jonny (nonvenomous)RE: https://dair-community.social/@emilymbender/116604745957981805
"links will become an afterthought" is not even coded language for "the rest of the internet is merely training data and we will own the entire means of accessing information online"
as Dr. Bender says upthread, the Rethinking Search paper just says this explicitly, also AMP, etc. I only mildly edited their figure here
The shift to "search journeys" is just another way of referring to "whole life immersive surveillance" where the intention is to slowly train you to expect more and more of your personal information to be visibly injected into search results as a surface for "personalization" and eventually move towards "zero-query search" where advertisements-i-mean-helpful-information are proactively volunteered to you.
this language appears in full form as early as 2018 and was chilling even then:
The zero-query search paradigm can be expressed with the slogan “the query is the user.” In practice, the context of the user is used to infer information needs. (Entity Oriented Search)
For I am CJ 
Who gives a single shit how good the "AI" can code when the purpose of "AI" is and always has been enclosure of all information in an ad-driven surveillance platform. If you love LLMs, this is the future you, yes you, are actively helping create.
There IS NO LLM USE not associated with the project to seize all information as a product. That is the WHOLE gamble being made that is driving all those billions into getting as many people as possible dependent on the most preposterously expensive and inefficient model of computing ever devised. It is only worth it if the upside is owning the whole economy.
Every step you take towards building LLMs into your daily habits and work ratchets the spring tighter on the mousetrap until, surprise! It clamps shut while your whole ass is wrapped around the cheese. Don't make me laugh with local models nonsense, if you think that those don't get deprecated the moment they pose the slightest whiff of a threat to the profit model - meta isn't releasing weights to be nice, it's to capture labor and control the tooling space. Don't be a sucker.
Don't reach for the easy dismissals - sure there are other tools in the category that aren't chatbots, there are assistive applications that are genuinely miraculous, and the rest. But the purpose of these technologies and their reality as products and mountains of capital doesn't have a nuanced "both sides" placation at the bottom of a thinkpiece: they are designed and built to own informational reality.
Oh wait no they just posted it and immediately blocked me but the post is still up. Well that's sad that that is what they chose to go with to block dunk me to their followers
Extremely funny that someone out there believes that
- opposing a technology of control whose whole selling point is supposed to be "less work" is being lazy (?)
- that by choosing to keep working in a way that is supposed to be harder means I cry about needing to work (??)
- that by working hard lazily to survive at a university without advancing the push for dominance being pursued by the largest companies in a world I am endorsing corporate slavery
- that they in fact have the moral high ground because they are using the corporate technology to cause me to be unemployed
And the correct moral choice is to oppose the bullshit that google and Microsoft use the AI for but completely separate that from the complementary but similarly motivated play made by corporate angel Meta and Moonshot and embrace the exact same thing wholeheartedly.
Its rare that someone says such an incredibly incoherent thing that you have to stop and untie the gift that they puked on your doorstep, but what the fuck did you just say to me listen here you little shit...
"You're fucking bucko'd kiddo" is now preventing me from sleeping
cholling
mathew@cholling @jonny "Right. Bananas. How to defend yourself against a man armed with a banana. Now you, come at me with this banana. Catch! Now, it's quite simple to defend yourself against a man armed with a banana. First of all you force him to drop the banana; then, second, you eat the banana, thus disarming him. You have now rendered him 'elpless."
thomkennon@jonny this dick has ZERO bite. Don't fret a minute. Solidarity jonny! 🖤✊
happyborg
veebs
Olivier Leroy@jonny thanks for your service!
Pieter Kuiper 🌍🇳🇱@jonny Is that what an AI-generated rant sounds like?
Bee O'Problem 
George B
Ani, fury unbound@jonny that post wasn't deleted, you just got reply-and-blocked
Raghav Agrawal@jonny The Navy Seal copypasta one is so funny
@impactology
I need to save it to my notes app so I can be quicker on the draw, this keeps happening
I need to save it to my notes app so I can be quicker on the draw, this keeps happening
@jonny textbook BPD
webhat
Alex
JP
Nini
joshbuddy@jonny I don't think I follow what you're saying about local models. Isn't the horse kind of well out of the barn at this point re: local models? It's not like some entity can take them away, and the best open models don't even come from Meta etc.
