The cost of creating lines of code is dropping. Everybody, with the help of a few prompts, can produce thousands and thousands of lines of code. This is also the case for words, and the cost of creating text is also falling dramatically. As we have recently seen, conference registered abstracts are skyrocketing (ICLR Went From 490 registered abstract to 62,000 in Ten Years1) and we start to see research papers made by people claiming openly that they did not and were not able to verify their content2.
As a consequence, where does the value of a research output lie?
Is it in the thousands of lines of code produced with a mere understanding of what the software does? Or is it coming from the knowledge of the data workflow, the architecture, the implementation of the algorithms, their reusability (in the widest, FAIR3, definition of the word), accuracy and impact?
Is it in the thousands of words that constitute a paper? Or is it in the framing of the problem, the understanding of the process, of the chain of thought, logical steps and the assessement of the results in a wider research context?
AI can be a very powerful ally and can speed up the production of knowledge, but can also have a real pernicious impact on human interactions and collaborations if it is not used in a responsible and collegial way, with mutual understanding and respect.
What follows is not about AI-assisted coding. It is not a list of best practices applied to AI generated code. It is about how, in the RSE team, we value the work we bring to research projects and how we value the work of our research colleagues, their time and their wellbeing, at any stage of their career (PhD students, Post docs and Academics). It is about how we want to keep research and technical rigour, human growth and development, and a positive team culture in an epoch where volume seems to be the only metric that matters. And finally, this is an attempt to maximise the positive impact of AI in research while trying to minimize the negative aspects.
Note: What is below does not supercede the University of Sheffield Good Research and Innovation Practices policy on the use of genAI in research. If you want to consult it you can find it on the University of Sheffield webpage4.
“A sub-Earth-sized TESS planet candidate in the nearby M-dwarf binary GJ 774, most likely orbiting GJ 774 A (TIC 351339274”. Quote from the author: “The author has not independently checked the analyses and does not have the expertise to verify every technical detail; the work relies instead on the validation steps listed below and on the public release of all code and data, so that others can check it.” It is worth noting that the paper has since been withdrawn from arXiv. ↩
Principles for Using GenAI in Research and innovation at the University of Sheffield ↩