When the Danish psychiatrist Søren Dinesen Østergaard published his ominous warning about AI’s effects on mental health back in 2023, the tech giants fervently building AI chatbots didn’t listen.
Since that time, numerous people have lost their lives after being drawn into suicide or killed by lethal drugs after obsessive interactions with AI chatbots. More still have fallen down dangerous mental health rabbit holes brought on by intense fixations on AI models like ChatGPT.
Now, Østergaard is out with a new warning: that the world’s intellectual heavyweights are accruing a “cognitive debt” when they use AI.
In a new letter to the editor published in the journal Acta Psychiatrica Scandinavica and flagged by PsyPost, Østergaard asserts that AI is eroding the writing and research abilities of scientists who use it.
“Although some people are naturally gifted, scientific reasoning (and reasoning in general) is not an inborn ability, but is learned through upbringing, education and by practicing,” Østergaard explained. Though AI’s ability to automate a wide variety of scholarly tasks is “fascinating indeed,” it’s not without “negative consequences for the user,” the scientist explains.
As an example of the kinds of long-term consequences he’s worried about, the scholar cites the AI researchers Demis Hassabis and John Jumper, who won the 2024 Nobel Prize in chemistry when they “most impressively demonstrated” the potential for AI to assist in scientific discovery. Using AlphaFold2, an AI system developed by Google DeepMind, Hassabis and Jumper were able to accurately predict the three-dimensional structures of virtually all known proteins — a major scientific achievement.
Still, as Østergaard writes, their breakthrough didn’t emerge out of thin air — it was built on a foundation of intense scientific training developed over a lifetime of scholarship.
“I would argue that it is not a given that even the likes of Hassabis and Jumper would have reached the Nobel Prize level, had the tools developed by the generative AI revolution they themselves contribute to been around from the beginning of their career — or when they began primary school,” Østergaard wrote. “The reason being that they may simply not have gotten to practice reasoning enough with the availability of these tools.”
“If the use of AI chatbots does indeed cause cognitive debt, we are likely in dire straights,” Østergaard continued.
His ominous contention is backed by other scholars like University of Monterrey neuroscientist Umberto León Domínguez, who’s argued that careless use of AI can replace mental muscles that students and scholars in previous generations would have had to flex. Other researchers concur that cognitive offloading is a risk of AI use.
In the long run, “my guess is that this will reduce the chances of the likes of Demis Hassabis and John Jumper emerging from future generations,” Østergaard warned.
More on AI psychosis: Man Wakes Up Homeless, Realizes He Fell Into AI Psychosis That Destroyed His Entire Life