LLMs Are This Close to Destroying the Internet

Impact of LLMs on Content Quality and “Model Collapse”

• Several commenters worry about “autophagy”: AI models increasingly trained on AI-generated output, leading to degraded, cliché-ridden models and harder-to-audit errors.
• Others argue this is manageable: humans can curate data, freeze models at peak quality, or avoid feeding them their own output.
• Some see these concerns as overblown “doomerism,” noting that data curation and gatekeeping are old problems, not unique to AI.

Journalism, Writing, and Jobs

• One side emphasizes real damage: writers and journalists losing work, news orgs hollowed out, and some literary journals and academic venues overwhelmed by AI-generated submissions.
• The opposing view treats job loss as normal technological displacement and argues good, personality-driven journalism will still find paying audiences.
• There is concern that most people will end up with cheap, low-quality “news,” while a minority pay for real reporting.

Spam, Search, and Discovery

• Many see LLMs as accelerating an existing decline: SEO spam, clickbait, ragebait, and backlink farms already degraded the web and major platforms.
• AI drastically lowers the cost of generating plausible junk (recipes, articles, comments, research-like papers), making good content harder to find.
• Some argue PageRank-style link analysis should mitigate AI junk; critics reply that backlink farms, expired high-reputation domains, and social/political pressures undermine this.

Curation, Walled Gardens, and the “Small Web”

• A recurring theme: the future may belong to human-curated islands—forums, newsletters, private communities, reputation networks, and possibly a revival of early-Yahoo!-style directories.
• Debate over whether these must be private/siloed to avoid AI harvesting; some fear that “good content in walled gardens” makes the public web worse.
• Others are optimistic that falling baseline quality will create demand for curated spaces and reduce social-media obsession.

Culture, Art, and Everyday Use

• AI-generated music and text are seen as technically impressive but often boring or generic; mass-produced mediocre art is not new, but AI scales it dramatically.
• Some argue most online “hint-level” content (recipes, Stack Overflow-style help) can tolerate fuzziness; serious work will still rely on primary sources and deeper research.
• A minority hope the mess simply pushes people to spend less time online and more in offline, word-of-mouth networks.