After I made this unrelated post, I was informed by @MysticMushroom1776@lemmy.dbzer0.com that lemmy.dbzer0.com was a pro-genAI instance.
It was honestly shocking for me to hear that these two things can coexist and I would like to share the discussion which was correctly moved for being off topic from the above linked post.
I think we owe it to ourselves to discuss topics like this with an open mind.
Here is my statements in reply to MysticMushroom1776, I wasn't able to record their reply before it was removed for being off topic bit I hope they can come here and we can continue the discussion.
Others, please also feel free to discuss with us.
'Wow, I didn't know [that lemmy.dbzer0.com was a pro genAI instance] As an anarchist myself I have to say that is really disappointing.
Anarchist communities [especially in real life] need humans otherwise there is no community. GenAl is a stolen facsimile of a human who can never understand morals or praxis. It can never understand freedom or control. It will however poison our communities (look at what SpaceX/Anthropic is doing in Memphis, TN) make us more predictable (and thus easier to control) and isolate us from other humans within the community which is so important to us.
Even with the most moral genAl you can possibly have, one that is totally FOSS and self hosted on a single gaming GPU, a lot of this is still true.
~~Obviously, this is the wrong place but I have to ask,~~ [prior to moving the discussion to this thread] what makes you think the extreme negatives of genAl (specifically to the anarchist community) can be excused?
... [I do not have a copy of mushroom's reply]
I pirate and seed. I'm also a fan of the megathread And for the most part, I think data is data. However there is a big difference between my pirating a book like the Lord of the Rings (which is credited and I can see the title of) and how Meta scraped Anna's archive and funneled all that data into an approximation of all combined work.
If I pirate the book, the Lord of the Rings to read it. I might enjoy it, I might talk about it to others, quote it, share it with my daughter, I might use the ring of power an an allegory for authoritarian control with a powerful technology, I might even decide to buy my own copies of the book.
What Meta did and then released as open source, won't allow a person the freedom to do any of that. The works were consumed and are used in silence to camouflage as human thought. No credit will ever be contributed. And so no one can even hope to appreciate the original works that was used in training.
The other concerns I have with (local) genAI that I think are deeply important to anarchists specifically.
- Isolation of humanity
- severing of our ties to community
- increasing predictability
- method of control
Photo is genAI made and sourced from the following article:
Sure thing.
I'll live over here in reality where someone having access to a screw is also going to include (in such overwhelming certainty that the negative can only be described as an exception existing in theoretical consideration only) the knowledge of the existence of a screwdriver.
I'll also note you have zero comment on any of my corrections of your misunderstandings about LLMs, so I'll go ahead and move on now.
It's because like the majority of anti-AI people, the "misunderstanding" is largely intentional. To get you to debate them. Hoping they can "win" and when they can't win through their arguments because it's all bullshit they just knock over the chess pieces and declare victory. Which is what this liberal chud is doing right now.
No corrections were necessary as you misunderstood my original point.
But to correct you anyway. Influence on llm models comes not just from instructions prior to entering text (pre prompting). But also from the inclusion, or selective exclusion of training data.
As a person cannot choose what they do not know about, so a LLM cannot respond about what it has not been trained on.
As such influnece exists in both models regardless.
These biases and their ability to influence a person. Due to the inherent sycophancy of ai and the tendency to anthropomorphise which humans have contribute to the phenomenon known as ai psychosis.
Thesetwo factors combine to the "gun pointed at your head that is inherent to LLMs".
Ai also hasn't existed since the 50s. It was theorised, although intelligent machines have existed long before that. An example was the mechanical turk, which was ironically powered by a human.
Anyway. I'm afraid you're wrong on me not being educated on how they work. I know all too much about how they work. I also know how people work, though that always seems to be a blind spot to ai enthusiasts.
Still comes down to those training.
Again, you seem intent on blaming technology for how people use it.
Are you trying to blame a model for how someone trained it? Or the user for using a model without knowing what it does?
How, in the wide wide world of sports, is this the fault of the technology?
I'm afraid its all the more clear that you do not.
Which isn't you. Hence the influence.
But your contention seems to be, I have no problem with the technoclogy merely because it was designed to kill me. That was the intent of the person making instead of the technology itself.
That's rather myopic.
Also, you didn't point out an intance where I "didn't understand" how ai works.
You are sorely mistaken on that point. The models I would say are 'in use' for me are models I've built.
Extremely specific and tailored to their purpose, built from scratch.
I would recommend reading my comments again then. I pointed out quite a few of your misunderstandings.
At this point though, I feel like I'm just repeating myself here. So I'll be clear.
You clearly do not understand the requirements to run an LLM. You clearly do not have any understanding on their history or development.
And most importantly, you are clearly blaming technology rather than the corporations behind the misuse.
Enjoy your day.
Do you host your models anywhere, or have documentation on the methods you used?
They aren't really general purpose, so no.
Though I'm working on a new classification model that may be good for more general use, so maybe that one I will.
As far as the methods, these are readily available - Build A Large Language Model From Scratch covers it pretty well. There's another for a reasoning model from scratch, but its more reasoning from scratch and starts with a base model.
This one?
Build a Large Language Model (From Scratch)
Thats the one
Covers a basic classification model and assistant model. You'll want some understanding of deep learning and python (its not too heavy, but you should have a decent grasp of the concepts and basics of python to understand what's going on).
Edit: On the easier side, you can do something like unsloth and modify a base model. Easier, but heavier requirements depending on what you start with. Good for testing concepts, but highly specialized models are the superior approach over general purpose IMO.
Thanks for the recommendation.
Ahaha. Sure, you selected vetted the model, adjusted weights and also have the computing power to train the model I doubt that.
No, these are custom models for hyper specific purposes. Built, not trained.
Doubt all you want, I could go that route on just my laptop if I wanted.
For the record, the biggest model I made is less than 2GB, and easily runs on the embedded hardware it runs on.
Also, no, nothing like flock or any of that. Ive turned those types of projects down.