If AI had the medical researcher’s job Premium

As a reflection on how a lot has changed in the last 50 years, the author says he is both amused and annoyed on how Artificial Intelligence (ChatGPT) is used to generate an entire medical research paper.

I read the article by Gemma Conroy (in Nature) on how scientists used ChatGPT to generate an entire research paper from scratch. I was both amused and annoyed when I saw this report. Having started publishing research papers in the 1970s, when I was an undergraduate medical student, this gave me a chance to reflect on how a lot has changed during the last 50 years.

When I started my journey in publishing research papers, there was no PubMed or Google, why, there were no computers! One had to go to the medical college library which, if one was lucky, would have a journal relevant to one’s topic, but the articles would have been published several months ago as the journals were being sent by surface mail from abroad. One had to go through articles manually, look up the references at the end of the article, tabulate the cross references, make a list of relevant papers needed and then start the process of trying to get those articles. A few would be available at the National Medical Library at Delhi, but it would mean a trip to Delhi to get the references, and that was not cheap. Moreover, after spending a day or two browsing through the journals, one would have to request for copies of the articles, which could take a few weeks as they had to be photocopied and sent to Chennai.

Regarding writing of the article itself, since there were no computers, one had to write them longhand and then get it typed out on a manual typewriter. One had to maintain a list of references separately and include them in the article as the paper progressed. By the 1980s, computers had arrived which made the typing of articles much easier. With the advent of PubMed and Google, life became much easier for the researcher, for, at the touch of a button, one could at least get the abstracts of all relevant articles published anywhere in the world. It still meant that one would have to write the articles oneself.

Soon, the concept of ‘ghost writing’ through medical writers arrived and some, particularly in the pharmaceutical industry, started employing these medical writers to write articles for them. Finally, we are now in the era where the entire research article is written by ChatGPT!

In all fields we are seeing skill sets declining. In clinical medicine the use of the stethoscope, or one’s fingers to palpate the abdomen or percuss or auscultate the chest are all rapidly disappearing as these are being replaced by X-rays, CT and MRI scans. I find some young doctors, who come to train with me, are unable to even write a patient’s case notes properly. Their reading skills have come down considerably and most of them use cell phones to Google whatever information they want. With the arrival of ChatGPT, this could be the death knell for scientific writing among junior researchers. ChatGPT is so clever that it is even difficult to detect plagiarism. It’s a great pity if our mental abilities and skills are not honed and we lose the skill of scientific writing.

Undoubtedly, we can make use of software to help us in our research, but asking programs like ChatGPT to write the full paper, defeats the very purpose of scientific inquiry. The word research literally means that we have to search and search again. When using the likes of ChatGPT, there is no searching or researching, instead, we are simply outsourcing the job to a computer.

I can foresee that if this trend continues, the number of publications, the number of citations, impact factor of journals, the h-index and other scientometric indices used to measure research, will all become redundant. It will also be difficult to detect scientific fraud, because as these programs improve, they will make fraud detection more difficult. For those of us who have been brought up in the conventional way of doing research, these developments are worrisome to say the least. It is even more frightening when we think ahead to what the next frontiers in computers and artificial intelligence will bring.