Comment by caseyy

Comment by caseyy a day ago

10 replies

There are many things to be said about open-source projects and, more broadly, the capabilities of the open-source community.

The most capable parts are for-profit organizations that release open-source software for their business imperative, public benefit companies that write open-source software for ideological reasons but still operate as businesses, and a tiny number of public benefit organizations with unstable cash flow. Most other efforts are unorganized and plagued by bickering.

Llama itself is challenging to take over. The weights are public, but the training data and process is not. It could be evolved, but not fully iterated by anyone else. For a full iteration, the training process and inputs would need to be replicated, with improvements there.

But could another open-source model, as capable as Llama, be produced? Yes. Just like Meta, other companies, such as Google and Microsoft, have the incentive to create a moat around their AI business by offering a free model to the public, one that's just barely under their commercial model's capabilities. That way, no competitor can organically emerge. After all, who would pay for their product if it's inferior to the open-source one? It's a classic barrier to entry in the market - a thing highly sought after by monopolistic companies.

Public benefit companies leading in privacy could develop a model to run offline for privacy purposes, to avoid mass consumer data harvesting. A new open-source ideological project without a stable business could also, in theory, pop up in the same pattern as the Linux project. But these are like unicorns - "one in a million years (maybe)."

So, to answer your question, yes, Llama weights could be evolved; no, an entirely new version cannot be made outside of Meta. Yes, someone else could create such a wholly new open-source model from scratch, and different open-source groups have different incentives. The most likely incentive is monopolistic, to my mind.

diggan a day ago

I think you've kind of answered a different question. Yes, more LLM models could be created. But specifically Llama? Since it's an open source model, the assumption is that we could (given access to the same compute of course) train one from scratch ourselves, just like we can build our own binaries of open source software.

But this obviously isn't true for Llama, hence the uncertainty if Llama even is open source in the first place. If we cannot create something ourselves (again, given access to compute), how could it possibly be considered open source by anyone?

  • caseyy a day ago

    I understand I was supposed to say “no” and question the open-source label. We’ve heard many arguments that if something can’t be reproduced from scratch, it’s not true open-source.

    To me, they sound a bit like “no true Scotsman”. Llama is open source, compared to commercial models with closed weights. Even if it could be more open source.

    That’s why I looked at it in a broader sense — what could happen in an open-source world to improve or replace Llama. Much could happen, thanks to Llama’s open nature, actually.

    • diggan a day ago

      > Llama is open source, compared to commercial models with closed weights

      Yeah, just like a turd is a piece of gourmet food if there is no other good food around.

      Sorry, but that's a really bad argument, "open source" is not a relative metric you use to compare different things, it's a label that is applied to something depend on what license that thing has. No matter what licenses others use, the license you use is still the license use.

      Especially when there are actually open source models out there, so it isn't possible. Maybe Meta feels like it's impossible because of X, Y and Z, but that doesn't make it true just because they don't feel like they could earn enough money on it, or whatever their reasoning is.

      • caseyy a day ago

        > Yeah, just like a turd is a piece of gourmet food if there is no other good food around.

        I didn't mean it's on a continuum, as you assumed. Apologies for phrasing it unclearly. I meant that the weights are public. They are open; there is no debate to be had about it. Generally and broadly, that is already considered open-source.

        And we all understand what "open-source" means in the context of Llama - it doesn't mean one of the idealized notions of open source, it means open weights.

  • ImprobableTruth a day ago

    I think the fact that all (good) LLM datasets are full with licensed/pirated material means we'll never really see a decent open source model under the strict definition. Open weight + open source code is really the best we're going to get, so I'm fine with it coopting the term open source even if it doesn't fully apply.

    • diggan a day ago

      > we'll never really see a decent open source model under the strict definition

      But there are already a bunch of models like that, were everything (architecture, training data, training scripts, etc) is open, public and transparent. Since you weren't aware those existed since before, but you now know that, are you willing to change your perspective on it?

      > so I'm fine with it coopting the term open source even if it doesn't fully apply

      It really sucks that the community seems OK with this. I probably wouldn't have been a developer without FOSS, and I don't understand how it can seem OK to rob other people of this opportunity to learn from FOSS projects.

      • pabs3 a day ago

        Not all of the community is OK with this, lots of folks are strongly against OSI's bullshit OSAID for example. Really it should have been more like the Debian Deep Learning Team's Machine Learning Policy, just like last time when the OSI used the Debian Free Software Guidelines (DFSG) to create the Open Source Definition (OSD).

        https://salsa.debian.org/deeplearning-team/ml-policy