The Cognitive Risks of AI: Dependence, Brain Activity Decline, and the Threat of Passive Use

The video addresses the cognitive risks associated with the use of artificial intelligence (AI), particularly language models (LLMs) like ChatGPT. It presents a fictional scenario set in 2029 where a military AI triggers a nuclear war, then shifts the discussion to a more immediate danger: cognitive dependence. By 2025, 80% of French high school students reportedly use LLMs for their homework, leading to a measurable decline in brain activity as recorded by electroencephalography (a study by MIT in June 2025). Current LLMs (GPT-4o, Gemini 3) offer multimodal functionalities and advanced document-processing capabilities, but their passive use reduces mental effort, akin to an amplified "Google Effect." A survey reveals that senior developers use AI as a productivity tool, while junior developers become dependent on it. A Harvard study shows that a well-structured AI tutor (PS2 PAL) improves physics performance, but its effectiveness depends on the user's willingness to engage. The video concludes that AI itself does not make people stupid, but unchecked usage can atrophy cognitive abilities.