El Podcast

AI Slop Is Flooding Science

Episode Summary

A computer scientist warns that AI isn't just changing science—it may be quietly corrupting it. As AI-generated papers, reviews, and misinformation flood academia, society risks entering a new era where distinguishing truth from statistical noise becomes increasingly difficult.

Episode Notes

An AI developer and researcher argues that large language models are accelerating the collapse of academic gatekeeping, flooding science with low-quality research, and creating a potential "epistemic dark age."

Guest Bio

@haversine.substack.com is a computer scientist, AI developer, programmer, and mathematics educator who has worked on language models, healthcare AI systems, and education research. His work focuses on how humans learn, how language shapes cognition, and the growing impact of AI on academia and society.

Topics Discussed

• AI-generated academic fraud
• LLM hallucinations in scientific journals
• The replication crisis in science
• Peer review and AI-generated referee reports
• Academic incentives and "publish or perish"
• Dead Internet Theory
• Stanford research fraud scandal
• AI's impact on truth and knowledge
• College, credentials, and declining academic standards
• Cognitive decline and education
• Silicon Valley incentives
• AI hype versus reality
• Dating, social media, and Gen Z
• The future of work and higher education
• Robert Gordon's innovation thesis
• Why society may be entering an "epistemic dark age"

Main Points

• AI-generated content is increasingly appearing in academic journals, including top-tier publications.
• Peer review itself is becoming automated, creating a system where AI-generated papers are reviewed by AI-assisted reviewers.
• Academic incentives reward publication volume rather than truth-seeking, making AI misuse almost inevitable.
• Language models risk contaminating future knowledge because they are trained on previous outputs, including errors and hallucinations.
• The guest argues society is losing its ability to distinguish expertise, competence, and genuine understanding from AI-generated text.
• Many AI companies overstate the capabilities of their systems while underplaying their limitations and risks.
• Higher education is suffering from credential inflation, declining standards, and growing dependence on AI tools.
• Social media and smartphones have fundamentally altered how younger generations form relationships, learn, and engage with the world.
• The decline in friction, boredom, and real-world challenges may be reducing resilience and critical thinking among young people.
• The biggest risk of AI may not be superintelligence but the gradual erosion of humanity's ability to know what is true.

Top 3 Quotes

"We're just in real time losing our ability to interrogate information."

"Everyone is asleep at the wheel, and I don't see how this gets fixed."

"The scary thing with language models is that they're going to calcify bad information, bad epistemics, and carry that forward forever."