Meta needs to stop open-washing their product. It simply is not open-source. The license for their precompiled binary blob (ie model) should not be considered open-source, and the source code (ie training process / data) isn’t available.
They've painted themselves into a corner - the second people see the announcement that they've enforced the license on someone, people will switch to actual open source licensed models and Meta's reputation will take a hit.
It's ironic that China is acting as a better good faith participant in open source than Meta. I'm sure their stakeholders don't really care right now, but Meta should switch to Apache or MIT. The longer they wait the more invested people will be and the more intense the outrage when things go wrong.
The blob itself isn't, exactly: you cannot just reencode a movie and claim copyright protection over the resulting blob.
What's protected is the content of the movie, and it's protected because it derives from human creativity.
> The copyright law only protects “the fruits of intellectual labor” that “are founded in the creative powers of the mind.”
> […]
> Similarly, the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.
No, derivative work require human creativity themselves. Compiling or re-encoding still doesn't count.
See : https//www.law.cornell.edu/uscode/text/17/101
A work consisting of editorial revisions, annotations, elaborations, or other modifications which, as a whole, represent an original work of authorship, is a "derivative work".
> Which exists as a binary blob.
Nope, for copyright protection it must exist at least as one binary blob, but having multiple binary blobs (with different resolutions) doesn't make it a different copyright piece. It's the underlying creation that is protected, not a particular instance of it. Star Wars, the Empire Strikes Back is what's registered at the Copyright Office, not Star_Wars_The_Empire_Strikes_Back.720p.avi.
> Copying that binary blob requires a license to do so.
Fortunately no, otherwise your internet provider would need a license from the copyright holders to copy the blob from Netflix server to your machine.
One last time: copyright isn't about the blob, it's about the creation stored on it. The process of creating the blob doesn't grant you any copyright protection of you don't own the underlying material.
>No, derivative work require human creativity themselves.
Then it would just be a copy then. Copies need a license.
>Fortunately no, otherwise your internet provider would need a license from the copyright holders to copy the blob from Netflix server to your machine.
No, I believe this is because internet providers do not save the content which means that a copy is not considered to be made. If copies were allowed of binary blobs people could legally make pirate sites sharing copies like that.
> No, I believe this is because internet providers do not save the content which means that a copy is not considered to be made.
Nope, that's not the reason, and that's why you don't need to give a copyright license to Apple before storing your personal pictures to iCloud either, nor does Apple need a license to store copyrighted material you got a license for (like software or paid downloaded movies). Copying a blob isn't a license infringement in itself, because the blob itself was never protected by copyright.
> If copies were allowed of binary blobs people could legally make pirate sites sharing copies like that.
No, because sharing is what you'd get prosecuted for.
Part of me thinks you should really try to start learning the basis of stuff before arguing on the internet about it, but who am I to judge your life choices. I did my best to help you learn something, but if you refuse to there's nothing more I can do.
> You do which is why it's a part of the terms of service for icloud.
404
> Copyright controls both reproduction and distribution.
Reproduction in the copyright sense isn't about blob copying. RAID 1 isn't a copyright infringement either… And neither is a Windows defragmentation (which is just the OS copying files around).
> You are being unnecessarily smug and condescending.
You are needlessly obstinate on a topic you don't understand.
This is actually my first impression while I am reading the post. Mentions "open source" everywhere but dude how the earth it is open source without training data.
Almost no company is going to release training data because they don't want to waste time with lawsuits. That's why it doesn't happen. Until governments fix that issue, I don't even think the "it's not really open without training data!!!" argument is worth any time. It's more worth focusing on the various restrictions in the LLaMA license, or even better, questioning whether model weights can be licensed at all.
What is the point of considering this hypothetical? Google, Microsoft, Nvidia, Apple, and many Chinese big-tech companies also release open weights models, most with fewer restrictions.
My issue with Meta's open-washing is that it is also not open-weight, given the license restrictions. It's "weight-available", I suppose. Try OLMo instead.
