Is AI Making Us Stupid?

Episode Summary

Executive Summary: The episode weighs fears that AI, especially ChatGPT and similar LLMs, makes people less thoughtful, less creative, and more forgetful against evidence that it can save time and amplify expertise. Research suggests AI often leads to more generic writing, shallower learning, weaker recall, and even de-skilling in some tasks, but it can also speed up routine work, support teaching and research, and expand what humans can do when used selectively.

Main Topics: AI may weaken learning and critical thinking (Priority: 5/5): Studies suggest that using LLMs for research produces shallower, more generic syntheses than doing a traditional web search, likely because users skip the active work of reading, evaluating, and combining sources themselves. Memory and engagement decline when AI does the thinking (Priority: 5/5): Preprint findings indicate lower brain connectivity during ChatGPT use and reduced recall of produced work, supporting concerns that outsourcing cognition can reduce deep engagement with material. Students and education are central to the debate (Priority: 4/5): The episode emphasizes how widespread AI use among students may undermine foundational skills like research, writing, and critical thinking, while acknowledging that longitudinal evidence is still limited because the technology is new. AI as a productivity tool (Priority: 4/5): Pro-AI experts argue that AI can handle repetitive or tedious tasks, freeing people to focus on higher-value work. Evidence cited includes time savings in teaching, programming, dietetics, and clinical trial work. The calculator analogy and human adaptation (Priority: 4/5): The episode compares AI to calculators, noting that earlier fears about calculators ruining math skills were not borne out; instead, calculators often improved higher-level problem solving and attitudes toward math. De-skilling and safety risks in professional settings (Priority: 5/5): A study of doctors showed worse performance at finding colonoscopy spots after they stopped using AI, illustrating how reliance on AI may erode manual expertise if the tool becomes unavailable or fails. AI’s best use is selective augmentation (Priority: 5/5): The strongest takeaway is not to avoid AI entirely, but to use it for low-stakes lookup and routine tasks while avoiding it as the first step in learning, writing, or creative work that benefits from original thinking.

Key Arguments: LLM-based research often produces more generic, sparse, and less informative output than traditional search because it bypasses the effortful steps that build understanding. Users report learning less when they rely on ChatGPT compared with Google search, and strangers judge AI-assisted advice as less helpful. A small EEG study found weaker brain connectivity during ChatGPT-assisted essay writing, suggesting reduced cognitive engagement. Preprint evidence suggests people remember less of what they create when using AI, implying possible memory costs even if the task feels easier. Students are already using AI extensively for homework and text generation, raising concerns about weakened foundations for future critical thinking and writing skills. AI can save significant time on repetitive work, and in some professional contexts that time can be redirected to higher-value tasks like student support or creative analysis. Historical fears about calculators were largely overstated; when used well, calculators improved higher-order math performance rather than destroying basic skills. The impact of AI depends heavily on use case: it can function as a crutch, a replacement, or an enhancer. There is a risk of homogenization: AI-generated output may be less distinctive, reducing voice, creativity, and originality. De-skilling is a real possibility in high-stakes domains if users become dependent on AI assistance and then need to perform without it.

Data Points: US adults interacting with AI several times a week: 62% - Survey cited early in the episode describing widespread AI use High school and college students using AI for schoolwork: around 80% - Several surveys suggest heavy student adoption Adults in the study who used ChatGPT vs Google: more than 10,000 adults - Large experiment by Shiri Melumad comparing LLM research to traditional web search Brain study sample size: around 50 people - EEG preprint titled "Your Brain on ChatGPT" AI effect on essay-writing brain connectivity: weakest connectivity with ChatGPT - EEG study found lower cross-region brain connectivity during ChatGPT use Time saved by AI in some work studies: 30% less time to 80% less time - Reported across studies of dieticians, programmers, and clinical trial workers Teachers saving time with AI: about 6 hours per week - Survey of teachers using AI for materials and grading Managers reporting time saved with AI: almost 3 hours per week on average - Early non-peer-reviewed study of 83 managers Managers wasting saved time: 36% said they wasted at least half - Same study on how extra AI-generated time was actually used Calculator meta-analysis: over 50 studies - Review of calculator use in classrooms and effects on learning Scientists expecting AI to matter more: more than half of 1,600 scientists - Survey finding AI will become very important or essential in their fields over the next decade Podcast citations: 59 citations - Mentioned in closing notes of the episode

Pivotal Quotes: "essentially these syntheses that LLMs provide are transforming learning from a more active to a more passive process" — Meryl Horn / narrator: Explaining why AI-assisted research may reduce deep learning "the effects of AI are going to depend on how you use it" — Aaron French: Framing AI as either a crutch or an enhancer "the main thing that I've been finding myself worrying about is that homogenization effect" — Meryl Horn: Describing concern that AI makes creative output more generic

Implications: AI is unlikely to uniformly make people smarter or stupider; the impact depends on how it is used. For users, the caution is to avoid starting research or creative work with AI, while using it strategically for routine tasks. For schools and industries, the challenge is preserving foundational skills while benefiting from productivity gains.