Murat Kaan@joshbuddy @jonny Plus LLM advances kind of plateaued, so a local model good enough for now will likely remain so for a while
@muratk5n @jonny It would be a very interesting world where large companies started coming after open weights. Seems wildly impractical, sort of like the govt coming after open computing, but again, would be an interesting world. In some ways, and not to seem overly accelerationist, I'd like this world. Would be nice to make piracy and open computing cool, get people interested in actually owning the means of computing.
And agreed on models seem like they're plateauing. Probably the best open models today are good enough for a long time. And of course, there could be some community effort to start re-creating open weights like what occurred in the world of Go.
@joshbuddy
@muratk5n
They are in the process of forcing age verification into Linux, no? It does sound impractical and difficult and like a wild overreach of any sensible bounds to government authority, but that doesn't really seem to be a deterrent.
There is a certain romance to a post apocalyptic cyberpunk dystopia, but all things considered, I would like for things to suck less and there be less extraction and suffering in the world.
Let me know when you decide the weights are "good enough" and you voluntarily stop updating, it would be a good test of theory I suppose.
@joshbuddy @jonny "there could be some community effort to start re-creating open weights like what occurred in the world of Go"
Exactly. Community effort, China who see open weights as a way to break the Western monopoly on LLMs.. There will be options, I am optimistic.
@joshbuddy @muratk5n Saying "open weights" is like saying "open binaries".
@mathew @joshbuddy Well, better than "no binaries" "you only access me through a cloud"
@mathew @joshbuddy There really isn't much to this tech, companies are exaggerating what it does, good or bad, for marketing purposes. There wont be an AI apocalypse (they are too stupid for that) and no one is going to monopolize it. The technology will be available to all, free or for a price.
@joshbuddy
What is the user share of local models vs hosted? Total guess, but I'd bet its less than 1%. How high would it need to be before they shut off the spigot of the weights they spent millions to train? 5%, 10%? I don't know but they don't pose a threat to the profit model now and building organizations and projects around LLM tooling of any kind certainly isn't making a world of independent information sources and work patterns any healthier.
If the AI companies stop releasing weights, how long do you think people who are using them in tool harnesses designed for the latest models keep using them? Do you think the largest companies in the world could afford a lobbying push to treat open weights as scary foreign weapons? The government bought TikTok, why would it be off the table to push Deepseek weights off huggingface and into the realm of piracy? Piracy is also a cat out of the bag, but it's not a threat, streaming companies are doing fine.
The idea that local models are on a course to catch up to or overtake the commercial models and then global capital will have to just do a Charlie Brown sulk back to hell sounds pretty far fetched to me. Given the course of corporate involvement in open source so far, it seems more like a temporary tactic in an economic and political war for dominance without any guarantee of longterm stability.
Also almost every time I see local models mentioned it's someone using them to derail a conversation about the harms of LLMs. Makes them a very useful tool for those big companies to keep doing all the damage they want and then just have someone say "but local models" when the companies' extremely antisocial policies are hilighted.
@gbargoud
@joshbuddy
Buying rhetorical cover is extremely cheap, actually, and people will just do PR for you for free.
@joshbuddy
Buying rhetorical cover is extremely cheap, actually, and people will just do PR for you for free.
Anarcho-Daddy@jonny @gbargoud @joshbuddy burn the data centers and keep the local models, the whole issue was always centralization of control imo...
@gbargoud @jonny So many of these conversations seem conceptually mushy to me. The benefits and harms of LLMs can be discussed separately from the benefits and harms of big tech.
Feels like the harms of big tech are very obvious and keenly felt whereas I'm still uncertain about the harms of LLMs specifically. They seem to make specifically crime easier and contribute to skills atrophying.
I'm trying to get clear on this stuff myself, it's not easy.
Jack 
@joshbuddy is anyone making LLMs who arent big tech? are there ways to train these things without billions in energy and data scraping? if so why are the big players paying these costs?
@jackperkins yeah, that's the problem with this "big tech" label, what does that even mean? I'm using it as shorthand for big tech companies that are using their technology to harm consumers and extract value out of them. Like Amazon, Google, Meta.
Arguably Mistral is not "big tech". Arguably DeepSeek doesn't fall until this either.
I just helped my partner run a campaign against big tech here in Denmark, and charting a course around this question of what is and isn't "bad" in tech .. yeah, its complicated. She got a lot of criticism for recommending Signal for instance, but in thinking about the harms of big tech, I kind of keep coming back to the book Surveillance Capitalism. Had lots of flaws but useful in thinking about how these companies extract value in really sinister ways.
