AI is hallucinating its way into research (thelibre.news)

In mid-2024, over 60,000 scientific papers were reportedly generated using large language models (LLMs), raising alarms in the academic community regarding the quality and integrity of research. This surge can be attributed to a pervasive "publish or perish" culture, where researchers, under immense pressure to produce more publications for careers and funding, increasingly lean on AI tools for tasks such as writing, reviewing, and even generating illustrations. While LLMs can streamline the research process, they also exacerbate the issues of junk science, as seen in a recent case where an AI-generated article featured absurd visuals and was quickly retracted due to questionable academic rigor. The significant adoption of LLMs and AI in research illustrates a systemic problem in academia, where there are strong incentives for quantity over quality. Current publishing practices, dominated by a few profitable firms, often prioritize rapid publication and citations over thorough peer review. This environment not only fuels the rise of predatory journals, which capitalize on the rush to publish, but also fuels redundancy in scientific literature, with many studies merely reiterating existing findings. As the global landscape of research shifts, especially with rising contributions from countries like China and India, the call for reform is growing louder, pushing for initiatives like the open data movement to foster more transparency and collaboration in science.