PhD students who write dissertation with AI less likely to stay in academia (www.nature.com)

A new working paper from the National Bureau of Economic Research reveals that approximately 30% of PhD dissertations in STEM fields in the U.S. contained AI-generated text in 2023, with this figure expected to rise sharply in subsequent years. The study, analyzing over 226,000 dissertations, indicates that PhD candidates utilizing AI tools for their writing are more inclined to pursue industry jobs rather than academia. For instance, graduates with AI-generated abstracts were found to be 17% less likely to enter academia while enjoying a 16% higher chance of securing industry positions. This trend raises concerns within the AI/ML community about the implications of relying on AI for essential academic work. Researchers, including study co-author Daniel Gross, question whether students are attaining the necessary depth of understanding in their fields when they delegate substantial portions of their dissertations to AI. As the integration of AI in academic writing accelerates, particularly in fields like mathematical and computer sciences, the need for robust research training becomes critical, especially in less charted areas of study where original inquiry is vital.