Dan Cîmpianu tested #OpenAI’s #Prism and found it’s “quite happy to suit my agenda” - generating fake research abstracts, defining non-existent technical terms, and recommending gibberish for publication:
[…] It’s the responsibility of academia to take a stronger stance on this, and introduce real penalties for researchers caught with slop. Because after all, if academic integrity isn’t grounded in actual integrity, with firm guidelines, which puts qualitative research first and foremost, focusing on the value of the scientific endeavor itself, and not sheer volume of publications, or monetary incentives like research grants, we stand to lose a great pillar of civilization itself to worthless slop and cause damage that we may never mend.
https://jadarma.github.io/blog/posts/2026/01/academic-slop-just-reached-a-new-low/
[…] [Rather,] we contend that slower science is better science and that, in fact, we can’t meaningfully do science at all unless scientists can work in community with each other and have the time and mental space to engage with each other’s ideas.
[…] Regardless of why they are produced, synthetic or partially synthetic scientific papers damage the scholarly information ecosystem, mixing unreliable texts that no one can really vouch for in among those that, in theory, other scholars could be learning from and building on.”
— Bender & Hanna, The AI Con, Chapter 5
More on NON rigorous research, see Dan’s 2024 analysis where he exposed #GitHub for turning 2 extra lines of code into a “13.6% improvement” claim - based on linter warnings, not actual errors.
https://www.theregister.com/2024/12/03/github_copilot_code_quality_claims
#theaicon #nerdreich #AI #slop #academia #research #AIEthics
