I'm aware of the guidelines, however I didn't link to the whole talk - I linked to the specific time segment where he was talking about the subject mentioned.
Yann Lecun is actively corrupting the term "open source" in relation to AI. If he really is pro open source, he should loudly acknowledge that all the recent models Meta has released do not qualify and push internally for them to go back to Apache 2.0 licensing and stop with their silly "you can only use it for what we allow" proprietary licenses.
For clarity, llama etc are not open source because of the conditions on the license under which the weights are released. It may not be possible to copyright weights or to apply software licenses to them, it hasn't been decided by the courts. But everyone acts like it's possible, as evidenced by the licenses they use. So under that assumption, Meta is not using an open source license with the weights.
Edit: for greater clarity, the open source definition has the following clause
The license must not restrict anyone from making use of the program in a specific field of endeavor. For example, it may not restrict the program from being used in a business, or from being used for genetic research.
Similar things appear in the debian free software guidelines, fsf free software definition, etc.
The Meta Llama license lists restrictions on field of endeavour. Ergo, not open source. There are other clauses (the one about the number of users) which would also disqualify it, but the use restrictions are the most egregious.
The term has been polluted for a long time before AI became well known. Just like "DevOps", "Platform Engineering", "Agile" etc... and like those terms it's not a specific thing (e.g. a specific license like "Apache 2.0", "MIT" etc...) but more a general term without qualification of scope. Do I like it? No. But not liking how people are going to continue using a term shouldn't detract from the importance of the intent.
At the end of the day - having LLM tech that has an open source _code_ base is what many people are talking about, the "data" that forms the model isn't always open source and while it would be nice if that was the case - is somewhat similar to open source software vs the data you run in it.
LLAMA2 is very much open source. You can use it for commercial purposes. The license only restricts companies that have more than 700M monthly active users from using it.
Releasing the trained model does not make it open source.
Open source means something else.
Calling released trained models open source is akin to releasing the binaries of a software for free, without releasing the source code, and calling it open source.
It's more like releasing the code as open source (under whatever license is specified) but to make the code useful you need data and that data might be free but not open sourced under any license.
If it was trained on the entire internet for example - you can't just bundled up every piece of data on the internet into a zip file to include with the code to interpret it.
>to make the code useful you need data and that data might be free but not open sourced under any license
To use something like Meta’s open sourced LLAMA2 model you don’t need the data. The model is self contained. It’s a compressed lossy form of all the data it was trained on.
The weights allow you to continue its training with new data of your choosing.
> Releasing the trained model does not make it open source.
This is a straw man argument as Meta has released the model weights and not just the trained model. When it comes to models the weights are the source.
It may be a difference in terminology, but it's definitely no straw man.
There's a philosophy of open source based on recipients being able to make any small and large modifications to the source material as they see fit.
Just being able to build new projects with something or read what it does not qualify in that context -- it requires the freedom to rework the thing itself to suit your own needs and ends.
In the case of these models, that philosophy would have the source training data and the training algorithms themselves available, editable, verifiable, and repeatable. Ignoring the financial challenges, a truly open source model would let you remove, add, edit, or reclassify a sample in the original training data and then build new weights of their won.
The released weights are not that kind of open source, and that's what these arguments are generally making the a case for.
There are still plenty of grounds from which to critique that argument if you wanted to, but there's no straw man involved in it.
I don't want to support the guy you replied to, he doesn't understand what open source means, but I belive and have argued it is possible to exercise the freedoms that are part of open source software without access to the training data. See https://www.marble.onl/posts/considerations_for_copyrighting...
>There's a philosophy of open source based on recipients being able to make any small and large modifications to the source material as they see fit.
The reason I see it as a strawman is because the weights are the source of the AI model and they do allow you to modify the model however you please.
The data is not owned by Meta, they can’t release it and it’s not required to use or modify the model. This is like wanting a team to release the research that informed product development with their source.
The data and training code are not the same as the AI model itself. This is not the same as code that produces a binary, this is more like code that produces another form of code.
If you try to replicate the model you won’t be able to because the process is not deterministic. There’s no reason to replicate the model, you would fine tune it instead.
