This AI knew the answers but didn’t understand the questions

Science News from research organizations This AI knew the answers but didn’t understand the questions Date: April 30, 2026 Source: Science China Press Summary: For decades, psychologists have debated whether the human mind can be explained by one unified theory or must be broken into separate...

Science News from research organizations This AI knew the answers but didn’t understand the questions Date: April 30, 2026 Source: Science China Press Summary: For decades, psychologists have debated whether the human mind can be explained by one unified theory or must be broken into separate parts like memory and attention. A recent AI model called Centaur seemed to offer a breakthrough, claiming it could mimic human thinking across 160 different cognitive tasks. But new research is challenging that bold claim, suggesting the model isn’t truly “thinking” at all—it’s just memorizing patterns. Share: Facebook Twitter Pinterest LinkedIN Email FULL STORY A cutting-edge AI model that appeared to mimic human thinking may actually just be memorizing answers. New tests reveal it struggles with true understanding, exposing a major gap in today’s AI systems. Credit: AI/ScienceDaily.com Psychologists have long debated whether the human mind can be explained by a single, unified theory or if different functions such as attention and memory must be studied separately. Now, artificial intelligence (AI) is entering that debate, offering a new way to explore how the mind works. In July 2025, a study published in Nature introduced an AI model called "Centaur." Built on standard large language models and refined using data from psychological experiments, Centaur was designed to simulate human cognitive behavior. It reportedly performed well across 160 tasks, including decision-making, executive control, and other mental processes. The results drew widespread attention and were seen as a possible step toward AI systems that could replicate human thinking more broadly. New Research Raises Doubts A more recent study published in National Science Open challenges those claims. Researchers from Zhejiang University argue that Centaur's apparent success may come from overfitting. In other words, instead of understanding the tasks, the model may have learned to recognize patterns in the training data and reproduce expected answers. To test this idea, the researchers created several new evaluation scenarios. In one example, they replaced the original multiple-choice prompts, which described specific psychological tasks, with the instruction "Please choose option A." If the model truly understood the task, it should have consistently selected option A. Instead, Centaur continued to choose the "correct answers" from the original dataset. This behavior suggests that the model was not interpreting the meaning of the questions. Rather, it relied on learned statistical patterns to "guess" answers. The researchers compared this to a student who scores well by memorizing test formats without actually understanding the material. Why This Matters for AI Evaluation The findings highlight the need for caution when assessing the abilities of large language models. While these systems can be highly effective at fitting data, their "black-box" nature makes it difficult to know how they arrive at their outputs. This can lead to issues such as hallucinations or misinterpretations. Careful and varied testing is essential to determine whether a model truly has the skills it appears to demonstrate. The Real Challenge: Language Understanding Although Centaur was presented as a model capable of simulating cognition, its biggest limitation appears to be in language comprehension. Specifically, it struggles to recognize and respond to the intent behind questions. The study suggests that achieving true language understanding may be one of the most important challenges in developing AI systems that can model human cognition more fully. RELATED TOPICS Mind & Brain Intelligence Language Acquisition Child Development Literacy Computers & Math Statistics Artificial Intelligence Computer Modeling Communications RELATED TERMS Evolutionary psychology Limbic system Robot Brain Thought Memory Computing Attention Story Source: Materials provided by Science China Press. Note: Content may be edited for style and length. Journal Reference: Wei Liu, Nai Ding. Can Centaur truly simulate human cognition? The fundamental limitation of instruction understanding. National Science Open, 2025; 5 (1): 20250053 DOI: 10.1360/nso/20250053 Cite This Page: MLA APA Chicago Science China Press. "This AI knew the answers but didn’t understand the questions." ScienceDaily. ScienceDaily, 30 April 2026. . Science China Press. (2026, April 30). This AI knew the answers but didn’t understand the questions. ScienceDaily. Retrieved July 13, 2026 from www.sciencedaily.com/releases/2026/04/260429102035.htm Science China Press. "This AI knew the answers but didn’t understand the questions." ScienceDaily. www.sciencedaily.com/releases/2026/04/260429102035.htm (accessed July 13, 2026). Explore More from ScienceDaily RELATED STORIES Scientists Discover How the Brain Rewires Itself to Truly Multitask July 12, 2026 — Practice may do more than make perfect. Researchers found that extensive training physically reorganizes the brain, allowing learned tasks to bypass the prefrontal cortex and run through specialized ... Intelligence Emerges When the Whole Brain Works as One Mar. 3, 2026 — For decades, scientists have mapped attention, memory, language, and reasoning to separate brain networks — yet one big mystery remained: why does the mind feel like a single, unified system? ... Why GPT Can't Think Like Us Feb. 21, 2025 — Artificial Intelligence (AI), particularly large language models like GPT-4, has shown impressive performance on reasoning tasks. But does AI truly understand abstract concepts, or is it just ... Largest Study Ever Done on Cannabis and Brain Function Finds Impact on Working Memory Jan. 28, 2025 — A new study explores the effects of both recent and lifetime cannabis use on brain function during cognitive tasks. The study, the largest of its kind ever to be completed, examined the effects of ... 'Alzheimer's in Dish' Model Shows Promise for Accelerating Drug Discovery Nov. 27, 2024 — A decade ago, researchers introduced a new model for studying Alzheimer's disease. Known as 'Alzheimer's in a dish,' the model uses cultures of mature brain cells suspended in a ... Robots Predict Human Intention for Faster Builds Apr. 5, 2023 — Humans have a way of understandings others' goals, desires and beliefs, a crucial skill that allows us to anticipate people's actions. Taking bread out of the toaster? You'll need a ... TRENDING AT SCITECHDAILY.com Astronomers Find a New Clue for Detecting Runaway Supermassive Black Holes Why You Hate Eating Bugs: DNA Reveals a 9,000-Year-Old Legacy Coffee May Protect the Liver in More Ways Than Scientists Realized The Hidden Health Risk of Sitting for Hours Without a Break

Source: ScienceDaily AI — Published April 30, 2026 at 06:44 UTC — Category: Research