Breaking scholarly news for AI enthusiasts - a preprint seen on ArXiv.org: Berglund, L., Tong, M., Kaufmann, M., Balesni, M., Stickland, A.C., Korbak, T. and Evans, O., 2023. The Reversal Curse: LLMs trained on" A is B" fail to learn" B is A". arXiv preprint arXiv:2309.12288.
ArXiv offers videos on some of its most popular newly published papers, so CS & AI folx may want to tune in for the details of the paper here: https://www.youtube.com/watch?v=1qkiGacDnm4
May we live in interesting times, as they say... times when scholars are still researching what works and what does not with large language models, such as ChatGPT, Bing's Chatbot, and Bard.
This is a reminder for all of us - that no matter how easy they are to use and convincing their responses - AI needs to be checked for accuracy. It's harder to ensure the truth of what the model is putting out when we don't fully understand a subject area, of course. This is a big challenge for students who are just beginning their learning and research journey.
Feel free to talk to a librarian if you want help fact-checking AI outputs. Maybe you'll be able to write a paper about what you learned along the way about your AI's truthiness (don't forget to acknowledge the help of your librarian partner, so we humans can keep working, too)!
Elihu Burritt Library
Central Connecticut State University, 1615 Stanley Street,
New Britain, CT 06050 - Map
CCSU Home | Central Pipeline | CentralSearch / Catalog | Sign In to CentralSearch | Search Library Website
0 Comments.