Will generative AI make us brainless?
For some time now, I’ve been thinking about the impact of “intensive” or “uncontrolled” use of generative AI tools on our cognition, in the same way as our overuse of social networks. I’m particularly worried about our children. We could be on the verge of creating a generation of idiots…
I’m worried about the impacts of Generative AI tools like ChatGPT on our brain. More precisely, the impact of abusive (uncontrolled) use of these tools on our cognitive abilities, that would be facilitated by the universal law of least effort and easy access, like our current overuse of social networks.
My thinking lies in the idea of applying the precautionary principle before it is too late [1]. I'm afraid we are on the verge of creating a generation of idiots... I also see it as a threat to our creativity.
The dose makes the poison and also the period of exposure. Generative AI tools used in moderation are probably a good thing, it’s the excess that worries me.
However, neurosciences are clear: « We become what we do with our brain. » The brain is affected by the way we use it. It is hardly far-fetched to expect that intensive AI tools use will change the human brain due to neuroplasticity [2], [3].
Asking an AI tool to solve problems or write texts for us may be convenient, but it short-circuits our brain's learning, bypassing the neural development that would accompany mastering the mental processes involved in solving problems or writing texts. Only us can create and strengthen our neural connections, and only through our own intellectual work [4].
For instance, it has been experimentally demonstrated that the so-called “intensive” use of GPS reduces the number of neurons and connections in our hippocampus giving hippocampal atrophy [5] . At the opposite, this explains why London taxi drivers have an overdeveloped hippocampus, because to pass the exam to become a driver, they must know London by heart [6].
This is likely to be particularly harmful for children whose brains are still in development. Ultimately, children who are overly exposed to AI tools risk suffering from underdevelopment of their neural networks.
The risk should be lower for adults who already have a mature brain with good cognitive abilities, but the risk remains that an intellectual laziness will set in. Also by virtue of the neuroplasticity of our brain, the lack of appropriate exercises could lead to an erosion of intellectual faculties. In a nutshell, this is easy to imagine the potential havoc in a working environment, where combining the universal law of least effort, the pressure to be productive, and the achievement of rapid results, without any regard for the workers' brains.
Some techno-optimists suggest replacing writing texts or solving problems with thinking critically about the solutions proposed by AI tools. But these people seem to ignore that exercising critical thinking requires solid prior knowledge and good intellectual skills. There is no critic worthy of the name without having a solid intellectual background, knowledge of the subject and analytical skills. These are prerequisites for critical thinking.
Other techno-enthusiasts make comparisons, let's say lame, with the use of a calculator. In the worst case, calculator abuse will create arithmetic illiterates [7]. But a calculator is an extremely limited tool when compared to a Generative AI tool like ChatGPT. It’s a bit like pitting a slingshot against an atomic bomb. Also, problem solving and writing texts are much more important activities for the development or maintenance of our cognitive abilities and thinking. It is more difficult to deprive yourself of the ability to solve problems or write texts. In fact, we can't do without thinking by ourself.
As an aside, I'm completely neglecting to discuss the capabilities of drawing and composing music which are also threatened by tools like DALL-E or Midjourney for images and AIVA or Lyria for music.
In education, thanks to techno-optimists, we are moving towards a Kafkaesque scenario, where students answer questions and do their homeworks with ChatGPT. Which assignments and questions will be prepared and evaluated by teachers using ChatGPT. As a skill, we will have to learn to write the best request (prompt) describing what we want ChatGPT to do for us. In part, to make better “pattern recognition”. Eventually this will be less and less necessary, as ChatGPT’s performance improves and with all the data it collects about our needs and preferences.
This is why it would be better to be very careful with generative AI tools. The use of generative AI tools like ChatGPT by children should be limited until basic cognitive skills are well acquired, in addition to essential knowledge. Let's say, not before post-secondary studies.
Today, our abilities to change the world are gigantic compared to the time of the invention of the printing press or the calculator, but this great power must be accompanied by great caution. It's all well and good to bequeath new challenges to future generations, but please avoid leaving them a legacy of cognitive disabilities because of our lack of wisdom.
Call for all
Finally, I'd like to make an appeal to everyone. If you know of other researchers in neuroscience, cognitive science, AI, philosophy, psychology or education who share my concerns and would like to deepen the argument, debate it more widely, conduct experiments to prove of disprove the possible damages and possibly alert public opinion to the subject [8], I'm ready to help.
