Artificial intelligence researchers hit by flood of ‘slop’
Artificial intelligence researchers hit by flood of ‘slop’
https://www.ft.com/content/54e274c5-de86-4b3e-96a9-95a46b5e48a0
Publish Date: 2026-02-01 00:00:00
Source Domain: www.ft.com
- Efforts to mitigate “AI slop” to maintain trust in AI research are intensifying.
- AI conferences are restricting large language model (LLM) usage in writing and reviewing papers due to a surge in poor AI-generated content.
- Scientists warn that the influx of low-quality AI material could erode the integrity and trust in AI research by introducing false claims.
- Studies reveal significant percentages of AI-written papers and reviews at top conferences, including 21% of reviews at ICLR and 9% of submitted papers at NeurIPS with AI-generated content.
- ICLR updated its guidelines to penalise the use of LLMs for creating low-quality reviews or undisclosed extensive usage in papers, rejecting submissions that don’t meet standards.
- The increase in AI-generated papers is largely due to commercial incentives leading to a focus on quantity over quality, risking model degradation and unethical shortcuts.
- While AI tools have legitimate uses in research like assisting and proofreading, responsible use is emphasized to ensure rigour and prevent negative outcomes.
- Companies are promoting AI models as “co-scientists” which could accelerate research but must avoid using uncurated AI-generated content for training to maintain model performance.
