‘No Man is an Island’ by John Donne

No man is an island entire of itself; every man 
is a piece of the continent, a part of the main; 
if a clod be washed away by the sea, Europe 
is the less, as well as if a promontory were, as 
well as any manner of thy friends or of thine 
own were; any man's death diminishes me, 
because I am involved in mankind. 
And therefore never send to know for whom 
the bell tolls; it tolls for thee.

In 2015, I attended a graduate course on data analysis near Montréal, Canada. A senior researcher came to give a lecture, about his career and accomplishments. What could have been inspiring turned into a sad and bitter litany of complaints about how Science had changed and how it was accelerating at a pace that was unsustainable with expectations for scientists to “tweet” and contribute to “blogs” to get traction on their works. The professor, you can imagine them to represent the typical stereotype of a scientist, had had a career in which research manuscripts were typed by hand and sent my post mail; in which his dear wife stayed home with the kids while he worked and came back to a lovely meal; in which correspondance between researchers across the world was much slower and data points were put on grid paper. It made me angry to listen to such a “party pooper” (pardon my bluntness ^^) rain on my parade and turn to gloom a room full of next generation scientists that just wanted to believe that there is a space for them in this world of Academia. It made me feel like he was someone that could not adjust anymore to the novel ways, “time for him to retire” is what I thought…

Fast forward to 2026, and I am wondering what is happening with the publication system in Academia and wondering if this retiring professor was right…

We have known for a while that the publication system in Science is broken. It has turned into this massive capitalist business in which journals churn scientific papers faster and faster every year. The existence of predatory journals and their presence online has exploded; good reviewers that really take the time to evaluate and comment the value of a manuscript are harder to find than ever; scientists across the world are found guilty of cheating their data or performing blatant plagiarism; and the confidence in scientific results is going down down down the drain… And now I wonder if LLMs, Large Language Models, and AI tools, are the nails in the coffin.

Oh they are great, very efficient and quite the help if you have the knowledge to guide them and use them as a support rather than as the main responsible of your scientific content:

• they can find mistakes in my code that would have taken me minutes to hours to days to fix during my PhD (probably saving me some tears and a great amount of frustration against this one missing comma haha);

• they can correct my typos and grammar mistakes (once I wrote “pubic” instead of “public”, never again!), you know English is not my first language just like for many other scientists across the world;

• they can take my ideas and spit out a first draft of a text in mere minutes;

• they can create the first version of a conceptual figure that I want to add to support my manuscript or presentation.

Oh they are good, sure. Because of this “enhancement” of our capacity, LLMs are accelerating the pace at which we can produce scientific manuscripts. And this amplifies even more the problems of our publication system:

• Scientists that did not know how to analyze their data are now using LLMs to do that task for them, without reviewing the statistical choices made, the real code performed, and the robustness of the results.

• Reviewers (let’s all remember this is a pro bono task that we do to ensure the quality of Science) that already did not have (or take) the time to review manuscripts well are now delegating LLMs to do that job, thus giving not so good reviews to already not so good manuscripts.

• Predatory journals are publishing more shitty papers but they look more polished because of LLMs. Even international journals estimated to publish good quality manuscripts are now outputting works that show the remnants of LLMs and their misuse.

• And no one reads papers anymore. No need, LLMs will find the important references, and summarize the literature.

So I wonder, if we are producing these papers and books at an ever accelerating pace, with the help of AI tools, then LLMs review these papers, and then it reads the papers for us to summarize the information, for whom am I doing Science anymore? For an AI company to train its own LLMs?

For whom the bell toLLMs? It toLLMs for thee?

Note: Thanks to Jonathan Rondeau-Leclaire for quick feedback including typos and suggestions 🙂

Oh and 2nd note: for those wondering, this post was also written with my own brain, no LLMs used in the process :))