Could everyone stop quoting the “hilarious” wrong answers given by google ai? It wasted a bottle of water to tell you that terrible joke, water is life, ai bullshit is death, what part of this is hard to understand
I'm a reference librarian and part of our work right now with weeding our collection has involved using search engines to see if some of our older materials are uploaded anywhere online. Right now I'm doing a couple hundred google searches each day and it's frustrating that whether I like it or not,
I'm forced to engage with AI. I can check archive.org and Hathitrust directly, but some of these books aren't there but are in other databases. I have to account for the fact that students who look for X or Y material may just do a google search.
I don't get a say in whether or not EBSCO is going to include an AI search option whenever a user interacts with our catalog. I can tell students that AI is useless or not to trust google results, but they are still going to reach out to me with requests based on that.
All that to say that yes there are people doing silly searches and posting them for clout, but there are also people posting serious questions and getting back AI results and I hope that in those cases they do get posted so that people can see how garbage this stuff is.
I had a long exchange with an EBSCO rep in my capacity as @reckoningmag.bsky.social publisher, trying to convince them AI was unethical/inaccurate, wasteful etc. She seemed receptive but ended up trying to handwave away my concerns. So they will not be carrying Reckoning content, sadly
My "favorite" thing with AI integration is all of the greenwashing. "Actually, we're using environmentally-conscious models!"
And then it's just carbon tax credits earned through like. "Awareness campaigns."
All of our tools we use to "prove" plagiarism via AI also either use AI or feed their gathered data into some model or other. I'm working on a paper about the ethics of engaging with AI in library sciences because we're really stuck between a rock and a hard place here.
The "AI to catch them" has a super bad accuracy rate, as I understand it. But worth using at all, it just doesn't work.
Better to look for different ways of measuring knowledge, probably. At this rate we'll end up totally back at aural exams before too long, huh? =P
Ah that's a really interesting approach! I didn't consider that they might just feed the exact prompt without providing any other instructions to the LLM!
