Marking student project reports. On checking the references I found many are made up, presumably by AI. When I query AI about them it gives a fake summary of the made up references. π€―
18.06.2025 10:32 β π 9 π 4 π¬ 5 π 0@yshendani.bsky.social
Like an interior decorator but the house is a *promising* material and the chairs are molecules. And none of it is real.
Marking student project reports. On checking the references I found many are made up, presumably by AI. When I query AI about them it gives a fake summary of the made up references. π€―
18.06.2025 10:32 β π 9 π 4 π¬ 5 π 0My friend Prof Andrea Sella, returns his Faraday Medal to Royal Society in disgust about the lack of action about Elon Musk - on.ft.com/4kDYwbS via @FT
10.06.2025 17:02 β π 51 π 11 π¬ 1 π 0Had the pleasure of watching "Insane Asylum Seekers" in Bush Theatre.
Great performance and such a packed narrative about being British-Iraqi. I was between tears and laughter the whole way through. β€οΈπ
www.bushtheatre.co.uk/event/insane...
I was in the mood to explore the scope and vibe of different journals so I got DeepSeek to summarise and more importantly, to throw some shade on them - rather amusing! π€π
Though I shan't divulge which journals and what shade π
And yes, the same might be used to write articles but we can much more easily and rapidly review and reproduce a study (in my field). Reviews are so long - would referees be able to catch the overly positive and not sufficiently critical writings of an LLM? Gosh, I hope so. π§
30.01.2025 16:36 β π 1 π 0 π¬ 0 π 0LLMs regurgitate an overly positive account of *everything working wonderfully* which unscrupulous parties may use to turbo write a review, thus transforming tentative statements into π’LOUD CLAIMSπ’
30.01.2025 16:36 β π 1 π 0 π¬ 2 π 0Each paper typically ends with a few tentative implications that we already know err on being positive. We might take these with a small pinch of salt and deduce caveats on a critical reading of the methods and results. The LLM is of course not doing any of that, AND then...
30.01.2025 16:36 β π 2 π 0 π¬ 1 π 0Reading review papers, I am becoming more concerned that with large language models, review papers might be especially easy to generate and they would contribute a whole lot more noise in the scientific literature πΆβπ«οΈ Consider this:
30.01.2025 16:36 β π 1 π 0 π¬ 2 π 0