We’ve been told that many things increase our risk of dementia — genetics, too much alcohol, not enough exercise, improper nutrition, high blood pressure, the risks go on and on.
Neuroscientist Vivienne Ming wants to add one more item to the list — artificial intelligence.
Scientists have already sounded the alarm that US dementia cases could nearly double by 2060, thanks to our aging population and rising rates of obesity, diabetes and hypertension.
Now, Ming is warning that AI could contribute to a “dementia crisis” as it weakens the brain systems responsible for curiosity, attention, high-order reasoning and executive function, among other duties.
“My own data shows that students using AI in the most common way — asking it questions and accepting the answers — show more than a 40% reduction in the gamma-band brain activity that indicates active cognitive engagement,” Ming, author of the new book, “Robot-Proof: When Machines Have all the Answers, Build Better People,” told The Post.
“Their brains are measurably less active than when they work without AI.”
Ming describes what an AI-powered dementia crisis could look like — and shares four early warning signs that suggest overreliance on AI.
How can AI affect cognition?
A survey last year revealed that 56% of US adults use AI tools, with 28% using them at least once a week.
Seeking information or quick answers is one of the top functions. For people who use AI this way, Ming said, changes to cognition are not immediately noticeable but build over time.
“When the answer is always one tap away, we stop developing the habit of wondering,” she explained.
“Without errors to drive learning, our brain’s reward circuits stop responding to mystery. We short-circuit the parts where wondering becomes exploration.”
Metacognition is also affected. That’s the awareness and regulation of your own thinking, also known as “thinking about thinking.”
Ming describes it as “testing your own understanding against reality, experiencing the gap and updating.”
When AI does that work for you, you get the illusion of understanding rather than true skill acquisition.
“We need ‘productive friction,’” Ming said, “the challenges that make life a little harder and thereby make us better.”
How certain are we that a crisis is on the horizon?
Many innovations over the years have changed the way we think — calculators and smartphones are just a few — but nobody blamed calculators for a dementia crisis.
GPS navigation is the closest analogy to AI, Ming said.
Habitual GPS use has been shown to diminish spatial reasoning by reducing activity in the brain region responsible for forming mental maps — the hippocampus.
“While GPS offloads one cognitive function, AI is designed to offload any cognitive function you’re willing to give it: writing, reasoning, planning, synthesis, judgment,” Ming noted.
“The substitution is nearly total and it’s self-reinforcing, because AI that’s better than you at something creates a temptation to stop practicing that thing, which makes you worse at it, which makes you more dependent on the AI.”
Experts have long recommended seniors engage in stimulating work, read and learn new skills to strengthen their cognitive reserve, potentially staving off dementia.
If there is an AI-fueled dementia crisis, Ming said it “would look like a gradual, population-level compression of cognitive reserve, the brain’s accumulated resilience against age-related decline.”
She acknowledged that AI tools haven’t been around long enough to measure long-term impact, “but the mechanistic evidence is there, the short-term behavioral evidence is there, and anyone claiming we need more proof before taking this seriously is applying a standard of certainty they don’t apply to anything else they care about.”
Can the brain positively adapt to AI use?
You want to use AI, but you also want to preserve your cognitive reserve.
The best way would be to have the AI act as an “adversarial collaborator” that does not blindly agree with your inputs, Ming suggested.
The AI would challenge your reasoning, identify weak evidence and force you to refine your arguments.
It doesn’t work if you “use AI the way most people currently use it — as a magic genie that provides answers,” Ming said.
Early warning signs that can suggest AI dependence
- You struggle to start a document without AI generating a draft first.
- You can follow a complex argument as you read it, but can’t reconstruct it an hour later.
- You can’t keep up with a long book without frequent summaries.
- You express confidence about skills you can’t demonstrate if tested.
“These are the signs of a monitoring loop that has stopped running,” Ming said.
