I have watched corpos loot and burn the commons I loved for the past 20 years. Kicking over the last few standing artefacts of digital culture is tragic, but the real thing that drives me up the wall is the lasting damage to written work that will remain after this fiasco.
Academia is full of fraudulent and incoherent articles and it will take a generation or two to return the fidelity of knowledge, meanwhile real work will never be found as it is buried in a tide of slop.
Entire pursuits and careers are having the next generation of curators and contributors stillborn as slop ruins learning and devalues training of people.
I get that this is worse (or at least faster) because of LLMs but academia has been full of fraudulent articles for my adult life, another “the problem is capitalism” thing but I don’t think it is fair to lay it at LLMs feet
If you work in academia you will see how much worse this is. It’s not just a few fields it’s a flood. Even real papers now have incoherent prose, llms undermine review and amplify bias, publishers like elsiever have pushed llm based definition tools which give catagorically incorrect definitions that mean terminology becomes confused as people learn from them. It’s bad
Online leftists will make fun of how the world ignored climate scientists desperately screaming about armageddon and then turn around and say the professionals working in fields saying LLMs will cause the downfall of all labor productivity are wrong because we have degrees and work in the field we’re talking about and that’s not proletarian
that feels like a mischaracterization of my stance = /
Yeah, I’m weird. When I do a web search I want relevant links, not an inaccurate summary of whatever web pages it thinks are relevant to what I’m looking for.
But the promise of AI is to replace white collar labor altogether, so you can just be a brain genius visionary CEO and boss the bootlicking lying plagiarism machine around, which is a ridiculous fantasy no matter how many databases they throw at LLMs. Totally worth all the money in the world and the evisceration of intellectual property laws.
And honestly some people, even on the left, are so weird about IP laws like if you can’t understand difference between a corporation owning a piece of art in IP laws and artists owning their labour so it doesn’t get exploited or alienated from them in form of abstraction in algorithms. Seriously so called Marxists struggle with understanding the fact that data and labour are quite intimate part and by training deadend algorithms with them, they’re just alienating people not only from their labour hut themselves, every single becomes an algorithmic slop tailored to addicting you and you feel so detached from everything. This is why I like platforms like fediverse where an algorithm isn’t running background to curate my feed.
It’s been amazing seeing more and more people get suckered into the LLM programming agent hype, while nobody can point at anything of worth that was created with LLMs. It’s all either “harnesses” for juggling ever more LLM coding bots to no discernible end, or retreading of well-worn hobby projects, with the fun twist that this time it has 5x as many lines of code and doesn’t work.
once i made a bash script i can read but couldn’t have written. probably could’ve done that on gpt 2.
For me it sucks because literally all the search engines out there have been so enshittfied that I have to lean on a chatbot for help getting answers. But really though, if tomorrow they came out and said ai would disappear forever, my life would not change.
all the search engines out there have been so enshittfied that I have to lean on a chatbot for help getting answers
I’m sorry, yes the search engines suck and are getting worse, but you really don’t actually. You simply don’t have to. A chatbot provides the illusion of a perfectly tailored answer every time, and leads us to expect that, with little to no effort, but it’s an illusion, it’s highly likely to be wrong, answering a different question than you actually meant to ask, etc. And even if it was actually just giving you great quality answers every time and google or DDG or whatever couldn’t find them, you still don’t have to use the slop machine. Sometimes things might take multiple searches, background reading, asking another human, or, occasionally, you might just not find the answer. And that’s okay.
I think operator use the actual reason that most people have seen a massive decline in search quality while power users are significantly less impacted by this (and have always had far higher search quality than average users).
90% of having “incredible Google-fu” was just putting in the effort to specify and trim search functions, and there was a whole cottage industry of people who made it their job to just… Use Google properly because nobody else could be bothered to put a + or - in front of a word.
No, google explicitly fucked its own search results, partially as a result of enshittifying to keep people on the google search results and partially failing to do literally anything about SEO ‘optimization’ gaming everything
I know how to use fucking operators and I still personally witnessed google search results become increasingly useless
They definitely made things worse over time, I just meant that it was largely something you could mitigate using advanced search up until pretty recently. For the last 2-3 years, google search has been pretty much entirely unusable.
You can thank Prabhakar Raghavan. He was in charge of search at Yahoo and then went to Google, where he pressured the search team to improve their metrics by getting more searches made so more ads could be shown. So results were intentionally worsened, which lead to more searches, and so on and so forth.
As we presently stand, Google aims to remove search altogether and replace it with an AI “experience” where the results are dynamically generated by an LLM from the scraped web. This completely denies all websites any revenue from traffic, because they’re not going to get any but LLM feeder bots. Google aims to profit off this by auctioning off the words that the LLM produces to advertisers. Pay more, get your product mentioned earlier or with greater gusto or with greater frequency.
