The End of Humanity’s Monopoly on the Brain
We are more predictable than ever
—all thanks to AI
It was believed that no one or no special abilities can accurately predict a person’s behavior and their thoughts. Not even psychics. What broke this long-believed notion was nothing more than statistical learning with a recent technological boom—AI.
How it started was simple text generation based on information on the internet. In fact, it was nothing more than a simple Google search, presented in a coherent manner. It was hardly an intelligence. However, soon the ‘intelligence’ part came into play, where AIs were able to conduct simple reasoning and produce creative content. The striking part was that its development was grounded similarly to how our brains worked.
Our brain is constructed with numerous neurons, which form tight connections to each other. If a specific part of the brain is utilized more, the connectivity of the neurons increases, and the specific cognitive ability strengthens. All of this process is based on learning, exercising specific neurons to activate more robustly. In turn, our abilities strengthen and ideas are created, with combinations of previous knowledge yet unique integration.
A simplified version of our neuronal process was the basis of the AI development. Train them with a wide array of data, learn the pattern, and produce accordingly. Within this same paradigm, Centaur was introduced at Nature in 2025, which is one of the first models to capture the foundational model of human cognition. This model, which was trained with more than 10 million decisions collected from over 60,000 participants across 160 psychological experiments, exhibited a pattern that was not shown before—predicting the subsequent behaviors. Previous research on AI focused on single decisions or behaviors, and explaining the reasons behind them. But this new model approached it otherwise. It didn’t learn the contextual human behavior from each experiment. It learned human behavior itself, marking the first glimpse of artificial technology invading the zone of humans. After sufficient learning, Centaur was not only able to accurately predict behaviors of common settings, but also novel and unique scenarios without any prior exposure.
Even though Centaur’s new experiment provided a significant step forward, it was not the sole research. Previous research has shown that AI models could reproduce the cognitive tendencies of humans in structured experimental settings. Specifically, AI-generated agents exhibited analogous behaviors ranging from anchoring effects, halo effects, in-group favoritism, to false consensus bias. Sometimes, they even replicate their mistakes or distortions. All of these were based on observation of humans, where irrationality changes to rational statistical calculations.
While how uncomfortable these researches may feel, they do highlight two important pieces of information. One, humans are naturally quite predictable creatures, as our actions themselves are learned through statistical probability. Second, AI is becoming closer to its name—artificial intelligence.
