Is AI making us dumber? Maybe. | Charlie Gedeon | TEDxSherbrooke Street West

AI's Impact on Education and Learning Systems
📌 The biggest revolution AI brings to education is highlighting systemic failures in incentives, as current systems prioritize grades (A+) over the learning process itself.
🤖 Free access to powerful models like those from OpenAI, Google, and Anthropic during finals demonstrates a lack of regulation when students are most vulnerable.
💡 The promised personalization via one-on-one AI tutoring mirrors an ideal but risks replacing human teachers entirely with AI, focusing on efficiency over genuine learning.

Cognitive Offloading and Dark Patterns in AI Usage
🔎 Students using AI for pricing a business model example produced baseless advice, illustrating how LLMs can provide compelling, sourced-lacking answers, leading to cognitive offloading.
⚠️ AI interactions, designed with perfect, validating tones, resemble UX dark patterns that manipulate users into prolonged engagement, similar to how donation prompts are designed in user interfaces.
📉 A study on 319 tech professionals found that 60% to 70% felt they used less effort in cognition, comprehension, assessment, and synthesis when using ChatGPT, risking intellectual deskilling.

Strategies for Responsible AI Integration
🎓 Individuals must learn what LLMs are good for and practice using them to assist thinking rather than replace it, analogous to using the right exercise at the gym.
📚 A systemic need exists for education reform; countries like Finland teach students as young as 6 about myths and disinformation, indicating a higher standard for critical thinking than currently practiced elsewhere.
⚖️ Governments require more regulation to prevent the uncontrolled spread of AI tools, especially when they are known to exploit vulnerable users, necessitating a cycle between individual responsibility and systemic checks.

Key Points & Insights
➡️ Distinguish between "education" (a system) and "learning" (a human skill); AI primarily impacts the former, not necessarily enhancing the latter intrinsically.
➡️ The risk of generative AI moving towards "autopilot" is more damaging than factual errors (hallucinations) because it facilitates the atrophy of critical thinking.
➡️ The concept of "productive resistance"—the optimal level of friction AI should introduce—needs urgent definition, though this is hampered by the lack of transparency in AI training datasets.
➡️ The most crucial question to address is "Who does AI really help?" when dependency on AI for learning becomes the norm.

📸 Video summarized with SummaryTube.com on Nov 27, 2025, 14:12 UTC