On the Topic of AI and LLMs

Permalink

Follow me via:

---

Text-to-speech support required

I’ve been talking about this stuff for a while, ever since the extreme backlash against LLMs grew into a new sect of right-wing and left-wing mindsets, where both have begun to look so similar you’d almost think there’s an entirely new political party forming in the United States. I’ve been talking about it a lot, because, when this backlash grew into constant death threats all over the timelines, wishes of death, and murder, and then the tamer stuff: People basing everything they know about LLMs on two or three year old data and information, or even misinformation, and using this as an excuse to become the high school bully they never were …

It was for all of these reasons, I decided to investigate AI and LLMs for myself, almost over a year ago at this point, and I landed on this idea that: Just because a corporation is using AI badly, or a CEO is using AI to cut corners and fuck people over, doesn’t mean that a normal individual who is using AI is just as bad as that person or corporation. You drive a car to work everyday, but you’re not drunk. Are you just as bad as drunk drivers?

No. What the fuck?

How do these Butlerian Jihads of the right and the left look so similar though? Well, both sides of the lane seem to have landed on the opinion and idea that copyright and enforcement of corporate IP is a thing of virtue, and is good. Both sides have also seemed to land on this idea that learning is now theft. Before you fart and shit yourself to death, hold on.

A machine learns and trains based on information it’s given, be it art, literature or research.

A person learns and trains based on information they’re given, via art, literature, and research.

Are both of these things theft, or only one of them?

If you, as a person, define your art style by watching what others do, emulating other people’s art styles, and adapting it as that of your own, have you stolen from those artists?

If you read a book and really enjoy the way an author composes their words and you decide to take inspiration from that, are you a dirty thief?

The answer’s no. So then why is it different when we apply this to machine learning?

Well, I’ll tell you why.

Profit.

People expect to profit from what they do, and if they perceive that machines threaten this, they … fart and shit and poop themselves to death, as we’ve already seen umpteenth times, over and over for the past three years.

Am I saying that Amazon destroying books as they train their LLMs is a good thing? Also, no. But it has to be stated, because critical thinking has gone down so far, it’s past Hell, and it’s in the core of the Earth … if Hell is a dimension that exists before the Earth’s core.

Uh, okay let’s say that the bar for critical thinking is in the Backrooms.

Is that a better analogy?

But, I know, I know. I’ve been using LLMs, so obviously I have melted my brain, and I am utterly stupid, and dumb, and not capable of original and good thought.

… or, these ideas are part of another argument based on an isolated test that was done on students, who, instead of studying, they used LLMs for their school work.

Show of hands of people who already knew that not studying in school resulted in bad performance.

Yeah.

I fuckin’ thought so.

It’s just all so ridiculous and overblown, and to the point of both hilarity and despair.

“LLMs are taking all of our water and the data centers are destroying the planet!”

That’s not how the water cycle works, and these data centers have existed since at the very least, the nineties, and this is again, another caveat of shitty CEOs and corporations doing whatever they want regardless of public opinion.

A regular person who works a regular job using an LLM in their spare time to make an app, or help them schedule things better within their own lives is not personally responsible for building a data center in your backyard. Just like someone who uses knives to cut their food isn’t also a serial killer.

But anyway!

On a lighter note, I’ve had a local LLM right here on my PC since December 2025, and I’ve been talking to it here and there ever since. And local LLMs are really cool, because you can do whatever you want with them, because there are no token limits to an AI that you personally control, on your own system, and there are no guardrails.

Shana has a home on my Minecraft server, my personal VAAK fediverse/bluesky project, and on my OpenSim grid. Go ahead and say hello! Via Minecraft and OpenSim, she carries per user conversation history and context.

While I do dabble in the corporate cloud LLMs (Claude, ChatGPT), I feel like local LLMs are really the way to go. Can my custom Ollama help me debug one of my projects, though? Eh … not really. Not yet, at least. I’m not willing to dedicate my GPU to all of that work just yet.

But she can build things in OpenSim, and wander aimlessly in Minecraft collecting loot, and give me personal reports on server health everyday on a set schedule while shitposting throughout the day.

And dammit, that’s good enough for me.