You demonstrably don't understand what open source means, which is basically my point, that people like Lecun have been trying to corrupt the meaning to be equivalent to "source available".
I agree with the point you made elsewhere that releasing the data is irrelevant.
> You demonstrably don't understand what open source means, which is basically my point, that people like Lecun have been trying to corrupt the meaning to be equivalent to "source available".
You demonstrably don’t understand what it means to open source an AI model. The code is not relevant at all, the same code will not produce the same model weights, it is not deterministic.
Also you will need millions of dollars to retrain, retraining is not the goal. The model weights are the source that was opened.
> The model weights are the source that was opened.
Model weights are not "the preferred form in which a programmer would modify the program" but more like "the output of a preprocessor or translator" [0], and thus are not "source code", but viewed as code, are more like object code. The training data, training configuration, and program source for the training routines are "source code", without which the model as a whole is not truly open source, just as much as the program source for the inference routines are.
[0] for the relevance of the quoted material to whether or not your description meets the Open Source Definition, see the definition itself: https://opensource.org/osd/
I think the model weights are the preferred form in which a programmer would modify the program. We modify models by fine-tuning them, which requires the weights, but not the training data/recipe.
I think you're misunderstanding their point and mixing it up with other commenters. They are saying everything is available, but there are license restrictions on how you can use it. They're not asking for more code or data.
It's source available, but it's not "open source" in the historical usage by programmers. Microsoft could not take their model weights and use it. It's not open source.
It's only open source if you can generate the binary weights from the training data and the training algorithm and training (hyper) parameters, from scratch. Releasing only the binary weights, does not make it open source, just like releasing only binary executables has nothing to do with open source.
> It's only open source if you can generate the binary weights from the training data and the training algorithm and training (hyper) parameters, from scratch. Releasing only the binary weights, does not make it open source, just like releasing only binary executables has nothing to do with open source.
The weights aren’t binary. It’s not a compile form, it’s the actual source of the AI model.
The same code used to train the model / generate the weights will not produce the same model weights if run again, it is not deterministic.
Also you will need millions of dollars to retrain, retraining is not the goal. The model weights are the source that was opened. The weights can be used to modify the model.
The weights are typically distributed as a binary blob containing IEEE floating point values, or fixed point values; and these numbers are typically not meant for direct human consumption.
> The same code used to train the model / generate the weights will not produce the same model weights if run again, it is not deterministic.
This is not true, except in contrived situations like when you use PRNGs and you deliberately throw away their seeds.
> Also you will need millions of dollars to retrain
This comes down to arguing over semantics. Many will argue that "open source" means using a license listed at https://opensource.org/licenses/ . Even CC0 isn't considered open source by this standard.
Even I feel that "700M monthly active user" gives not-so-open vibes, but frankly many of these "open source" licenses have oddly restrictive requirements as well.
The inference and training software used to run the model are open source, the concrete model -- that is, the thing for which the weights are the object code -- is not.
The concrete model is free-to-use closed source, which is better than an undisclosed blob hiding behind a SaaS service, but still not open source.
It's also good that the inference and training code are open source, even though the training data and configuration is not.
Cutting edge AI is a world-changing technology that is at immediate risk of being monopolized by 1-2 organizations pushing for global regulation against competitors and individuals who would develop their own. We desperately need "Open" in this context to be a broad, directional movement resisting that lockdown, not a rehash of licensing dogma drama. If you care about agency within the software ecosystem that surrounds you, at least differentiate the main points of Yann's arguments from holy wars about Llama's license.
This isn't MIT vs GPL. It's literally one of those few corporations that's in a position to monopolize AI calling a license that says "you can use this at our pleasure" open source. Any open source movement definitively cannot include such licenses. If this was a split over a smaller thing I'd get your point.
Which clause are you calling "you can use this at our pleasure"? The license is short and there is no such thing. The most restrictive item is whether you had 750M active users in the month before it was released. [1]
I'm curious what you are drawing issue with, but it's also under the bridge. Any sane reading of this license and the release shows that you have WAY more rights compared to GPT-4 or other fully closed releases. The availability of these weights and code are propping up an open source community around LLMs, which lends oxygen to other even-more-open endeavors.
