It’s really super simple what turned me against it. The absolute torrent of mindless garbage. Blogs and articles, and videos, and people who can’t hold a conversation without saying “Hang on, let me ask ChatGPT”, garbage code that people submit and demand I fix because “Claude said it should work” only to find out they couldn’t explain what it does or why, let alone figure out how it works themselves. The absolute sea of AI generated spam, the robo-calls, the infinite, endless baying of, infinite idiot electric sheep. People who literally cannot read without a chat bot reading it to them.

This is hell. This is hell and every single AI is an legion of demons, and all I have to slay it is the lump of bacon-flavoured jello with anxiety between my ears to slay all the trash it generates.

What should we do about it? I’m not putting that on the internet. Nice try though. If you want to fix it, you fix it by abandoning the digital wherever possible and living in the real world. Touching grass, talking face to face with real human beings. You’re not going to find the solutions in digital spaces. You find them in the faces and hands of your neighbours.

I got a degree in data science cause machine learning is cool as hell. Interesting math that does cool shit, if a bit fiddly. We can process cancer scans faster and more accurately than ever before!

Then the fucking crypto bros showed up and thought you could use it to replace literally anything, stole the entire contents of the Internet, and gave us shit quality tools (validating my opinion that the tool is support, not replacement), and are destroying the planet in the process.

And I’ve given it a shot. Only tool that was useful was Claude and I kept it on a tight leash. Every app based on AI (gave up distinguishing LLM long ago) hasn’t been been worth the bits it was written with.

Edit: oh right, that thing where it’s also making us dumber. That’s in there too.

I’ve been playing around with chatbots since they arrived in the mid 90s. It was fun.

I tried the first iterations of the new LLMs and decided they too could be fun, but are useless and untrustworthy for real life situation whether personal or business. Definitely not worth the cost to the environment, society, or the nature and wielding of power in the world.

What cemented it was when someone wrote, paraphrasing here, that LLMs are going to be like asbestos. Initially seeming like some kind of wonder solution to all kinds of problems until we realize it creates much worse problems. And like asbestos, the abatement of LLMs will be long, slow, and costly. And will ruin a lot peoples lives in the meantime.

It’s less about the AI and more about how the AI is used by folks. It all boils down to how it lets people who don’t care about the quality produce way more than it lets people who care about quality produce, even as it helps both.

My experience browsing video content to discover new stuff is pretty much ruined. Folks that really don’t have a good idea prompt up crap. AI speeds up folks that don’t care about quality orders of magnitude more than it speeds up people that care about quality, so the flood favors low quality. Then there’s the knockoffs. Something looks like a creator that I liked and I’m not paying much attention and then 30 seconds in I’m wondering why the hell it’s so soul meltingly hollow, then I catch on it’s an AI knock-off riding the wave of a more popular channel.

In software development, folks that formerly trusted in the developers to do the job as they see fit now think themselves experts on software development. They prompted up “hello world” type fodder and now they are micromanaging folks with decades of experience. This happens to some extent with every tech fad, but this one is just way worse. Then there are people who have basically been pundits, high on opinion, low on actionable anything. They’ve always been annoying and pretty much wrong, but you could pretty much tune them out because everything they said amounted to nothing. Yeah, it’s annoying that they get paid to basically talk the way executives like, but at least they didn’t actually impact anything. Now they prompt up their bad ideas into pretty bad concepts, and there’s increased demand to pay attention to “almost realized concepts” that they convinced their executive friend looks good.

In terms of what to change, basically I don’t see a good way other than going back and just not figuring out this tech in the first place. I don’t see a path for leveling the playing field to penalize slop, for the narcissistic know-it-alls to go back to being on the sidelines.

I’ve always held this view from the beginning, there was no catalyst.

Ai is trained on stolen data, and poses an incredible environmental risk. It outsources critical thinking. Everyone I know who uses it regularly has shown significant cognitive decline in the last few years because they’re not thinking for themselves anymore.

