The Memory Crisis: How AI Is Changing Our Brains
We can't remember phone numbers anymore. That's just the beginning. How outsourcing cognition to AI reshapes memory, attention, and deep thinking.
My grandmother knew her entire extended family's phone numbers by heart. Every birthday. Every address. Today, I can barely recall my own mobile number. This is not a failure of memory. It is an adaptation — a reallocation of cognitive resources away from information storage because external storage became cheap and reliable.
The question is whether all cognitive reallocation is equally benign.
What memory actually is
Memory is not a filing cabinet. It is an active, reconstructive process that shapes how we think, what we notice, how we make connections, and who we are. The memories we consolidate and maintain are not neutral records. They are the substrate of reasoning — the raw material from which we generate new ideas by recombining existing ones.
When we outsource memory to AI, we are not just freeing up storage space. We are changing the composition of our mental substrate. We remember less, which means we have less material for the generative, associative thinking that produces insight.
The attention crisis underneath
Perhaps more serious than the memory shift is the attention shift. Memory requires attention — you cannot remember what you never truly attended to. As AI generates information faster than we can process it, the tendency is to skim, scan, and summarize rather than read deeply. The result is a layer of superficial familiarity with many topics and genuine depth in very few.
Deep thinking — the kind that produces genuinely novel work — requires sustained attention over time. It requires holding a problem in mind long enough for non-obvious connections to emerge. This process is slow, uncomfortable, and entirely incompatible with the attention patterns that AI-accelerated information consumption encourages.
The research we should be watching
We are, collectively, running a large-scale uncontrolled experiment on human cognition. The results will not be clear for years, possibly decades. But early indicators — declining performance on certain memory tasks, reduced reading comprehension depth, increasing difficulty with extended concentration — suggest that the effects are real, even if their magnitude and permanence are not yet established.
This is not an argument for limiting AI use. It is an argument for intentional AI use — for maintaining deliberate practices that exercise memory, attention, and deep thinking, as a conscious counterweight to the cognitive tendencies that AI-abundant environments naturally encourage.