Text generation by artificial intelligence (AI) will fundamentally change academic publishing. In recent years, several AI systems have demonstrated the production of image and text content that is increasingly indistinguishable from human-generated work. This has created new opportunities for intellectual workers almost overnight, but has also raised equally strong concerns. While artists and journalists are clearly more at the forefront of this nascent revolution, it is not difficult to imagine a researcher looking at the frustratingly poor draft of a research article and wondering, “Could a machine write it for me?”
This question was considered a pipe dream until recently, as machine-generated scientific arguments were easily distinguishable from human work, and article-creation software, in particular, highlighted the permeability of the peer review process from nonsense processes to nonsense articles. But these technologies have advanced so quickly that we have probably entered a new phase where machine-generated texts can be seamlessly integrated into human-produced scientific articles.
AI and scientific publishing
Several experts who track down problems in studies believe that the rise of AI has exacerbated the problems in the multi-billion dollar sector. All experts stressed that AI programs like ChatGPT can help with writing or translating documents if they are thoroughly tested and disclosed.
It’s not always easy to spot the use of AI. One clue, however, is that ChatGPT tends to favor certain words. Andrew Gray, a librarian at University College London, combed through millions of essays looking for excessive use of words like “meticulous,” “complicated,” or “praiseworthy.” He found that by 2023, at least 60,000 essays included the use of AI—over one percent of the annual total.
Meanwhile, according to the US group Retraction Watch, more than 13,000 articles were retracted last year, more than ever before.
Revolutionizing borders
Recent breakthroughs in the use of AI in complex strategy games have shown how surprisingly easily AI can outperform humans on problems considered intractable using computational approaches. Similar breakthroughs in the use of AI for scientific advancement could be achieved by combining (i) precise goals, i.e. a clear definition of what we consider to be a successful scientific observation, (ii) algorithms capable of efficiently optimizing the results of these observations toward these goals, and (iii) structured and accessible scientific data.
In this sense, we can imagine AI systems that propose new experiments and descriptions of observed phenomena and summarize data in figures to support their conclusions. An AI system capable of doing original scientific work could revolutionize all of science, for example, by being less bound by the boundaries of scientific disciplines than humans are and taking multidisciplinary science to a new level.
