The Verge argues that contemporary AI functions like mental junk food: instantly appealing yet potentially harmful to long-term cognitive health. Human brains have been sculpted by millions of years of evolution, developing complex systems for navigating the physical world and social environments. When AI tools are consumed habitually, they may interfere with the very processes that sustained our species.
The traditional view treats the mind as a computer, tracing progress from simple calculators to neural networks that mimic human thought. John von Neumann, a pioneer of computer science, warned that such a computational metaphor cannot capture the extraordinary intricacy of the nervous system. Neuroscientist Paul Cisek of the University of Montreal proposes a feedback-control perspective, illustrated by an outfielder who simply keeps the ball steady in the visual field rather than performing hidden calculations.
Feedback-based models align with the evolutionary record, which shows that as vertebrates gained mobility they evolved specialized navigation circuits that later became the hippocampus. This structure supports the storage of episodic memories, allowing organisms to recall specific journeys and outcomes. Cisek argues that such mechanisms illustrate how brains continuously adjust behavior in response to environmental cues, a process the computational paradigm tends to overlook.
Human cognition is also shaped by profound social feedback loops that began when ancestors mastered imitation, enabling the transmission of tool-making techniques and coordinated rituals. The emergence of oral language amplified this capacity, while the shift to settled agriculture introduced abundant high-fat foods that our bodies were not evolved to store in excess. The article likens this dietary transition to the current surge of AI, both representing shortcuts that can erode adaptive systems.
The piece characterizes AI as a 'cognitive hot dog',tempting in the moment but capable of clogging the mental arteries that support deep learning. It warns that sporadic interaction may be harmless, yet habitual reliance could diminish the brain’s ability to synthesize knowledge and navigate complex problems. This framing suggests that the danger lies not in the technology itself but in its integration into daily cognitive diets.
In a July 2025 podcast on AI’s impact on education, OpenAI’s vice president of education Leah Belsky claimed that learners comprised more than half of the platform’s 900 million average monthly users, describing ChatGPT as the world’s largest learning platform. Anthropic has responded with a program called Claude for Teachers, which supplies educators with free access to its premium models. Both initiatives aim to embed generative AI deeper into classroom practice.
Former Google engineer François Chollet has labeled this trend ‘cognitive automation,’ defining it as the encoding of human abstractions into software that then performs tasks traditionally done by people. An interdisciplinary consortium of scholars recently argued that large language models should be seen primarily as cultural and social technologies that amplify collective information, rather than as autonomous intelligent agents. The article concludes that understanding AI through feedback-oriented models may better reflect its true influence on human behavior.