llm
I hate how LLMs are often used, but as a model of cognition I think they're fascinating. Take this paper, for example.
Here they look at reasoning over long sequences of data that can't fit into working memory. The latest LLMs can solve problems like this, using hybrid memory models that consolidate "short-term" memories in an LRU cache into "long-term" memories in the form of persistent weights (which is good to know about generally).
However, this paper shows that this is still a very hard problem. In particular, it's hard to organize the consolidated information in a way that's useful for doing future work, and that a "sleep-like" process of iterative offline processing can help significantly.
It remains to be seen how important this will be and how "sleep-like" it really is. Still, I love how this work draws inspiration from nature and could potentially help us learn about animal minds and cognition generally. Much better than many LLM studies.
ICYMI: some dork used #Claude #LLM to 'fix' an issue in #rsync and broke it (v3.4.3).
#GenAI is shit and shld not be used for code that needs to work, even if you have the morals of a billionaire skunk.
Lots of critical tasks, such as backup 🤦♂️ will be broken as people update their #Debian #Linux thanks to this stupid obsession and psychosis.
How much will that cost? On top of all the costs inflicted on people who never touch the stuff and don't want to run anything built with it, like me.
YAY! New York is the first US state to ban smart glasses in all its courthouses. https://www.engadget.com/2211151/new-york-first-us-state-ban-smart-glasses-all-courthouses/ #AI #LLM
“Google Chrome is reaching into users' machines and writing a 4 GB on-device AI model file to disk without asking. The file is named weights.bin. It lives in OptGuideOnDeviceModel. It is the weights for Gemini Nano, #Google's on-device #LLM. #Chrome did not ask. Chrome does not surface it. If the user deletes it, Chrome re-downloads it.”
#privacy #sustainability #consent #values
https://www.thatprivacyguy.com/blog/chrome-silent-nano-install/
AI paranoia is a symptom of a larger problem: a corporate culture that values quantity over quality.
https://lithub.com/its-getting-harder-to-spot-ai-in-contemporary-publishing-and-thats-very-very-bad/
"[Leo] referenced The Lord of the Rings in his encyclical about AI—an expert (if unintentional) troll of tech billionaires who keep misinterpreting the series."
Which see:
“It is not our part to master all the tides of the world, but to do what is in us for the succour of those years wherein we are set, uprooting the evil in the fields that we know, so that those who live after may have clean earth to till.”
https://www.wired.com/story/pope-leo-schooled-the-tech-bros-on-tolkien/
Behind every failed revolution there's an #LLM
It is a sad technology emerging from a sad field of study initiated by sad men in a sad place sat atop a sad hill in New Hampshire. The technology generates sadness. It is soaked through with sadness. So many who use it feel sadness when they do. So many feel sadness when they watch others use it.
LLMs are like the sadness a child would feel if you showed them a beautiful cake and told them it was a real cake and they could have the whole thing, but it turned out to be styrofoam. Nathanael West once wrote that Hollywood is where the American Dream goes to die. LLMs are that sadness delivered to your bedroom (*). LLMs are DoorDash bringing you the styrofoam cake over and over forever while you hope it'll be a real cake each time.
It's OK to be sad. We all are from time to time. But what nuance is there in telling yourself the sadness is something else? Humans have been doing that forever. If there is nuance I think it lies in mourning what the existence of this technology is telling us needs mourning.
#AI #GenAI #GenerativeAI #LLMs #LLM
(*) I assert that reading Nathanael West's oeuvre over and over is better for you than using LLMs.
I know I've used vibe coded accessibility stuff, sometimes accidentally because I didn't know it was AI assisted, other times because I was curious. *sigh*
AI assisted/Vibe coded software genuinely sucks. The quality is so poor that, to me, there is no difference in AI assisted and or vibe coded software. Even my beginning baby scripts are better than software slop. The only vibe coded thing I kinda sorta liked, until I used it more and more, was SonarPad but then I began to realize that, hey, these features? Utterly useless to me, but then there were just obviously missing things I'd think of that would have been revealed in actual testing or people actually using the fucking software they code.
I soon realized I wasn't using a quarter of the features in SonarPad so I then went back to Notepad or my fork of the Accessible Markdown text editor and haven't touched SonarPad since.
The blind community is flooded with software slop, which should say a lot about the accessibility and accessible design of developer tools. If you hate software slop by blind developers, then ask developer tool creators and developer tool maintainers why doesn't their coding IDE work better with Braille and speech?
As for the tools I use...
The conviction that technology is neutral is not itself neutral. It paradoxically functions as the conviction that this neutrality is a good thing so as to positively sanction the technological development of more "neutral" tools.(this is my own transcription from the video; any errors or discrepancies are mine).
