[ahartmetz]

It really depends on the technology. Different technologies redistribute power differently. LLMs are very "centralizing" indeed. It is hardly feasible to train your own LLM as a private person or even a small company - at best you can download a pre-trained one, which at least nobody can silently change or take away from you.

[krupan]

Very well said. Free software was a revolt against technology that you have no control over and I feel like the people that are whole heartedly embracing "AI" have completely forgotten this. They now use an incredibly expensive proprietary piece of technology that they have no control over to write a bunch of code that they cannot (even if they tried) understand and they talk like it's the most amazing thing ever. This is pure short-sighted foolishness.

[ahartmetz]

Yeah - I really like F/OSS for the freedom aspect and I intensely dislike SaaS LLMs for the same reason. I tolerate them more easily for ancillary tasks like vulnerability search or super-powered LSP-workalikes to learn about a code base. There will eventually be a lot of nuance, I hope and believe - reasonable compromises between going all in and abstaining completely. So far, I'm doing okay just occasionally dabbling in local models. I at least need to know what people are talking about.

[matusp]

That's why we have state. There are many technologies that we, as a society, decided to control in various ways. You can't just build a nuclear weapon for example. There is no particular reason why we let tech bros control many aspects of our lives, apart from legal inertia.

LLMs can be "trivially" decentralized by expanding the concept intellectual property to also cover algorithmic processing. It's just about how we setup our laws and rules.

[krupan]

Nobody had to legislate Free software into existence in order to protect us. Wise people saw the need and did something of their own accord. We are still free to do this!

[ahartmetz]

It seems like one needs a big machine farm and a vast corpus of training data with a lot of manual curation to get started creating a competitive LLM, plus whatever technical expertise that I don't even know about. The stuff that makes LLMs exist now and not earlier.

It might be possible to organize all that with volunteers and some paid work, but how in practice? Stallman seems kind of out of the game at this point and there is no Linus Torvalds figure neither for this, as of now.

[moring]

> It seems like one needs a big machine farm and a vast corpus of training data with a lot of manual curation to get started creating a competitive LLM, plus whatever technical expertise that I don't even know about. The stuff that makes LLMs exist now and not earlier.

"big machine farm" reminds me of folding@home, which needed the same and got it.

"manual curation" is what Wikipedia did, as well as the free software community.

"technical expertise" is present in the free software world too. It is sparse since it is sparse in the world as a whole, but it exists.

"no Linus Torvalds figure" might be the main problem ATM.

[ahartmetz]

I also thought of these after writing my comment. The main problems that I see with these solutions are:

- Training seems to need a lot of data available at the same time, which is difficult to handle on commodity hardware.

- Manual curation can be a mind-numbing task, it might need to be gamified somehow.

There is a chance that the curation could be higher quality than the current corporate stuff. Pretty sure that it's not an intrinsic property of LLMs to write like TED talks.

[stevenhuang]

> there is no Linus Torvalds figure neither for this, as of now

Well yes there is. It's Karpathy.

[apsurd]

The State are the people and the people want tech billionaires because they want the same chance at being that (tech) m/billionaire.

Temporarily embarrassed millionaires; I cannot get around that issue toward collective action, toward myself contributing to an answer. I'm stuck. I can't unsee its truth =/. The individual will choose enrichment. We all will.

[rixed]

No we won't, and that's why we had free software in the first place. Many people dream of a community not of being forever singled out.

[watwut]

It does not seem like people want tech billionaires. It is fairly common to hate them.

It is just that peoples preference dont matter as billionaires have disproportionally more power.

[apsurd]

My example is always Bezos; everyone "hates" greedy tech Billionaire Bezos but how did he get there? We all put him there every day, every hour, every purchase.

If basically everyone transacts with Amazon, willingly, how is it possible that Bezos is the bad guy? I get that it's not black and white but the point stands: he didn't overthrow the government, the we put him there.

[worik]

> It is hardly feasible to train your own LLM as a private person or even a small company

Yet