How do we get to the point of not having ai? Pretty easy. Stop using it. Don’t buy products that have it, don’t use products that have it, don’t use software that is written with it. Boycott it entirely. Honestly it’s not hard

Im not against them in particular but have only seen limited positive value from them. A lot of money and resources are being spent based on some dream of some future potential they may have when it does not make sense economically, socially, and environmentally. Certainly it should exist and be worked on but the whole race thing is bad. I use them as an alternative to search but check the references it gives me.

It’s destroying software because juniors are no longer using their brain to learn.

It’s destroying intelligence because most people will use it as an external brain.

I’ve had my fanfics scalped by them, and they had the fucking audacity to leave ‘I used an AI detector on this fic, heres the link to the detector’ shit comments too in their wake.

A LLM can only regurgitate whats put in. It cant craft a story with a deliberate theme and subtext. It cant choose to evoke certain ideas and thoughts.

Some shitbrain a while back decided to go on about how writers would be out of a job because of it, when the mess was first starting. Still wish i could go back and blast him a new one over it.

This is not how I reached this point, but it is an example of an experience that would get me to this point.

Earlier today I was watching a video on passive ventilation. It talked about windcatchers and such, which I had already known about, and showed videos of them. It then talked about medieval european ventilation designs, which I hadn’t known about, and showed videos of windmills? I kept watching the video until the four-minute mark when it showed a diagram of a house with some depictions of temperature flow. I noticed that the image was oddly grainy, weirdly generic, and oh- this is an ai image!

People in the comment section had noticed this as well. Apparently it became more obvious later in the video. The imagery was generated using ai. Also apparently the voiceover was ai generated as well.

I fact-checked some of the information given and it was just wrong? There wasn’t even much information in the video, and it was still wrong! They didn’t even give specific information but their general information was still wrong!

Tldr: a burger that on the outside appears to be a burger, yet on the inside is merely grey goo

I think that popping the ai bubble would be a good start. The person responsible for the creation of this video should have needed to pay more money in order to churn out misinformation from the banality machine. If I had the power to shift everything as I wanted, I would politely suggest that we construct a society based around the value of human experience. Moving away from capitalist systems (and not towards something worse) would help with a lot of similar problems in general I think.

The couple times I’ve actually used LLMs, it was comically wrong.

I asked it to write some pyqt code and it made up things that didn’t exist. It wasted my time and was worse than useless.

And then I tried a Starcraft 2 replay analyzer/coach LLM. It told me 4 spores (anti-air defense structures) was too many to defend against 2 banshees (air to ground attack airplane). In the next sentence it told me I needed 2 per base, of which I had 4. So it told me 4 was too many but 8 was the right amount.

Now I turn it off everywhere and avoid it. My brain works and I don’t need computers to lie to me.

I work in IT for a tech company that is embracing AI to an extent. We have a ticketing system for managing all the work we do. Tickets can range from simple tasks that take literally 5 minutes to complex multi-part collaborative projects that take weeks.

Recently I’ve been getting tickets that read like a miniature novel, with formatted sections labeled things like “summary”, “acceptance criteria”, and so on. It’s all clearly AI generated, and it makes it difficult to understand exactly what is being asked. I don’t need (or want) three pages of word salad when the requestor could have simply said “write a script to alert us when the API key ‘X’ is set to expire in 7 days”.

When I see tickets like these the first thing I do is go back to the requestor and ask them exactly what they want from me. Often times what they want is not what it sounds like their AI generated ticket is asking for, or at least specific details are wrong. And I have absolutely no problem wasting their time when they waste mine with this AI slop.

• LLM is good for language-related stuff. i.e., translation(also convert legalese to common street), summaries???, removing noise words from huge texts.
• generative models are suped up photoshop-bandaids

those are good at what they initially are made for.

• LLM for search: nobody asked for.
• LLM for ask jeeves?: search has already been doing this. website localized search is even better.
• LLM for writing: only for shitposting.

the marketing and shoe-horning around it are BS. if i can change something, i wish the improvement was more on protein-folding, weather prediction, farming and logistics.

