The internet made us stupid. AI promises to make it worse

Floating along on my bicycle on a daydream of a country road, up behind me came a man on a hissing e-bike, going fast, headed somewhere important. I thought at that speed — 25 miles an hour, I estimated — and with dusk gathering, he risked hitting a bear that would promptly eat him and the bike. But no such luck.

I saw him later at the crest of the hill that I’d climbed. He whizzed in circles and headed back down while I rested. The hill had been a tough one for me. As you weaken with dependence on the machine, I muttered to myself, I grow stronger. There is one certainty in the cycling world: Trad bikers will outlive e-bikers, who are fools to give up the physical benefits, the spiritual joys, the liberty and independence of a human-powered mechanism.

Our machine dependence, of course, is growing at an exponential rate, as AI comes into wide usage. If the internet, per author Nicholas Carr, has made us stupid, AI promises to make us even stupider. Carr has argued, correctly, that with its endless distractions and fragmented structure, its flashing rabbit-holes, its emphasis on speed and constant switching (between subjects, links, pages, images, etc.), the internet causes cognitive damage, a rewiring of the brain so that we’re less able to ponder and meditate, to think at length and complexly — to go deep. His 2010 book “The Shallows” remains the most important inquiry into technological immanence and its consequences since Neil Postman’s “Technopoly” (1992).

Now comes AI, and we face a new disaster for human cognition, as the prospect of artificial intelligence-induced imbecility is confirmed in study after study of the use of large-language models such as ChatGPT.

One researcher looked into the “future of critical thinking” in an LLM-saturated environment and found “significant negative correlation between frequent AI tool usage and critical thinking abilities.” The breakdown of critical thinking was due to the obvious factor of the increase in “cognitive offloading” that apps like ChatGPT afford. Instead of staying fit doing hard work, the brain “muscle” atrophies as it allows the machine to carry the load. There were echoes of Carr in the conclusion to the study, which noted that AI dependency can “diminish users’ engagement in deep, reflective thinking processes.” Younger people were found to be particularly vulnerable, exhibiting “lower critical thinking scores compared to older participants.” Kids, teens and young adults, in other words, are the most endangered by the technology. (Think about that, you adults who are making money off peddling and proselytizing tech — you are hurting children.)

A team from Microsoft and Carnegie Mellon University published similar findings in February. Use of “generative AI,” i.e. large language models, can “inhibit critical engagement with work,” diminish skill-sets “for independent problem-solving” and — this should be so obvious it need not be said — lead to “long-term overreliance on the tool.” In this telling, AI is a brain-rotting narcotic; the heavier the use, the greater the addiction, the more damage done.

Finally, there’s the MIT study published in June, titled “Your Brain on ChatGPT.” Tasked with writing an essay, study participants were broken into three groups: brain-only, those who completed the essay without machine aid; search-engine users, who were allowed access to Google or the like to supplement the writing; and large-language model users, who were free to have ChatGPT do the heavy lifting. During the writing of the essays, the researchers measured “brain connectivity” with electroencephalography. Their conclusion: “Brain-only participants exhibited the strongest, most distributed networks; Search Engine users showed moderate engagement; and LLM users displayed the weakest connectivity.”

Cognitive activity dropped with each increment of increased use of the machine. To reiterate: The more dependent on tech for crafting your thoughts, the lower your mental performance, the stupider you get. And the stupidifying endures over time. “Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels. These results raise concerns about the long-term educational implications of LLM reliance.”

Between February and April, the number of ChatGPT users worldwide increased from 400 million to 800 million (and this from 50 million in January 2023). In the U.K., the proportion of students who did not use ChatGPT or other LLMs collapsed from 47% last year to 12% this year — only a tenth of students surveyed were holding out against the machine. By the middle of 2024, almost 90% of students at Harvard were employing LLMs for their studies. More than 70% of American adults report regularly using AI, and a third say they use it every day.

Our cognitive disaster is unfolding as if it’s the way of the world, like sun-up over the horizon and the light that blankets the land — inevitable, irreversible, the nature of things. This may be so, given that bowing before the tech god is inherent in industrial civilization, an expression of the fanatical worship of innovation that goes unquestioned in our society.

Meanwhile, I am reminded that the rider of the e-bike who passes us by with his artificial speed is day by day turning into a flabby mess. The few who retain their strength and their independence, who refuse to bend the knee to the machine, will only grow more powerful, more intelligent — perhaps to win a Darwinian race in the long run, when the feeble-minded and servile dependents on AI will be eliminated after the machine system fails.

Christopher Ketcham is writing a book on environmental revolt against industrialism. He is the author, most recently, of “This Land: How Cowboys, Capitalism, and Corruption Are Ruining the American West.”