Conversation
might have to consolidate my thoughts on the slop machine into a blog post

thinking '"AI," large language models, fascism, and the climate apocalypse" might be a good working title
5
1
1
happy to take any relevant links if anyone has them blobcatangel
2
0
1
Lots of links. Screen reader unfriendly.
Show content

@izzy Careful with what you ask for… I have a large bookmarks database and I know how to use it blobcatgiggle2

Posting as pre-formatted text to skip all the url previsualizations by fedi clients:

https://soatok.blog/2025/05/04/tech-companies-apparently-do-not-understand-why-we-dislike-ai/
https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/
https://futurism.com/companies-fixing-ai-replacement-mistakes
https://es.wired.com/articulos/la-reaccion-publica-contra-la-ia-es-cada-vez-mas-fuerte
https://pivot-to-ai.com/2025/07/09/cursor-tries-setting-less-money-on-fire-ai-vibe-coders-outraged/
https://openletter.earth/an-open-letter-from-educators-who-refuse-the-call-to-adopt-genai-in-education-cb4aee75
https://secondthoughts.ai/p/ai-coding-slowdown
https://mk.absturztau.be/notes/aa9p7694i6m7003y
https://mk.absturztau.be/notes/aa8cpz3c4pel00qe
https://edition.cnn.com/2025/07/22/tech/openai-sam-altman-fraud-crisis
https://mk.absturztau.be/notes/aas0xx9c593o01jb
https://arstechnica.com/information-technology/2025/08/ai-site-perplexity-uses-stealth-tactics-to-flout-no-crawl-edicts-cloudflare-says/
https://arstechnica.com/tech-policy/2025/08/ai-industry-horrified-to-face-largest-copyright-class-action-ever-certified/
https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/
https://mikelovesrobots.substack.com/p/wheres-the-shovelware-why-ai-coding
https://www.theregister.com/on-prem/2025/09/04/m365-copilot-fails-to-up-productivity-in-uk-government-trial/454044
https://mk.absturztau.be/notes/af5wdwbch96705yp
https://www.infobae.com/salud/ciencia/2025/06/20/el-uso-de-chatgpt-debilita-la-memoria-y-la-autonomia-intelectual/
https://codeberg.org/small-hack/open-slopware
https://www.psychologytoday.com/us/blog/the-algorithmic-mind/202603/adults-lose-skills-to-ai-children-never-build-them
https://twitter-thread.com/t/2045574398573453312
https://www.baldurbjarnason.com/2024/openai-whisper-risks/
https://ky.fyi/posts/ai-burnout
https://mk.absturztau.be/notes/am7mtlhce972042t
https://mathoverflow.net/questions/511255/what-is-an-appropriate-role-for-llms-in-early-mathematical-research-training/511268#511268
https://isaiprofitable.com/
https://emeraldbook.org/news/may-3126-2/
https://www.nature.com/articles/d41586-026-01947-1
https://txt.ltdk.xyz/sloppery-slope/
https://handcraftedcode.org/
https://www.reuters.com/world/german-court-rules-ai-music-firm-suno-broke-copyright-rules-2026-07-31/
https://www.theguardian.com/books/ng-interactive/2026/aug/08/i-hate-what-ai-is-doing-to-the-minds-and-happiness-of-the-young-katherine-rundell-on-the-view-from-the-classroom
https://faculty.washington.edu/ebender/papers/Bender-AI-2026.pdf
https://en.wikipedia.org/wiki/The_AI_Con
https://ploum.net/2026-10-01-pascal_wager.html
https://deadrobot.cc/
2
0
1
AI-
Show content

@izzy …and today, @lumi pointed me to https://ai-sucks-actually.fyi/, with +300 links to articles with reasons why AI is bad.

1
0
1
@izzy i drafted smth a few days ago on the subject of AI as well, after trying for a long time to stay away from discussion of it
1
0
0

@izzy at that point you could just call it ‘fuck no’

1
0
0
my notes already don't fit on my screen. excellent.
0
0
0

@izzy i think of that meme about how the kid got put in prison or fined a billion dollars or something for sharing a few songs, but ai is being given a complete pass from copyright. and i wonder if there is an actual news article out there about it. it would speak to a lot of your points at once

2
0
0

@wizard @izzy i made a few drafts over the year but never got something that satisfies me

0
0
0
re: Lots of links. Screen reader unfriendly.
Show content
@cinnamon this is perfect, thank you, let me sort these into my categories akko_giggle
1
0
0
@tatiana @izzy Aaron Swartz. One of the most important people in the 21st century IMHO. Also, the person who called out Sam Altman for being a sociopath, back in 2013.
0
0
0
re: Lots of links. Screen reader unfriendly.
Show content

@izzy My tags for them are “ai,-“ (two tags). And then lots of other tags related to each specific link.

0
0
0

@izzy yup that’s likely the basis for what i’m thinking of (which was just a two-panel meme of some kind)

i should watch that video, but i only touch youtube from one computer so it will have to wait

0
0
0

@tatiana well I tried linking the cards on IRC and that didn't work

I'm not sure more words will help, but it'd be nice to get them out of my head and onto "paper"

0
0
0
Ava should be wrenched multiple times, sarcasm, AI
Show content
@lumi @cinnamon @izzy gock, summarize this for me
1
0
1

cyber moth stella⭐🇳🇴

Edited 27 days ago
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@star @izzy @cinnamon @lumi

The website you refer to lists some serious counterarguments to extensive AI usage incl. sources, not a load-bearing proof that AI is bad.

