AI moan

Aug. 22nd, 2025 10:15 pm
vivdunstan: Muppet eating a computer (computers)
Getting beyond exasperated with the number of people on forums and groups I'm on, who reply to questions, needing specialist and reliable knowledge to answer, and post a response from ChatGPT. Usually & consistently wrong. Those folk may learn it's not a good source but others later will use it too.

I'm also beyond exasperated by the misuse of generative AI in creative fields, including art, interactive fiction and writing. I'm also an academic whose own academic writings have been used without my permission to train AI, including for Meta. And then there are the environmental costs. Just stop!

My 88-year-old Mum who's never used a computer or smartphone in her life, and never will, also rants about AI. I'm quite impressed that she knows about it and has enough knowledge to have opinions :)

Incidentally I've been using and involved with some forms of AI since the 1980s. But I really really dislike the modern version and where things are going with it.
vivdunstan: Part of own photo taken in local university botanic gardens. Tree trunks rise atmospherically, throwing shadows from the sun on the ground. (Default)
Heads up for fellow academics as well as other authors. I'm not a prolific published academic, but at least two of my academic journal papers (on Scottish book history and library history) have been pilfered for AI training purposes. All done without my permission. You can search for these at https://www.theatlantic.com/technology/archive/2025/03/search-libgen-data-set/682094/

Lots of academic papers (co-authored) from Martin on there as well. And loads by my historian and computer scientist friends. It's also amusing seeing author namesakes. I now know which academic with a similar name (my maiden name - I have some early published academic work under my maiden name) is publishing prolifically about healthcare and medical matters! That Academia.edu keeps emailing me unhelpfully about, referring to "Vivienne Dunstan"!

MacWhisper

Oct. 30th, 2024 12:10 pm
vivdunstan: Part of own photo taken in local university botanic gardens. Tree trunks rise atmospherically, throwing shadows from the sun on the ground. (Default)
After battling to hear and understand the whispered Sea Devil dialogue in a Big Finish audio yesterday a friend kindly suggested that I try an auto transcription service. I've just been trying MacWhisper, which I can run locally on my Mac, no uploading the audio, which I wouldn't be happy with for copyright reasons. And it's done surprisingly well. The large AI audio model (available with a one-off pro purchase) copes best of all. But even the free small model was pretty impressive. Couldn't catch all the Sea Devil dialogue, but still did way better than me. I've now tried it with me blethering at it, and then also transcribing the audio for my 11 minutes talk about my gggg-granddad, Dandie Dinmont Terrier breeder Francis Somner. I'm really impressed with how it managed that last test, the large audio model especially. It even coped with many of the Scottish place names and counties, albeit not all. Definitely a higher level of accuracy than Apple's own internal dictation facility. So yup, I think I will find it useful. Though it was essential that it run locally on my Mac. And for once I am really grateful that I have an extremely whizzy Mac processor wise. Because I really needed it here!

P.S. I never had a script for this talk. I just burbled my way through it, using the PowerPoint slides as my prompts. It is remarkably free of ums and ahs!

vivdunstan: Part of own photo taken in local university botanic gardens. Tree trunks rise atmospherically, throwing shadows from the sun on the ground. (Default)
There's a viral message currently going round where people think they can object to Meta (e.g. Facebook and Instagram) using their content to train AI simply by posting a message in a post on their feed. This won't work. You have to formally object to Facebook using the correct process. Go to https://www.facebook.com/privacy/center and click the "object". Then you have a simple form you can "submit" to opt out. Make sure your Facebook and/or Instagram email address is listed there (it will be prefilled if you are logged in on the web browser). You can add a reason in the text box if you want. I put "I do not want any of my Facebook or Instagram data to be used to train generative AI. Please respect that." but it shouldn't be necessary. Key thing is to click "submit". You will get an email a few minutes later from Facebook/Meta saying you are opted out. This objection form used to be vastly harder to fill out. They have greatly simplified it. Please use this to opt out, not a fake viral message.
vivdunstan: Art work for the IF Archive including traditional text adventure tropes like a map, lamp, compass, key, rope, books a skull, and a sigh referring to grues (interactive fiction)
IFComp have today added new questions that authors must answer about any use of AI generated content in their works for the imminent competition. I was very pleased to be able to tick the “no AI” option for my game. Not even for the cover art, which I did myself. I am not at all a fan of the use of AI in creative fields like this, whether for text content or art. The human cost to creatives working for a living in these areas is too great, as is the environmental cost of running generative AI queries.

Profile

vivdunstan: Part of own photo taken in local university botanic gardens. Tree trunks rise atmospherically, throwing shadows from the sun on the ground. (Default)
vivdunstan

September 2025

S M T W T F S
  123456
78910111213
14151617181920
21222324252627
282930    

Syndicate

RSS Atom

Most Popular Tags

Style Credit

Expand Cut Tags

No cut tags
Page generated Sep. 2nd, 2025 10:27 pm
Powered by Dreamwidth Studios