My Relationship with AI Is Changing as AI Slop Spreads

Summary

In this personal essay, Stephen D. Turner says his view of AI is becoming more cautious after seeing a rapid increase in AI-generated “slop” across inboxes, social feeds, blogs, slide decks, and scientific writing. He argues that long, polished documents can conceal the author’s judgment and that papers are becoming weaker signals of productivity, expertise, and contribution. He cites Anshul Kundaje’s criticism of publication-based academic assessment and describes TMLR editor Nihar Shah’s conversations with authors whose submissions were headed for rejection: several could not explain basic ideas or technical details, while another paper contained a major error in a central claim. Turner notes that these interviews took more than 20 hours and cannot scale, while the flood of cheap content could push readers toward institutional prestige and make it harder for researchers outside prominent institutions to be noticed. He also says AI-generated prose is making him less interested in reading because he reads partly to understand how another person thinks. His own experiment with avoiding AI convinced him that writing unaided improved recall, understanding of structure, revision, and enjoyment. Turner still uses AI for coding, educational web applications, and a multi-agent workflow that checked citations in a biosecurity preprint, although the models eventually refused further fact-checking on that subject. He concludes that he will continue using AI selectively, while trying to retain responsibility for the writing and thinking attached to his name.