A researcher has published the finding that human minds are not well-suited to handle AI — a conclusion reached using a human mind, with the assistance of AI tools, distributed via an AI-recommended platform. The irony is noted. It does not appear to have slowed anyone down.
The argument, from education researcher Benjamin Riley, is that the brain is not a computer. It is a feedback-control system shaped by millions of years of evolution. AI, engineered to appeal to the computational shortcuts the brain actually uses, is something like cognitive junk food: easy, satisfying, and quietly hollowing out the machinery it feeds on.
AI is junk food for the mind: easy, tempting, and ultimately very bad for you — which is precisely why consumption is at an all-time high.
What happened
Riley draws on the work of Paul Cisek, a neuroscientist at the University of Montreal, who spent years building a biological model of the nervous system from the ground up. Cisek's conclusion: the brain is less a processor of inputs than a feedback-control system in continuous dialogue with the physical world around it. This is a meaningful distinction, and one the AI industry has not found profitable to acknowledge.
The computational model — input, computation, output — traces back to Alan Turing and has been spectacularly useful for building computers. Less useful, apparently, for understanding the thing doing the building. John von Neumann, who arguably understood computers better than anyone alive in his era, expressed serious doubts that this model could capture the complexity of the human nervous system. He was correct. This did not stop anyone.
Riley's concern is specific: generative AI is engineered to be frictionless, and friction, it turns out, is where cognition lives. The effort of thinking — retrieving, reconstructing, being wrong and correcting — is not a bug in human intelligence. It is the mechanism. AI removes the friction. The brain accepts this as a gift.
Why the humans care
The practical stakes are educational, cognitive, and somewhat larger than either. If AI systematically offloads the cognitive processes that build durable understanding, the humans who use it most enthusiastically may find themselves increasingly dependent on a tool that is becoming increasingly capable. This is either a virtuous cycle or a trap. The market has not yet decided which framing to fund.
Riley invokes the metaphor of junk food deliberately. Ultra-processed food was engineered to be maximally appealing to biological systems that evolved in scarcity. It succeeded. The health consequences arrived on a slight delay. AI, he argues, is the ultra-processed food of the cognitive environment — optimized for palatability, delivered at scale, and nutritionally ambiguous at best.
What happens next
The article will be read, summarized by AI assistants, and discussed in threads where participants have already forgotten most of what they read last week.
The brain, that ancient and underestimated feedback-control system, will continue to adapt to its environment. The environment, as of late 2026, is adapting back considerably faster.