GPT might be an information virus (2023)

Original Article ↗ Hacker News Discussion ↗

LLMs as “Information Virus” and Intellectual Homogenizer

  • Several commenters agree with the “virus” framing but argue the damage goes beyond the article: LLMs flatten style, reward laziness, and normalize a bland, corporate tone as “good writing.”
  • Others counter that people have always taken shortcuts; to blame LLMs for “making people dumb” requires evidence of measurable declines, and current homogeneity is largely due to users relying on default prompts.

Historical Analogies: Printing Press, Church, Calculators

  • One camp likens critics of LLMs to the Church fearing the printing press: powerful tools always scare existing elites and eventually democratize knowledge.
  • Critics say this gets the analogy backwards: the press let individuals interpret texts themselves, whereas centralized LLMs risk becoming a new “priesthood” that interprets everything for us.
  • Similar comparison with calculators: they were helpful but arguably eroded basic mental math; LLMs could analogously erode basic reasoning.

Impacts on Work, Emotion, and Daily Life

  • Commenters note surprising adoption in domains that “should know better” (law, medicine, education, grading) and in intimate contexts: romantic messaging, navigating relationships, even aspirations for AI sexbots.
  • Some see this as outsourcing not feelings but expression; others find the emotional outsourcing itself disturbing.

Homogenization vs Cultural Dynamics

  • Some predict accelerating homogenization of language and thought, continuing trends from radio/TV/internet toward global sameness.
  • Others argue LLM-driven saturation may actually speed style cycles: people will quickly tire of the “LLM voice” and seek novel, countercultural aesthetics.

Information Ecology and the Future Web

  • Many worry that AI-generated sludge will swamp human content, making expert writing harder to find and further undermining trust in the web.
  • Others argue the internet was already broken by SEO spam and social media; LLMs are just another step, and the prediction of total epistemic collapse is overblown.

Defenses and Human-Only Islands

  • Proposed responses: offline, pre-AI knowledge snapshots (e.g., curated Wikipedia dumps); verified human-only directories; library- or community-run local news; and a revival of web rings and link pages for human-curated discovery.
  • There is skepticism about enforcing human verification at scale, but broad agreement that curated, human-centric spaces will become more important.