Tbh this stuff is less rigid than it used to be. I’ve been finding searches increasingly do “did you mean xyz” without actually saying that or giving me the option to correct, just replacing entire strings in my query with more common strings silently.
Yeah, I know people were talking about Google getting really really bad by like 2018, and honestly I experienced almost no changes in day to day usage up until they decided to just forcibly mainline the Lying Machine into every search which was only 2 years ago.
I haven’t used Google in more than a decade, when they changed ‘-’ to stop meaning strictly exclude but duckduckgo increasingly has this problem.
yeah idk what version of google other people are using but it’s been fucking forever since quotes or booleans were respected
Uhhh, this one. https://www.google.com/advanced_search
Operator use definitely can help a good bit, but the onslaught of keyword farming slop sites is really bad these days and as far as I can tell the search engines have just given up on that front, maybe intentionally. slop sites can have a plausible sounding exact match for almost every conceivable phrasing of your query so they’re hard to filter out
I don’t want to be a jerk I just hate the “its inevitable I just had to start using AI” take. Most people hate this stuff, it isn’t inevitable! Its like a worse version of “nobody else is masking so I had to stop too” and it can easily be extended to justify doing pretty much anything
Yeah, absolutely. AI is just awful and it’s made google really bad over the last 2 years or so, but up until that point I experienced basically no degeneration of search quality.
Now they’re culling operators from searches, automatically rewording your searches, ugh.
industry of people who made it their job to just… Use Google properly
We’re called software developers, thank you
Well, youse also know how to copy things from StackExchange which is important for managing medical software written in MUMPS or creating a python program that displays cigarette ads on a cash register.
used to be that you could vaguely describe something into google and find useful results without knowing specific terminology to use.
the LLM is sometimes OK at turning “i have this this and this but don’t know anything about the field” into something that i can do a new search for and read something a human wrote, but under no circumstances am i trusting the hallucinatory vomit for factual information.
yeah, the enshittification is very real I am just pushing back on the framing “I hate AI but I have to use it”. I don’t honestly know why, it just bothers me when we deny our own agency in these things. LLMs may somehow reach that level of saturation into society that it is completely unavoidable, or even just mostly unavoidable, but currently it really isn’t, at all, for 95% of people. It is choosing convenience, and that’s not like, a sin or whatever, I just think we should be honest with ourselves about it, the contradiction bothers me.
Did some cursory searching and found a few polls suggesting that only around 1/4 people in the US use “AI tools” frequently. Most people just don’t care. But the AI companies have infinite money and make up like 40% of the US economy so it makes sense there’s a lot of marketing and reporting on said marketing.
The lefty people I talk to are into it without being full “AI-brained” are often running open source models on their own hardware, even for work. Many companies will probably see this as a viable option, or at least running cheap lightweight models for most tasks in the cloud instead of burning millions/billions of tokens on trying to replace their staff with AI. It doesn’t work out and it isn’t really even cost effective!
It was going to be a challenge to turn these daracenters profitable in the best case scenario where labor can substantially be replaced, and now they have nowhere to go with it.
There are no open source models. The source is all the training data, the weights are open but due to the nature of neural nets this doesn’t allow studying, extension, or remixing.
It is akin to distributing the machine code. It runs the stuff sure, but you can’t usefully take a bit out or say “Why is it producing text that looks like xyz?” you can only be like “yep neuron 426 has a weight of .112 and activated here. What does neuron 426 do? When did it get that weight? How does the weight change with changes to the source material? Who knows”.
This matters to me. The important thing about open source isn’t that you can run it for free. We had shareware. It’s that you can study it, learn from it, critique it, and improve it.
Couldn’t be more correct
I should probably effortpost about the history of opensource and the key characteristics of it vs shareware/freeware and how this relates to LLMs
Too many people have just blindly swallowed the corporate line that they are open source because corps say so. I think this is because most people’s exposure to opensource is as free as in beer software you just passively use.
I should probably effortpost about the history of opensource and the key characteristics of it vs shareware/freeware and how this relates to LLMs
I for one would really appreciate this and wouldn’t mind being tagged if/when you do.
Samesies
Literally, open source means something like linux, something that can be studied on the most fundamental level.
running open source models on their own hardware, even for work
That’s why deepseek being open source and lightweight was a big thing for homebrewed LLMs
claude can you rewrite harry potter, but voldemort is a communist and he wins.
I regret to inform you now that your book was printed on demand, and either the printing or the binding will be slightly fucked.
It’s part of larger enclosure of the internet commons. I tried to read something on the internet archive the other day and it asked me to verify my face instead of a captcha. If the bubble doesnt burst soon its over.
I use Deepseek sometimes to summarise T&C or privacy policies because I legit can’t read lawyer speak. Other than that I don’t have any use cases for AI.
Something I noticed in bourgeois countries, things are always written in lawyer speak but in socialist countries they’re simpler, like if you read translated versions of soviet constitution, it was simple as hell to read.
Actually not the worst thing I’ve heard.