He says this in the context of an argument that, especially in digital/online contexts, neutrality is conflated with truth, reason is identified with data and a quasi-scientific objectivity, and these conflations are damaging to democracy.
#internet #dev #tech #LLM #GenAI #GenerativeAI #democracy #politics
These days when I write an email I explicitly add a note at the footer that says something like:
“This email is made and corrected without the use of LLM’s”
I just want people to know it’s me and not a bot
Sad it has come to this
I like Kim and her work on StopGenAI well enough, but this post of hers is unwarranted and unfair in my view, as is the fact that several people chose to block me in response.
If you read the full context, which I link below, you'll see the objection boils down to the fact that I used the word "current" in a critique of LLM-based generative AI systems, in an attempt to be precise about what I was saying. Obviously people should block whomever they want for whatever reason they want, but I am not going to be cajoled by threats and ultimatums like this one into making sweeping claims just because other people want to see that. I've been crystal clear about my critique of generative AI for over 3 years now and my writing on it here stands for itself. I have no record of it but people who know me know that I was also quite critical of the hype around GANs in the late 2010s.
For context, here's my post: https://buc.ci/abucci/p/1782335950.858712
and here's Kim's objection: https://zeroes.ca/@kimcrawley/116807184909806502
which reads:
If it weren't for your "current" modifier, your comment would be great.For reasons I don't understand the thread rapidly degrades from "this post would be great without that one word" to "you invited a murderer into my house": https://zeroes.ca/@kimcrawley/116810548823823670
#AI #GenAI #GenerativeAI #LLM #tech #dev
If instance admins allow AI Agents on their platform and they keep harassing us I have no other choice then to silence that instance
Again, I do not pay these massive costs each month to host robots
Let's keep the #Fediverse human shall we? 
Before working for 2 years on the Apache APISIX API gateway, I was mainly oblivious to API gateways. It’s only by working with them that I understood their value. Decoupling the client and the server unlocks a lot of options: moving authentication to the API Gateway, securing APIs, deduplicating API requests, etc.
In this post, I want to describe how the same pattern applies to #AI.
#AIGateways work in a similar way.
I created an issue on the plugin's GitHub about removing the MCP functionality into a separate plugin, and the maintainer immediately converted it into a discussion, which is here: https://github.com/coddingtonbear/obsidian-local-rest-api/discussions/272
I'm getting the impression from all the LLM-related discussions and commits that this is a lost cause and I'll probably have to fork this plugin if I want an MCP-free version of it. Looks like they took up slop coding earlier this year too, which is another area of concern.
#AI #GenAI #GenerativeAI #AgenticAI #LLM #Claude #Copilot #AIPoisoned #NoAI #Obsidian #MCP #REST #API
New article: I feel tricked by AI - why I won't be using it anymore.
https://stusarticles.com/i-feel-tricked-by-ai-why-i-wont-be-using-it-anymore/
RE: https://infosec.exchange/@mttaggart/116830015225440372
This is a really important read for understanding both what's under the hood with LLMs, but also why it is impossible to secure any #LLM.
In the thread, I posted a couple of extra links to experts explaining that last sentence better than I can. But read the article Taggart linked first.
From the linked article: "The Oumi analysis also found that 56 percent of the correct Gemini 3 answers couldn't be backed up by the sources Google linked. The AI is giving answers whose origins users can't trace." This technology is not fit for purpose.
Frankly, given how much is at stake, I find it unacceptable that multi-trillion-dollar corporations are playing word games with the legal system. They should release their products on Steam and XBOX if they're for entertainment only and get them out of search engines and office suites, tools people use to do real work every day.
Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers: https://the-decoder.com/landmark-german-ruling-declares-googles-ai-overviews-are-googles-own-words-and-makes-it-liable-for-false-answers/
#AI #GenAI #GenerativeAI #LLM #Google #GoogleGemini #Gemini #AIOverviews
Hey, can I get some legal experts in here to tell me I’m wrong about what I think this ruling means? Also, how does international case precedence work?
A #German court has ruled that Google is directly liable for what its #AI #search overviews say. Previous case law shielding search engine operators from liability doesn't apply to AI overviews.
That’s freaking massive. Google’s AI responses are wrong almost 10% of the time, make up sources, and infer facts not in evidence, and cause real harm. Germany says publisher immunity does not convey when the company product, the ai, is stating things as fact.
Losing publisher #immunity is a really, really big deal. Especially if we can get a similar ruling in the US, and if this ruling flows into EU precedent.(I don’t know how any of that works)
In any case, go German law writers.
#infosec #truthiness #llm https://the-decoder.com/landmark-german-ruling-declares-googles-ai-overviews-are-googles-own-words-and-makes-it-liable-for-false-answers/