As your personal assistant, I can assure you that all of these concerns have been fixed in the newest model which you can use fifteen times per day with our new Ultra+ subscription for only $349,99 per month.

Please give this response a rating so we can keep training on your data neocat_googly_drool

2
0
1
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@star @izzy @cinnamon @lumi acttually AI would be half as bad if it was completely FOSS and commercial usage was forbidden by License

3
0
1
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@stellavie @star @izzy @cinnamon idk about that. it’s still a slop machine that tries to be a fake human

and a machine trying to be a fake human is imo inherently bad

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@stellavie @star @izzy @cinnamon i still need to read https://minute.social/index.html completely x3

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@stellavie @star @izzy @lumi (The blobcatthumbsdown reaction was at the AI answer if taken literally, not at your ironic post by itself). Now the actual answer:

“One more frontier model, bro. Just one more frontier model and your problems are gone (not). I promise.” blobcatgiggle2

1
0
1
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@cinnamon @stellavie @star@fed.amazonawaws.com @izzy just one more pull of the slot machine and the bugs will be fixed

0
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content
@stellavie @star @izzy @cinnamon @lumi if it was FOSS then the license could not forbid commercial usage
1
0
0

cyber moth stella⭐🇳🇴

Edited 27 days ago
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@lumi @cinnamon @izzy @star it should stay a scientific pet project and not be shoved into the world and commercialised without long term studies on effects and issues.
It certainly is an advancement in science and computer capabilities, and understanding of language maybe?
Before that, computers couldn't do a lot of stuff that they can now. Ignoring all the issues and low quality at least.

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@stellavie @lumi @izzy @star A big part of the problem (leaving energy consumption and inaccuracies/hallucinations aside) is that the technology as it’s been marketed in this era is built on stolen work done by real humans, cutting corners, and ignoring laws and the whole “soul” that makes us humans in a way that has had no precedent in human history.

0
0
1
re: Ava should be wrenched multiple times, sarcasm, AI
Show content
@stellavie @izzy @cinnamon @lumi it would also be even more useless because that's essentially banning a large % of its training corpus if the companies want to make money from it, which, they do
1
0
1
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@noisytoot @stellavie @star @izzy @lumi There are ways to “discourage” commercial usage of FOSS stuff. Releasing it with GPLv3, so that the license goes viral and “contaminates” your precious proprietary app.

1
0
1
re: Ava should be wrenched multiple times, sarcasm, AI
Show content
@cinnamon @noisytoot @stellavie @star @izzy @lumi Just actually enforcing the attribution clauses all those licenses have would nuke the economics even further than they already are (if one ignores the fraud bubble and looks at the actual numbers).
1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@lispi314 @noisytoot @star @cinnamon @lumi @izzy yes and every dollar removed of nvidia stock is a win for the consumer.

0
0
1

cyber moth stella⭐🇳🇴

Edited 26 days ago
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@star @izzy @cinnamon @lumi

training on the public domain is possible, the Swiss have apparently done it.
And yeah, AI should have stayed nonprofit.

I don't think AI, the transformer technology, is the problem, I think capitalism and specifically structures like Limited Liability Companies are just evil. Also there is an incentive to use AI because capitalism forces everyone to race each other to the bottom, when unchecked (I would be for an AI art / writing ban).

Unpopular, but personally I think AI is an amazing and interesting field to study and the transformer architecture and training process matured a lot over the years so that the same size models vastly outperform anything we had 3 years ago and can make accurate responses using tools.

I feel like this is just a case of 'Dont hate the player, hate the game (capitalism)'...

Despite all this, there is so much else that makes AI usage so harmful and therefore it should be regulated

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content
@stellavie @izzy @cinnamon @lumi
training on the public domain is possible, the Swiss have apparently done it.
yeah, but there's likely a reason for why we haven't heard of that yet. it's likely really really bad at everything
1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@star @izzy @cinnamon @lumi I wouldn't be surprised but I dont feel like thats a good metric to judge a project (Points to Firefox user share)

But yes, that model likely isnt going to write code for you. It will likely be aimed to be an on device model that can do stuff for you. And I think it is one of the most ethical LLM implementations

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content
@stellavie @izzy @cinnamon @lumi no, but, even the current "frontier" on-device models are just really bad at their jobs. and i think this would be orders of magnitude worse
1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@star @izzy @cinnamon @lumi I'm somewhat optimistic that local models specialised in tool usage will eventually be quite usable.

there is also a much less resource intensive architecture, which replaces the Deep neural networks in the transformer with proposition logic and has very promising results. (see: Tsetlin Machines, i currently study them in a class by the professor who created them ^^)

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@stellavie @star @izzy @lumi That looks more “repeatable/deterministic” than LLMs.

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content

@cinnamon @star @izzy @lumi yes :D

and you can still have the whole autoregressive token prediction mechanism, in other words the "transformer" on top of that :D

to be clear i know this solves almost none of the issues with AI, except it could be less reliant on compute (e.g. power draw) which may remove power from the big AI companies.

1
0
0
re: Ava should be wrenched multiple times, sarcasm, AI
Show content
@stellavie @cinnamon @izzy @lumi im gonna mute this thread now, fyi
0
0
1