Last week, it reported a nearly 79% revenue surge in the first half of this year. About half of sales came from outside China. In an exclusive interview with CNBC on Wednesday, Quan predicted that human-like robots will become part of many households by 2030, thanks largely to advances in generative artificial intelligence. And before then, he expects, Roborock can make its latest, high-end cleaner with an AI-powered robotic arm so cheap that the mass market will be able to buy it — for at most a few hundred U.S. dollars. "If we only focus on the premium segment, in the end, other than being the best robotic vacuum cleaner company in the world, we will have nothing," Quan said in Mandarin, translated by CNBC. He noted that robot vacuums still don't have a very high household penetration rate. China, the largest market for the robot vacuums by value, has a penetration rate of only 5.6%, while the United States, the second-largest market, has a 22% penetration rate, according to Euromonitor estimates for 2025. The firm predicts penetration in the U.S. will tick up to 24.1% over the next two years, and edge down to 5.5% in China. Competition in the robotic vacuum + robotic arm category heated up earlier this year at the U.S. Consumer Electronics Show, with Roborock and at least two other Chinese competitors releasing demos. The AI-powered arm removes obstacles from the cleaner's path as it rolls autonomously around the house. So far, only Roborock has started selling one, called the Saros Z70 — but with a hefty price tag of around $2,600 on The site shows 141 reviews and a 4.6 rating. Roborock did not share specific sales figures. Initial reviews of the Saros Z70 from U.S. tech sites such as Mashable and Wired weren't impressed, especially given the price, but hoped for more capable versions in the near future. Both recommended that consumers stick with the more traditional Roborock Saros 10R — which retails for $1,600. Robot vacuum cleaner companies should develop products that "bridge cutting-edge technologies and mainstream price points to accelerate adoption," said Jin Liu, senior analyst of small appliances at Euromonitor International. But even if the price comes down, it would only be a small step toward having a robot help with cooking and other household chores. Vacuum cleaners are the "only successful application [of robots] in our homes to date," said Jeff Burnstein, president of the Association for Advancing Automation (A3). "This is after four decades of talking about how we're going to have robots in our home." "What made the [robot] vacuum cleaner so successful is it didn't cost that much," he said. For the same thing to happen for humanoids to enter homes, he said, there needs to be a compelling quality for the price. Humanoids, such as those from Chinese startup Unitree, still cost tens of thousands of U.S. dollars and don't have clear household use cases yet. Despite its mass market ambitions, Roborock said that because of tariffs, it had to raise the Saros Z70 price by $700 from the original $1,899. Quan said Roborock started working with suppliers late last year in Vietnam, where he said the company can fulfill all its North American orders. Looking ahead, he said the company is considering global supply chain partnerships, but not necessarily to invest in building its own factories. Roborock's plans for a Hong Kong listing are primarily to raise capital for international expansion, Quan said, noting the company is also expanding beyond vacuums. Despite Roborock's 79% revenue surge, the company more than doubled its spending, largely on research and development, steepening the company's losses in the first half of the year. Quan said the company has hired nearly 100 AI experts this year and is still hiring — with an eye to add a total of nearly 200 AI experts this year. He said many of the new hires have overseas education or work experience. The company built a dedicated AI lab in Shanghai and a research institute in Shenzhen, soon after Roborock's founding in 2014. When asked about computing power, Quan said there are many solutions and that buying Nvidia chips aren't the only option. As for improving the AI-powered robotic arm and making it cheaper, "the challenge lies primarily with the algorithm and data," he said, not the hardware. As AI becomes more critical for household robots, Quan has an even bigger vision. "If this robot in your home needs to clean, then it will have to integrate the cleaning knowledge that Roborock has accumulated over the years in algorithms, models, data and training," he said. "Then it can be installed onto the robot like an app." "This robot may be Tesla's, or Unitree's, or someone else's, ... but in the area of cleaning, it will be inseparable from Roborock," he said, claiming the company has the best data on cleaning tasks. Another company might have the best data for robots to cook, he said. The humanoid market will likely reach $5 trillion by 2050, with $800 billion in China alone, according to Morgan Stanley estimates. "With humanoids, if they can't do more than one thing, then they're competing against an existing form factor that can do one thing very well," Burnstein said. But ne noted companies around the world expect there's a big market for safe, affordable humanoids that can cook, clean, help the elderly and do other things. "We're not there yet with the technology, but maybe we'll get there and maybe that multitasking would be the differentiator potentially," he said. "So you wouldn't need 5 robots. You might just need one."