That said, I have no illusions. The environmental crisis taught me hard the thankless role of Cassandra. I've contacted several researchers, some of them very well-known. Without denying the problem, they haven't lifted a finger. No doubt, they are very busy. Maybe I'm starting too early, people prefer to wait until it's too late...
Some people who share all or part of my concerns [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19].
Notes and references
[1] Other examples where the application of the precautionary principle would have been appropriate. Knowing the water cycle, it would have been easy to predict that plastics would end up in the ocean and then microplastics in our blood. Also, it was predictable that the massive use of fossil fuels would cause a greenhouse effect through the accumulation of carbon dioxide in the atmosphere and that would alter the climate. We could also have predicted the creation of echo chambers and the massive disinformation caused by social networks.
[2] BARBER, Nigel « Can Artificial Intelligence Make Us Stupid? », blog post, Psychology Today, July 29, 2015.
[3] KORTE, Martin. « The impact of the digital revolution on human brain and behavior: where do we stand? ». Dialogues in clinical neuroscience, 2020, vol. 22, no 2, p. 101–111.
[4] Here I restate an idea from physicist Tom Murphy who expressed concerns about science students wanting recipes in order to solve physics problems in his book « Energy and human ambitions on a finite planet » 2021.
[5] BIRKEL, Lindy. « Decreased use of spatial pattern separation in contemporary lifestyles may contribute to hippocampal atrophy and diminish mental health. » Medical Hypotheses, 2017, vol. 107, p. 55–63.
[6] JABR, Ferris « Cache Cab: Taxi Drivers’ Brains Grow to Navigate London’s Streets », Scientific American, December 8, 2011.
[7] With maybe some difficulties to learn more advanced mathematics
[8] An interview in French, I gave to Radio-Canada Alberta «L’intelligence artificielle et le développement cognitif » — November 29, 2023.
[9] Abdelkrim Alfalah has published a related post about « Cognitive Outsourcing » on LinkedIn.
[10] David Anctil has published articles in French on related subjects, first « Copilot, l’assistant virtuel qui s’apprête à bouleverser le monde du travail » in the pages of Québec Science, and the article « L’éducation supérieure à l’ère de l’IA générative » in the journal Pédagogie collégiale.
[11] Researchers is Education and English Literature from Susquehanna University « This is your brain on ChatGPT: how AI might be affecting your creativity ».
[12] The neurologist David Fortin, from Université de Sherbrooke, in French « Les écrans laissent leur empreinte sur le cerveau des jeunes ». Dr Fortin is also the author of a fascinating popular book in French « Le cerveau, une galaxie dans votre tête ».
[13] This brings grist to my mill… A post in French in UdeMNouvelles where Serge Larivée, professor and researcher specializing in human intelligence, expresses his fear that ChatGPT contributes to the regression of the intelligence quotient (IQ). « ChatGPT nous rendra-t-il plus stupides? »
[14] An interview with AI professor and researcher Laurence Devillers from Université Paris-Sorbonne in the journal L’ADN. « Les IA génératives pourraient provoquer un déclin cognitif »
[15] AHMAD, Sayed Fayaz, HAN, Heesup, ALAM, Muhammad Mansoor, et al. Impact of artificial intelligence on human loss in decision making, laziness and safety in education. Humanities and Social Sciences Communications, 2023, vol. 10, no 1, p. 1–14.
[16] León-Domínguez, U. (2024). Potential cognitive risks of generative transformer-based AI chatbots on higher order executive functions. Neuropsychology. A popularization paper in PsyPost
[17] Meta-analysis 2020 — AL SAGR, Abdullah Nasser et AL SAGR, Nora Abdullah. « The effect of electronics on the growth and development of young children: A Narrative Review ». Journal of Health Informatics in Developing Countries, 2020, vol. 14, no 1.
[18] Second Meta-analysis 2022 — VÁZQUEZ CANO, Esteban, PARRA GONZÁLEZ, María Elena, SEGURA ROBLES, Adrián, et al. The negative effects of technology on education: a bibliometric and topic modeling mapping analysis (2008–2019). 2022.
[19] Third Meta-analysis 2023 — In French — Institut national de santé publique du Québec « L’utilisation des écrans en contexte scolaire et la santé des jeunes de moins de 25 ans : effets sur la cognition », 2023.
