What 81,000 people want from AI
Anthropic’s “81,000 interviews” piece on what people want from AI is met with skepticism: many see it as a glossy, front-end–heavy marketing exercise based on self-selected, already-positive users and vague categories like “professional excellence.” Commenters highlight moving individual stories of AI helping with learning, entrepreneurship, and even medical or emotional crises, but worry these are survivorship-biased anecdotes used to sell a technology that may worsen job insecurity, overtesting in medicine, and dependence on chatbots for friendship or therapy. There is broad concern about who captures the productivity gains from AI—workers vs. shareholders—and a recurring call for independent, academically rigorous research on AI’s social impact rather than company-authored narratives.
Website and UX
- Many complain the Anthropic page is extremely heavy, slow, and CPU-intensive, especially on phones; several describe frame drops and fans spinning up.
- Blame is placed on “vibe-coded” / Next.js-style frontend bloat, telemetry, and animations; some see it as emblematic of modern web regressions.
- A PDF mirror is shared as a workaround.
Survey Design and Sample Bias
- Several note the respondents are active Claude users prompted to give positive visions first, so hostile or “I want AI to go away” views are structurally excluded.
- Headline categories like “professional excellence” and “personal transformation” are criticized as vague, generic, and marketing-friendly.
- Some compare this to classic corporate “thought leadership” pieces; calls are made for independent academic analysis instead.
- Survivorship bias is raised: the study tells us what satisfied users say, not what the broader or AI-averse public wants.
Interpretation of Quotes and Use Cases
- Many find the individual stories the most interesting part: acceleration of research, emotional support in war zones, educational help for kids, assistance in self-understanding and health advocacy.
- Others are uneasy: AI as quasi-friend/therapist, emotional dependence, medical prompting leading to over-testing and false positives, or people outsourcing personal creativity and care.
- Some dramatic claims (e.g., years of work compressed into weeks) are met with skepticism about realism and user competence.
Economic and Social Impact
- Strong concern that current AI deployment is framed as labor replacement, benefiting shareholders more than workers or consumers.
- Debates over whether productivity gains will raise wages for skilled workers or simply increase exploitation and precarity.
- Suggestions include focusing on societal supports (healthcare, housing, safety nets) rather than only higher dev salaries.
AI Capabilities, Risks, and Communication Style
- Desire for AI that can say “I don’t know” and firmly correct users, not just generate fluent answers.
- Some think Claude is relatively better at avoiding “bullshit,” but unreliability remains a top concern.
- Multiple commenters suspect that parts of the article and even user quotes were edited or AI-written, citing stylistic “tells” and fluffy, content-light prose.
- Overall tone: mix of fascination with real benefits and deep skepticism about marketing spin, methodology, and long-term societal effects.