More and more scientists and researchers are using artificial intelligence to render study reports faster and write research papers. The arrival of this technology is shaking up the academic world, and involves risks.
We knew ChatGPT author, journalist, and even hacker. There is now a ChatGPT author of scientific studies. The artificial intelligence developed by OpenAI has been phenomenally successful since its launch at the end of 2022. Used by many professionals in their day-to-day work, ChatGPT has transformed many professions – and the latest may well be that of scientist.
It is The Daily Beast which reports the news in an article published on May 6, 2023: more and more scientists and researchers are using artificial intelligence to write their publications. In an academic world where the number of publications in scientific journals determines careers, the arrival of ChatGPT does not go unnoticed.
ChatGPT author of radiology studies
The Daily Beast article takes the example of Som Biswas, a radiologist working at the University of Tennessee in the United States. After having carried out several attempts to achieve a satisfactory result, and by doing several prompts, Som Biswas succeeded in perfecting his technique. He says he used ChatGPT to write 16 scientific publications, 4 of which appeared in specialized journals. In Pediatric Radiology, where his study ” The Impact of ChatGPT in Radiology was published on April 28, 2023, he is noted as the sole author. Only a mention at the bottom of the text specifies that the text was written by ChatGPT, and that he edited and formatted it afterwards.
The Daily Beast explains that the example of Som Biswas is not unique. ChatGPT and other generative AIs are being used by researchers to help them write their research faster. However, not all scientists are as honest as Som Biswas, and many hide the involvement of AI in their rendering.
Elisabeth Bik, a microbiologist and specialist in the question of integrity in science, explained to The Daily Beast to have seen the number of researchers ” abuse of ChatGPT usage increase. Some would have used artificial intelligence to manufacture ” the chain dozens of articles in recent months. According to her, the authors would also not have specified that ChatGPT had generated the texts.
The reason behind this behavior is easy to understand: the citations [dans des journaux] and the number of publications are the two measures of the success of academic researchers “, explains Elisabeth Bik. The more studies an author has published, the more competent he or she is considered. In this context, it is logical that some try to use tools like ChatGPT.
The chatbot developed by OpenAI is however far from being a scientific reference. ChatGPT regularly lies, invents facts, and does not source the information it gives well. On top of that, it’s also susceptible to racist and sexist biases, like all artificial intelligences. These biases must be taken into account in research to achieve neutral results that are as close as possible to reality.
Despite these shortcomings, many continue to use AI. Stefan Duma, a professor specializing in engineering at the University of Virginia and editor of the Annals of Biomedical Engineering journal, said he has to reject more and more low-quality ChatGPT-generated articles. ” People do anything with AI. People sometimes send me the same letter 10 times, changing only one word “, he regrets.
Subscribe to Numerama on Google News to not miss any news!