Overview of Why people really hate AI (The Vergecast)
This episode of The Vergecast (hosts David Pierce and Nilay Patel) examines a growing cultural and commercial problem: mainstream distrust and dislike of AI. The conversation is sparked by an internal OpenAI memo (Fidji Simo) urging a pivot to enterprise use cases, and ranges across polls showing negative public sentiment, VC and founder messaging strategies, concrete product controversies (NVIDIA’s DLSS 5, Samsung’s Z Trifold), debunked viral AI claims, and policy/First Amendment concerns tied to FCC rhetoric. The hosts argue the core issue is simple: AI companies are asking for huge resources and permissions without delivering consumer products people genuinely love.
Main arguments & framing
- AI firms have not yet produced a clear, widely-loved consumer product comparable to the internet, smartphones, YouTube, or Instagram.
- Public sentiment is more negative than positive; this hurts adoption and makes it harder to justify the industry’s demands (data, compute, copyright access, data centers).
- The industry’s messaging mistakes — from apocalyptic “jobless future” pitches to blaming media and consumers — have backfired.
- Companies are reacting by pivoting to enterprise/B2B (where monetization is clearer), but that won’t solve the broader reputational problem.
- Policymakers and platform gatekeepers (e.g., FCC rhetoric) are creating chilling effects on journalism and speech when they mix regulatory threat with political aims.
Evidence & data cited
- NBC News poll: 57% believe AI’s risks outweigh benefits vs. 34% who say the opposite.
- Pew study: 53% think AI will worsen creative thinking (16% say it will improve); 50% think AI will worsen meaningful relationships (5% say it will improve).
- Anecdotal reports from tech executives: Gen Z is especially skeptical or hostile toward AI.
Industry responses and narratives discussed
- OpenAI (Fidji Simo memo): push toward enterprise and coding use cases; “acting as if it’s a code red.”
- VC/Founder messaging:
- Two tactics: (1) Claim media/consumers are misled and adoption will come once products reach scale; (2) Use alarmist “doomer” scenarios (AGI/jobless future) to raise huge sums.
- Both tactics are criticized: environmental/energy arguments don’t change consumer behavior; doomerism can raise money but damages trust and public support.
- Microsoft/Satya Nadella: calls for “social permission” — AI must demonstrably improve health, education, public sector efficiency, or business competitiveness.
- Result: companies risk losing social license if they can’t show tangible public benefits.
Problems with current AI products (user experience & ethics)
- Consumer products are brittle and inconsistent (examples: AI mis-answers, factual errors).
- Monetization struggles: ChatGPT has broad adoption but costs far exceed revenue at scale (subscriptions/ads/commerce haven’t solved it).
- Resource/externality demands: huge energy and hardware requirements (data centers, GPUs, RAM) create public friction.
- Copyright and data-collection practices: industry often relies on scraped data, provoking legal and ethical backlash.
- Aesthetic and agency concerns: examples like NVIDIA DLSS 5 show the risk of platform-level aesthetic imposition (the graphics card altering an artist’s intended look).
Notable episodes, controversies & examples covered
- OpenAI memo (Fidji Simo): company pivot to enterprise; criticism that OpenAI has been scattered with "side quests."
- NVIDIA DLSS 5: an upscaling/AI filter that sparked backlash because it can override developers’ artistic choices (memes, “yassification” of characters); Jensen Huang’s defensive response at GTC was seen as tone-deaf.
- Samsung Galaxy Z Trifold: Allison Johnson’s odd hands-on — suspicious eBay purchase, phone likely a China model, Trump Mobile SIM, Samsung’s cancellation of the product; raises questions about foldable viability and why Samsung ghosted reviewers.
- Foldables market: high prices, heavy devices, no clear killer consumer use case yet; Apple fold (if it appears) could be decisive.
- Meta / Metaverse: mixed strategy (shutdowns, reversals, layoffs), Supernatural VR fitness acquisition and community fallout; Meta’s brand/trust deficits persist.
- Threats to media/First Amendment: FCC chair Brendan Carr’s public comments tying broadcast license scrutiny to coverage of the Iran war; concerns about chilling effects and politicized enforcement.
- Viral AI hype debunked:
- “Fly uploaded to a computer” — overstated; researchers and reporters debunked the claim as non-equivalent to an uploaded animal.
- “ChatGPT cured a dog’s cancer” — debunked: human researchers and concurrent therapies were the real factors; ChatGPT did not design a cure.
Notable quotes & lines
- Fidji Simo (OpenAI): “We are very much acting as if it’s a code red.” (all-hands)
- Satya Nadella (Davos): “We will quickly lose even the social permission to actually take something like energy… if these tokens are not improving health outcomes, education outcomes…”
- Hosts’ synthesis: “The industry is asking for so much and they haven't delivered a product people love.”
Key takeaways
- Social permission matters: without clear consumer value or demonstrable public benefits, AI will face sustained opposition.
- Messaging and tone are strategic: alarmist or blame-shifting narratives (doom, “you were lied to,” media blame) erode trust.
- Monetization remains unresolved for many consumer AI services; enterprise adoption is the current safer path for profitability.
- Platform- or hardware-level AI changes (e.g., DLSS 5) that override creators’ choices provoke disproportionate backlash.
- Journalistic rigor is crucial to cut through hype: many viral AI claims are unverified or false; careful reporting matters.
Actionable recommendations (for different audiences)
- For AI companies:
- Prioritize consumer value: build one or two killer experiences that people genuinely want and will pay for.
- Be transparent about resource use and data practices; show concrete public benefits (health, education, government efficiency).
- Avoid sensationalist or fatalistic messaging aimed solely at fundraising.
- For policymakers:
- Seek clear, narrowly tailored rules that preserve press freedom and avoid chilling effects.
- Reward demonstrable public benefits for high-energy/high-data projects (social license criteria).
- For journalists & newsrooms:
- Maintain rigorous verification standards; debunk viral hype quickly and clearly.
- Report both technical strengths and the real-world limitations of AI products.
- For consumers:
- Evaluate AI tools on real utility and privacy trade-offs, not just hype headlines.
- Demand transparency: how was a model trained, what data was used, and what are the costs?
Episode structure / other segments mentioned
- Hiring plug: Vergecast producer & Decoder supervising producer roles.
- Interview tease: Allison Johnson on Galaxy Z Trifold.
- Lightning round: DLSS 5, debunks of “uploaded fly” and “ChatGPT cured dog’s cancer,” Meta/Metaverse updates, FCC/Brendan Carr free-speech critique.
- Plugged shows: Decoder episodes and Version History podcast.
Final synthesis
The episode argues the core reason “people really hate AI” right now is not a single scandal or a few bad headlines — it’s a systemic mismatch between what AI companies are asking for (compute, data, legal leeway) and what they have given in return (reliable, delightful, money-making consumer products). Until companies produce clear, everyday value or convincingly demonstrate public-good outcomes, public skepticism will persist, and political/regulatory backlash will grow. Journalists and careful reporting have an outsized role in separating hype from reality and highlighting where AI actually improves lives.