annamillsoer.bsky.social
Writer, community college writing teacher, obsessed with AI in education, #OER advocate, author of HowArgumentsWork.org.
annarmills.com
895 posts
5,915 followers
3,108 following
Regular Contributor
Active Commenter
comment in response to
post
Oh dear! Yes, we are on pretty strict rations...
comment in response to
post
I'm intrigued by your claim that the machines may answer the questions more cogently...
comment in response to
post
😂
comment in response to
post
@FrankPasquale.mastodon.social.ap.brid.gy's book The New Laws of Robotics was recommended to me by Lauren Goodlad of the Critical AI Institute at Rutgers @criticalai-journal.bsky.social. It's been foundational for me...
comment in response to
post
Wait, do you mean that the humanities folks are largely saying we shouldn't live with it?
comment in response to
post
Thank you! Can I quote you over there?
comment in response to
post
💯. I did raise this possibility over there... I think it's a big part of it. The "Why the silence?" comment is a little naive after all that's happened to Twitter.
comment in response to
post
We have done what we could in the MLA/CCCC Task Force on Writing and AI: lots of guidance for teachers who assign writing in higher ed, but we only made one statement to broader audiences: aiandwriting.hcommons.org/2023/07/17/t...
comment in response to
post
Oh, that makes my day!
comment in response to
post
Yeah, I like that too!
comment in response to
post
No one should pooh pooh the rhetorical potential of LLMs!
comment in response to
post
Thanks, I'd love to see more on that! The mirage of AI intentionality?
comment in response to
post
"Bug" would minimize it, though. I think the exploration of "bullshit" in this context is more helpful for showing that the way the software is designed implies a fundamental estrangement from an orientation to truth.
comment in response to
post
Hmm, I suppose in the sense that whatever we see in it wasn't there by intention it's a mirage communication? So if an AI output "claims" something, it's a mirage of what a human claiming something means.
comment in response to
post
Yes, at least a feature in the sense of a core characteristic... ! There was a viral tweet from Andrej Karpathy to that effect. Something like "it's hallucination all the way down."
comment in response to
post
It's desirable when you're looking for new examples of old patterns.
comment in response to
post
🙏
comment in response to
post
Thank you--I'm so honored you think it'll be useful!
comment in response to
post
Oh, I'm intrigued by "strange error." Sorry we missed this! I'm thinking we need to update... maybe the spreadsheet of alternative terms should be ongoing. Where is the best place to read about the reasons for choosing "strange error"?
comment in response to
post
Yes, "glitch" suggests that the system is designed to work a certain way and something is stopping it, but making up with predictions that synthesize patterns from the training data is just what LLMs do.
comment in response to
post
Wow, thank you for sharing! This is the first I've heard of anyone suggesting mirage besides us, but you were ahead. And we didn't even think of the term; ChatGPT produced it when I gave it criteria for "hallucination" alternatives.
comment in response to
post
Oh that's a good one! Except for the fact that it will always need explanation... There are so many good ones; the collection of alternatives is interesting in itself.
comment in response to
post
Nice.
comment in response to
post
What would be an (inadequate) English translation - AI haunting?
comment in response to
post
Definitely on the list! Also botshit!
comment in response to
post
Lauren Goodlad of @criticalai-journal.bsky.social
comment in response to
post
@maryjacob.bsky.social @mcorbettwilson.bsky.social @mdmccradden.bsky.social @pennyjw.bsky.social @ramonalvarado.bsky.social @rleon1.bsky.social @shahaoul.bsky.social @sskapica.bsky.social @thefordon.bsky.social @warschauer.bsky.social And others we couldn't find on BlueSky: bit.ly/Hallucinatio...
comment in response to
post
@emartinsen.bsky.social @emilymbender.bsky.social y.social @eryk.bsky.social @evenannaliese.bsky.social @holdspacefree.bsky.social @jadelgador.bsky.social @jwrobertshaw.bsky.social @kconrad.bsky.social @khademinori.bsky.social @leaton01.bsky.social @mahabali.bsky.social @martc.bsky.social
comment in response to
post
In the paper, we build on suggestions from social media discussion with the folks tagged below, as documented here: bit.ly/Hallucinatio...
Thanks to @balazskegl.bsky.social @brianlfrye.bsky.social @bsbutler.bsky.social @cogdog.bsky.social @djjrjr.bsky.social @dorialexander.bsky.social
comment in response to
post
Love the meme! 😂
comment in response to
post
I like "Truth exists and it matters." Thank you!
comment in response to
post
I suppose another way to talk about mirage/hallucination is as an AI output that reflects a claim not found in the training data? Thoughts, @xolotl.org?
comment in response to
post
Thanks, exactly! I would argue that even though we don't want to be naive about truth, in the end that's what we mean here...
comment in response to
post
There was some tension between my approach and @xolotl.org's that we were trying to reconcile with that wording. Are we defining mirages/hallucinations as deviation from human expectations or deviation from truth?
comment in response to
post
Thanks for highlighting that phrase--it may not be ideal! We definitely struggled with the wording there and went back and forth quite a bit.
comment in response to
post
😂 The Hallucination region of France is now my dream vacation destination.
comment in response to
post
Join the convo about #AIMirages kicked off by this paper I wrote with @annamillsoer.bsky.social.
Reply here, or directly in the margins of the paper itself: xolotl.org/annotation/a...
comment in response to
post
Yes! I like that one... however it does have associations with lying and intention, no?
comment in response to
post
Thanks for the kind response.