How AI slop is causing a crisis in computer science (www.nature.com)

The recent rise of large language models (LLMs) in research has sparked a significant crisis in the computer science community, characterized by an overwhelming influx of low-quality submissions, colloquially termed "AI slop." This term refers to papers that are either entirely AI-generated or contain significant inaccuracies, known as hallucinations. Since the launch of tools like OpenAI's Prism, researchers have increased their productivity dramatically, leading to more than double the number of submissions to major conferences like the International Conference on Machine Learning (ICML) compared to previous years. Consequently, the traditional peer review system is struggling to cope, with rejection rates surging as oversight becomes increasingly difficult. To address this challenge, various measures are being implemented across the academic landscape. Initiatives include mandating that authors peer review each other's submissions and requiring new submitters to undergo eligibility checks. Some conferences have also introduced fees for additional submissions to discourage repetitive work. More radical approaches, such as shifting to a rolling, journal-based publication model, are under consideration. As the integrity of scientific research comes under threat, experts warn that unless these issues are tackled effectively, trust in computer science could be significantly eroded.