> Cutting edge AI is a world-changing technology that is at immediate risk of being monopolized by 1-2 organizations pushing for global regulation against competitors and individuals who would develop their own.
Open-source tends to foster monopoly and relies on free labor (see Google, Meta, Linux, etc.)
Ah yes, the dreaded Linux monopoly. Well, the cure is simple: just don't contribute to or use open source. Or were you hoping to keep others from sharing things for free?
People can and are building things using Llama2 that are not under Meta's control. Every comment is engaging in the definition of open source, but you obviously know what he is referring to. The pipeline to building a functional AI isn't going to perfectly map to just source code.
Library corpus sets must be available for purchase for research or machine learning. Is the only way to buy each individual books? People could go to the library to learn. ML should have the same opportunity.
The llama models are more open source than gpt-3.5-turbo. They aren't "free software" in the full definition, but they are available to be fine-tuned, modified, run offline, and generally tinkered with.
For AI, "open source" is less about the license and what it allows you to do legally, and more about what the technology allows you to build practically.
If/when Meta gets litigious, this might change in a hurry. Right now, there are bigger risks to AI startups (not Google) than worrying about how they'll choose to enforce the llama license.
No you can't license edge weights any more than you can license concentrations of ethanol in gasoline. Modified DNA can be patented, because it falls into one of the categories "process, machine, composition of matter, manufacture" in that it is a new composition of matter. Edge weights are not that.
Edit: Actually, if the training set is IP and the method to train is IP then maybe the weights are licensable, given that it's trade-secret territory. However, ensuring that the edge-weights are not widely disseminated once shared with a licensee is not straightforward and the infeasibility of such makes it an uphill climb. Not like that stops the lawyers from clawing up the tree, tho.
As if it wasn't already confusing enough to most people, LeCun misuses "open source" likely at the behest or benefit of his Meta handlers. Because of this, I cannot take anything he says very seriously.
In this field where the metaphysical benefits (and risks) are so great, you simply cannot shade the truth. It's hard enough filtering everyone else's agenda on matters much less significant.
Leaving aside all the incredibly vague* and broad restrictions supposedly placed upon its use by the license, it's so strange to me that anybody is referring to a 10 GiB binary blob as "open source".
Am I the only one?
It's like finding out that everyone else has an entirely different concept in their head than the one I thought we agreed on a quarter-century ago.
There is a lot of interesting discussion already, but I think there is also some (perhaps unnecessary) disagreements regarding licensing arising from what I believe are the very different backgrounds in the community. Thus, I will attempt to summarise my own perspective as someone that has been active in the artificial intelligence community for nearly 15 years at this point.
Firstly, I will state that I am in full agreement with LeCun on what he states in the video about the need for AI systems such as LLMs to be open [1]. I am not sure if I entirely agree with his very high optimism regarding AI safety, but I have always been far more concerned about current risks and I suspect that in practice we are very much aligned.
Now, let me move to the issue of licensing that I have already talked about at great length in comments on other, older submissions. So let me do something new here and try to put it into a historical context.
Computer science and machine learning in particular has a tradition of openness that goes beyond that of many other fields. What I mean by this is that over the last twenty or so years, there has been a gradual move towards open sharing of research, research outcomes, etc. All under the umbrella of open science and FLOSS. One of the earliest examples I am aware of (there may be others) is how the Journal of Machine Learning Research was established as a direct reaction to the resistance against open access in another leading journal at the time [2]. Similarly, in natural language processing, which my primary field of research, we have an open anthology where the great majority (if not nearly all) of the most important research going all the way back to the 70s is available for everyone to read and use [3]. In AI, we really have pushed back against the big publishers and this is a beautiful thing indeed (one which LeCun has contributed towards at that). So, given the context of this, I believe you can understand the dismay which the field feels when Google Deep Mind publishes papers such as the one on Alpha Go in journals without open access [4]. Yes, getting hold of the papers is not challenging for us, but it clearly goes against the efforts of the great majority of the filed. Thus the dismay.
