Dr. Casey Fiesler

This is fascinating: www.reddit.com/r/OpenAI/s/I... Someone “worked on a book with ChatGPT” for weeks and then sought help on Reddit when they couldn’t download the file. Redditors helped them realized ChatGPT had just been roleplaying/lying and there was no file/book…

My video about how LLMs are not search engines has led to many, MANY comments telling me that I should be using Perplexity. Some insisting that Perplexity does not hallucinate. Out of a list of 26 papers it just provided me (in "Research" mode) 4 were real. FOUR. 85% hallucination rate.

Tip for PhD applicants reaching out to prospective advisors: Don’t use an LLM to write these emails. It’s not a great first impression to express your admiration for a recent paper the faculty member published if that paper does not exist.

This is the part people don't seem to get. (And I'm not sure what I mean by "people" here, maybe like "in my social media comments saying things like 'humans get things wrong all the time!' or 'LLMs are more accurate every day!'") Many. People. Assume. Accuracy. www.nytimes.com/2026/04/07/t...

This is a fascinating exchange for multiple reasons. www.instagram.com/mostlyhumanm... Summary: Sam Altman is shown a viral video of someone asking ChatGPT (in voice mode) to time a run. Even though they stop after just a few seconds, the response is that the run was 10 minutes. [cont.]

I actually find this article more unsettling than reading about e.g. big tech lay-offs. I mean I do think that right now most attempts to do this would fail. BUT "startup culture but literally only one person gets everything" is some serious dystopia. www.nytimes.com/2026/04/02/t...

I just tried to google "pathetic dot" just to get to the wikipedia page to link my students to, and my finger slipped and I accidentally searched for "pathetic dog" instead. 😭

Well... my university system might have signed a $2mil deal for ChatGPT access with zero faculty input, but at least there's "training" required to access it that takes 30 seconds to click through.

I mean I'm glad that I'm helping educate young people on TikTok and all that but also my 72 year old father just told me that he told all of his friends that he meets weekly at Waffle House in south Georgia to watch my Facebook videos about how large language models work.

I am asked a LOT about pros and cons and experiences of my social media presence as an academic. The second chapter of this book covers that in great detail - and specifically from me as well as some other academics Mel interviewed!

HCI/social computing scholars! DID YOU KNOW that the 2027 ACM GROUP conference has a deadline (for PACM-HCI publications) coming up very soon (abstracts 3/24, papers 3/31)? And also that the conference has a location and dates? St Simons Island, GA Jan 10-13, 2027! group.acm.org/conferences/...

I made a playlist for my in-progress series of short videos on how large language models work. Hopefully this is more easily accessible/shareable! www.youtube.com/playlist?lis...

My first reaction to this story was "doesn't Grammarly have any lawyers?" and then I saw the statement from the CEO that the lawsuit is "without merit" (LOL) and realized that maybe their legal strategy is to ask an LLM to (badly) impersonate a real lawyer without that lawyer's consent.