I dabbled with ChatGPT when it first came out, then this happened:

https://www.npr.org/2023/02/24/1159286436/ai-chatbot-chatgpt-magazine-clarkesworld-artificial-intelligence

That was the catalyst that turned me against it. I’m a pretty big reader and I love finding new authors and stories. Knowing that people were cheating the system to make a quick buck did not sit well with me. Knowing that a scifi magazine that has published some of my favorite authors received so many bad submissions that they had to shut it down for a while made me angry.

Then came the information about inaccuracies, environmental issues, AI psychosis, and it being forced into Google searches made me not want to have anything to do with it. I was so disgruntled that I switched every one of my default search engines to https://noai.duckduckgo.com/ so I can avoid it.

One of my biggest gripes is that people use it for an easy quick solution but do not verify the information is correct through other sources. I have a friend who is very into AI, to the point he told his employee that he must use it during work hours when trying to find solutions. I told him, if my boss said that, I would start looking for a new job immediately.

That said, I’m not entirely against LLMs in specific circumstances. I can see how it’s useful in research spaces, but it should be triple checked by a human. Or in creating documents and summaries of meetings, but again should be triple checked by real humans.

I don’t trust the output, I don’t trust the companies behind the output, I just don’t trust it.

Two main catalysts:

1. Seeing how dumb 99% of products are.

2. Seeing how dumb 99% of users are.

Call my naive, but I truly did not see it coming. I didn’t expect cognitive surrender. I didn’t expect the whole world to start continvoucly morging. I didn’t expect lawyers to submit fake case citations, again and again and again.

I didn’t expect society as a whole to just shrug their shoulders and decide that accuracy doesn’t matter.

I also didn’t expect how much or how rapidly the bubble would inflate. GPU prices were already crazy, but now RAM and even SSD prices have roughly doubled in the past year, with no end in sight, with all production toward data centers. This is a disaster. We’re seeing higher prices for downgraded machines, like Microsoft’s 8GB Surface.

Seeing as we’re all on Lemmy, I hope that I will not need to belabor the point that centralization is bad. The shift toward data centers and away from personal devices is a shift toward centralization. It’s a shift toward greater censorship and away from freedom.

It’s not coincidence that this is happening at the same time as fascism is rising all around the world, that online ID laws are passing all over the world, that privacy-protecting technology like VPNs and end-to-end encryption are under greater and greater attack, and that knowledge repositories like The Internet Archive and Wikipedia are under attack. The shift is toward governmental control of the Internet, of access to knowledge in general.

Now the US government gets the final say on who will have access to ChatGPT 5.6. Surprise, surprise.

I’m not anti-AI or anti-LLM per se. I am anti-corruption and anti-bullshit. I am pro-consumer, pro-privacy, pro-individuality. In practice, that means I am anti-AI. Or more generally (but less strongly), I am anti-cloud.

As for what I would do: well the real solutions are the same as the solutions to most of modern society’s problems. We need extensive economic reform. But barring that, we need to do whatever we can to shift the balance back toward personal, private computation. We don’t need more datacenters. We don’t need trillion-parameter models at all. The only thing they are good at is basic stuff you don’t need them for (unless you’re an idiot), and generating well-masked bullshit.

Edit: Oh, and I would also hold all the corporations accountable for their obviously-illegal behavior, like pirating all the copyrighted material in the world. We definitely need more transparency in terms of training data.

When I watched about 5 maker videos in a row. Each one was struggling with a build. Then they all say “now until this point I had being using AI and realised it was lying to me/from the start. After searching the web for 20mins I found the answer” I genuinely said wtf out loud.

Shamefully I’ve also done this myself.

Its gernly not faster to seek information via LLMs even when they give you links to the data they’ve referenced.

I’ve also seen a high level exec use it to make a 10year department plan in the 5 minutes before a meeting.

Also both staff and managers using it for their relevant parts of performance reviews. They could just skip the review altogether and save time.

One thing we can do is stop calling it AI. It’s just a LLM and complex guessing machine.

Somethings (like performance reviews) need to be explicitly declared to NOT use AI.