@joshbuddy i understand that mistrail and deepseek are not yet google/ms/apple scale of dominance but do their products costs billions to produce? if so they will need these positions in the end to recoupe their investments. giving away the result of this investment as a free trial doesnt mean its open or public, in the way something like the linux kernel arguably is despite significant investment (tho orders of magnitude smaller than LLM)
maybe in 5-15 years some clever computer science will reveal ways to make these things in your bedroom, and which point i will no longer be concerned with enclosure or social control. but until then its a bit silly to make a split between how something is produced and the people who are producing it. lets talk about what actually exists!
@jackperkins yes totally. these things are free as in beer, not open source. not clear to me that deepseek is looking to re-coup, seems more like a ploy to undercut american companies, so, they might not be looking for a return on anything.
but if models are plateauing and these open models are good enough, we don't need to be overly concerned about them being a tool for control, people can and do run them locally.
i was thinking about the opening to @pluralistic book "The Reverse Centaur's Guide to Life After AI" and he makes the distinction that ML is useful and one point of cleavage between people who like them and people who despite them is agency. The people who like using them had a choice in the matter. But we're seeing how LLMs and other shitty tech is being foisted on people with no choice.
I had a fun talk with a friend at FB and they told me that now managers are expected to have 50 direct reports and manage the workload through LLMs and everything is miserable. I hope they implode.
Mx. Chara Aznable (they/them) of Pnictogen@joshbuddy @jackperkins "Arguably Mistral is not 'big tech'. Arguably DeepSeek doesn't fall until this either."
False. Mistral is a metastatization of big tech—it was founded by Google and Meta people and therefore surely was founded with the same diseased capitalist assumptions that pervade all tech companies. DeepSeek was founded by an entrepreneurial grifter whose purpose for exploiting AI is stock-market gambling—there's no more corrupt purpose for exploiting technology than trying to swindle money through market gambling.
@mxchara @jackperkins I mean, I think this gets at the definition of "big tech" but I certainly don't agree with your characterization, by this definition pretty much every start up is "big tech", so the label ceases to be useful.
@joshbuddy a million is one thousand times smaller than a billion. many startups received handfuls of millions of dollars in investment, whereas openai burns something like a billion a month in operating costs.
what is the cost mistrail or deepseek paid to produce their "open" models?
@jackperkins to be clear, I would certainly call OpenAI big tech, and I'm not at all a fan of what they do. And granted that creating any large language model is probably capital intensive because of acquiring the source material and the GPU and energy costs.
Curiously, it looks like at least one group is working on distributed training, so more like how the OpenGo weights were made
https://www.primeintellect.ai/blog/intellect-1
Mistral looks like they took very traditional funding. DeepSeek looks like they are funded out of a hedge fund.
Gosh, this is all pretty depressing. Technically, it should be possible to make a completely open LLM, like the INTELLECT-1. I know that money in tech means we'll eventually get screwed. The only silver lining in this is at least the open weights are free beer, no one can take them away and they are useful for people.
After reading Winners Take All I just wanted to exit tech. It's hard not to look at the direction of all this and just wish for a time machine or something.
INTELLECT–1: Launching the First Globally-Distributed Training of a 10B Parameter Model
We're excited to launch INTELLECT-1, the first globally-distributed training run of a 10-billion-parameter model, inviting anyone to contribute compute and participate. This brings us one step closer towards open source AGI.
No it is not possible to separate the technology from the context of its production: https://neuromatch.social/@jonny/116606065797032036
Here is my attempt at providing concrete history and context: https://jon-e.net/surveillance-graphs/
You can stick around for the whole thing or just skip to the punchline: https://jon-e.net/surveillance-graphs/#the-near-future-of-surveillance-capitalism-knowledge-graphs-get-chatbots
jonny (nonvenomous) (@[email protected])
@[email protected] Except that its entire design, technological lineage, and deployment is inextricably bound to its nature as a product developed for this purpose. As far as you can reach in the tech stack and prior art that leads to LLMs you will find information companies whose business model is informational enclosure dumping oceans of money into research on how to better enclose information. Not every tool is a hammer, there are in fact tools to which analogies that apply to one category of tool do not apply. It's not some magical bush out there in the woods that we would have inevitably come across. It's not a byproduct of physical law. It was *made* for profit and control. You can imagine any number of different directions that information automation could have gone if not designed as a product given this particular web platform landscape (that was itself designed and built, rather than natural). I don't know, spend even one billion dollars on a p2p information system where people can better co-curate information and social spaces so it can be found, conversationally, on demand, through the magic of other people. Spend even one billion dollars on foundational coding tools and frameworks that avoid the need for generating boilerplate altogether. The idea that technologies are morally neutral and it is only their use gives them moral valence is only coherent if ignoring their history, and maybe more importantly, the many alternate futures their existence foreclosed.
