LLMs and the Epistemic Apocalypse
The article warns of an emerging 'epistemic apocalypse' driven by the widespread use of large language models (LLMs) that erode the reliability of knowledge production, particularly in academic and scientific domains. It highlights a case study of a flawed paleontological paper that made unsupported claims about ancient octopus intelligence, exemplifying how low-quality, AI-generated content is infiltrating peer-reviewed research. The author argues that the degradation of written communication and scholarly rigor threatens our ability to discern truth in the post-2023 information landscape.
- ▪LLMs are contributing to an erosion of reliable knowledge by generating convincing but often inaccurate or incoherent text.
- ▪A scientific paper claiming to identify a giant ancient octopus species lacks proper measurements, images, and basic biological accuracy, such as misidentifying the largest invertebrate.
- ▪The paper incorrectly links jaw asymmetry in fossils to advanced intelligence, extending speculative claims beyond what the evidence or established science supports.
- ▪Academic publishing is increasingly vulnerable to low-quality or AI-generated 'slop' due to misaligned incentives and declining scrutiny.
- ▪The author sees this trend as part of a broader 'epistemic rot' that undermines trust in written communication and expert knowledge.
Opening excerpt (first ~120 words) tap to expand
The slow creep of epistemic apocalypse.A feedback cycle of information drifting from a measurable reality.HaversineMay 17, 2026ShareI will skip the typical “LLMs hallucinate and they aren’t objective and reflect the biases of their training data” and assume any reader understands this by now. We have been inundated with a firehose of slop from these schizophrenic stochastic parrots for years, so I am ringing the alarm around a much less visible problem that I see brewing in the bowels of peer review, research labs, and universities. What we’re experiencing is an erosion of our ability to know things about the world and each other.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Hacker News (Newest).
Discussion
0 commentsMore from Hacker News (Newest)
