Flutterby™! : A few AI notions

Next unread comment / Catchup all unread comments User Account Info | Logout | XML/Pilot/etc versions | Long version (with comments) | Weblog archives | Site Map | | Browse Topics

A few AI notions

2025-09-09 20:21:07.719145+02 by Dan Lyke 3 comments

Ben ‪@franzferdinand2.bsky.social‬

I remember someone saying that's how you know AI can never truly be a writer. If you tell AI to stop using em dashes, it'll stop. If you tell a writer to stop using em dashes they'll tell you to fuck yourself and that you can pry them from their cold, dead fingers.

‪Dr. Damien P. Williams Wants A Better Sociotechnical Paradigm‬ ‪@wolvendamien.bsky.social‬

LLM responses which do not reflect consensus reality & facts are produced via the Same Process which generate responses which *Do* broadly conform to consensus reality & facts.

The Same Processes.

The Same Ones.

"Hallucination" is just a word to distance yourself from the outputs you don't like.

Modern Diplomacy — Finance — Gemini AI Predicts Best Meme Coins to Buy in Q3: Dogwifhat, Fartcoin, Snorter (Via).

JP @jplebreton@mastodon.social

@mhoye being 100% right about VR, then crypto, then NFTs, then the metaverse, then LLMs doesn't feel good. it feels like being the sole adult in a peewee football match. just pitiful. we could be living in a world that is solving real problems.

[ related topics: Interactive Drama Writing Sports Marketing Cryptography Currency Artificial Intelligence ]

comments in ascending chronological order (reverse):

#Comment Re: A few AI notions made: 2025-09-12 00:50:26.379281+02 by: Definitely Not a Bot

"Hallucination" is a category error. It falsely implies perception.

#Comment Re: A few AI notions made: 2025-09-13 00:30:48.635873+02 by: spc476

I prefer "confabulation" myself, but that never caught on.

#Comment Re: A few AI notions made: 2025-09-15 09:36:57.565378+02 by: Dan Lyke

I'll have to work on gtting "confabulation" into my lexicon.

And, yeah, on "hallucination": Anthony @abucci@buc.ci

The only hallucinating happening with LLMs is when you, the user, look at the output and think it is real, or true, or correct; or that the LLM is answering a question or following a command. It's not unlike looking at a movie set and believing you're looking at a real town.

#AI #GenAI #GenerativeAI #LLM #Copilot #Claude #Gemini #ChatGPT