@jonny @gbargoud I don't think I understand the point "No it is not possible to separate the technology from the context of its production"
Isn't it safe to say that the tech has escaped its original intended purpose? Doesn't technology escape its original intended purpose all the time? This was kind of my takeaway watching James Burke and Connections.
@joshbuddy
@gbargoud
No it has not escaped its intended purpose. Its purpose was to provide a ubiquitous any-machine that cultivated dependence as a necessary passage point of information that once reaching saturation can be exploited for stupendous profit in a multi-sided market selling access to consumers, attention to advertisers, influence to.... and so on. It is doing exactly what it is supposed to do.
Read da linked piece, its much more coherent and comprehensive than I'll be here.
@jonny @gbargoud I did read what you linked to btw, I was kind of hoping for more background and argumentation. I mean, I think it's an interesting argument "LLMs are made for info enclosure", and it would explain some of my sickening feelings regarding them, but I'd like to see some more facts rather than assertions.
but maybe expand this into something a little more full? then you might have something here.
@joshbuddy
You stayed up all night, read all 26,000 words with 276 references and concluded that it didn't have enough facts and wasn't full enough?
You stayed up all night, read all 26,000 words with 276 references and concluded that it didn't have enough facts and wasn't full enough?
@jonny I'm obviously looking at the wrong thing, sorry, I thought you meant the thread itself.
I see now you meant https://jon-e.net/surveillance-graphs/ Can't wait to read it! Thank you
Surveillance Graphs
Vulgarity and Cloud Orthodoxy in Linked Data Infrastructures - A critical history of the semantic web and linked data, grappling with the next generation of surveillance capitalism, where grand corporate knowledge graphs devour the planet and sell it back to us as a glassy-eyed LLM personal assistants, will we remain stuck in the ideology of the cloud, or can we have better dreams?
@joshbuddy @gbargoud @jonny "The benefits and harms of LLMs can be discussed separately from the benefits and harms of big tech." false
J. R. DePriest 
:EA DATA. SF:@mxchara @joshbuddy @gbargoud @jonny
They can discuss "the benefits and harms" of LLMs and big tech separately but it is disengenous to do so.
To me, it's about intent and about scale.
Big tech created the technology. Big tech went out of their way to fire the people who brought to any ethical concerns while doing it.
LLMs would not exist right now in the way they exist without their hyper-capitalist, anti-human, above the law, surveillance mindset.
It doesn't matter nearly as much what outsiders, anarchists, and open or free software advocates do because the horrible businesses that made it possible are still in the middle of killing the planet, putting profits over morality, rushing to embrace authoritarian fascism, and powering autonomous people killing machines for the government.
When you say "LLM" or "AI" or "generative AI" 97% of people think about ChatGPT, Gemini, Claude, Anthropic, OpenAI, or Google. They don't think about people training or running their own models on their own hardware. They didn't think about open weights.
You can't separate the two right now, not while the bubble is stealing water, polluting the air, consolidating power, and excusing mass layoffs. We are being put in our place and LLMs are the engine.
(roll m3tti)@jonny i've lost a lot of faith in humanity due to this and also all the other incidents that happend like edward snowden, epstin, cumex and all other stuff i don't want to get into detail. But this showed me the public is not interessted in what happens to them and who controlls their lifes and yet i have no clue how to tackle this problem besides better education. But this is also rigged :/. Cyberpunk here we come
@jonny but than there are also folks that just do awesome work and don't let this bad stuff in their lifes like all people #permacomputing or the #noai movements around the world. So yeah i guess there will always be a resistance. The sadest part is that most of my tech "friends" or people that are renowned for tech expertise praise the ai stuff which i totally not get. Due to the fact that its not intelligent. I thought always that people in it are quite smart and know a little bit of math but i might have been misslead. Those statistic systems will and can't be intelligent due to the nature auf statistics its just getting somewhere near but never exact. But somehow i have the feeling that people always go for the solution that is "good enough" which totally is covered by statistics.
Cassandrich