Written on 15.10.2026

Research over the years

In the old days, before the internet era, people would conduct research at their institutes mainly depending on their libraries. Access to knowledge was only through libraries and discussions among colleagues and students. Occasionally, researchers would also meet at national or international conferences. Capacity was limited, so production of new results was low but also quite healthy. Exceptional minds, capable of independent thinking and having great creativity would stand out.

Then came the internet era. With the advent of the WWW people obtained access to electronic material, be it official publications, unpublished work, discussions, or blogs. Web search engines developed to a stage where it became possible to find related work quite easily. On-line    and offline remote communication channels as well as discussion threads made it easy for researchers who are not physically close to exchange ideas and collaborate. Electronic preprints of new findings were made available before the official publication dates of journal articles and before conferences took place. However, at the same time, the internet gave access to unreviewed, less credible research results, whose volume has increased over the years. It was up to the research team to consider or disregard published findings.

Then came the LLM era. Research gradually shifted from an art, where you would have to think of a  valid problem, search for related work, design and cultivate a solution, analyze it, implement it, and test it, to engineering, where you use LLMs to help you identify the problem and assess its value, design and implement the solution, and finally test it and generate content (even the paper itself). IT is no longer just an assistant for finding knowledge but also as means for producing output. Several issues arise. First of all, novelty is challenged, as new findings are due to a human-AI collaborative effort. Second, research ethics issues emerge: how far should a scientist go into using AI for conducting research? There are thin lines there. Third, what is the real value of a human-AI product, opposed to a traditional research output in the past? Fourth, does it still make sense to present works in conferences? New research findings are produced in an unprecedented pace and they can instantly be communicated and integrated into models that assist in the production of new research. Fifth, researchers playing by the (old) rules are clearly disadvantaged, in the same way as researchers denying to use the opportunities offered by the Internet in the 90's were. Sixth and most important, there is much less incentive (and less joy) nowadays in conducting research, as it becomes less and less meaningful once AI-generated content can surpass human-driven. I guess the same holds for all Arts: if a model can generate great new songs, what's the point of being a composer, even when you have the talent? 

Disclaimer: While authoring this I made no use of AI or internet search (which is AI-driven nowadays). Forgive any spelling and other mistakes. I guess I'm repeating others here...