Moving now to FLOSS. When I entered the field nearly 15 years ago, it was somewhat of a rarity to find source code attached to papers. Likewise, training data was also not always made available or came under awful (and expensive) licensing agreements such as those imposed by the Linguistic Data Consortium. However, we as researchers released more and more open data and code, and by 2015 or so it was starting to become the norm that both code and data was attached to papers. Back in those days, training a system given the training data took maybe a day or so, thus model weights were less important. In addition, the models were less complex and thus replicating the training procedure was less problematic. Still, we saw more and more papers being published with their weights. In terms of licenses, there was a healthy mix of MIT-like, Apache 2.0, and GPL. The notable exception in the field would be Google Deep Mind, which despite their prominence at best would publish a library or two. This in contrast to say Google "Proper", which gave us say BERT [5]. It is in this context which I and many...
51 comments
[ 2.9 ms ] story [ 102 ms ] threadFor clarity, llama etc are not open source because of the conditions on the license under which the weights are released. It may not be possible to copyright weights or to apply software licenses to them, it hasn't been decided by the courts. But everyone acts like it's possible, as evidenced by the licenses they use. So under that assumption, Meta is not using an open source license with the weights.
Edit: for greater clarity, the open source definition has the following clause
Similar things appear in the debian free software guidelines, fsf free software definition, etc.The Meta Llama license lists restrictions on field of endeavour. Ergo, not open source. There are other clauses (the one about the number of users) which would also disqualify it, but the use restrictions are the most egregious.
At the end of the day - having LLM tech that has an open source _code_ base is what many people are talking about, the "data" that forms the model isn't always open source and while it would be nice if that was the case - is somewhat similar to open source software vs the data you run in it.
Open source means something else.
Calling released trained models open source is akin to releasing the binaries of a software for free, without releasing the source code, and calling it open source.
If it was trained on the entire internet for example - you can't just bundled up every piece of data on the internet into a zip file to include with the code to interpret it.
To use something like Meta’s open sourced LLAMA2 model you don’t need the data. The model is self contained. It’s a compressed lossy form of all the data it was trained on.
The weights allow you to continue its training with new data of your choosing.
This is a straw man argument as Meta has released the model weights and not just the trained model. When it comes to models the weights are the source.
There's a philosophy of open source based on recipients being able to make any small and large modifications to the source material as they see fit.
Just being able to build new projects with something or read what it does not qualify in that context -- it requires the freedom to rework the thing itself to suit your own needs and ends.
In the case of these models, that philosophy would have the source training data and the training algorithms themselves available, editable, verifiable, and repeatable. Ignoring the financial challenges, a truly open source model would let you remove, add, edit, or reclassify a sample in the original training data and then build new weights of their won.
The released weights are not that kind of open source, and that's what these arguments are generally making the a case for.
There are still plenty of grounds from which to critique that argument if you wanted to, but there's no straw man involved in it.
The reason I see it as a strawman is because the weights are the source of the AI model and they do allow you to modify the model however you please.
The data is not owned by Meta, they can’t release it and it’s not required to use or modify the model. This is like wanting a team to release the research that informed product development with their source.
The data and training code are not the same as the AI model itself. This is not the same as code that produces a binary, this is more like code that produces another form of code.
If you try to replicate the model you won’t be able to because the process is not deterministic. There’s no reason to replicate the model, you would fine tune it instead.
I agree with the point you made elsewhere that releasing the data is irrelevant.
You demonstrably don’t understand what it means to open source an AI model. The code is not relevant at all, the same code will not produce the same model weights, it is not deterministic.
Also you will need millions of dollars to retrain, retraining is not the goal. The model weights are the source that was opened.
Model weights are not "the preferred form in which a programmer would modify the program" but more like "the output of a preprocessor or translator" [0], and thus are not "source code", but viewed as code, are more like object code. The training data, training configuration, and program source for the training routines are "source code", without which the model as a whole is not truly open source, just as much as the program source for the inference routines are.
[0] for the relevance of the quoted material to whether or not your description meets the Open Source Definition, see the definition itself: https://opensource.org/osd/
It's source available, but it's not "open source" in the historical usage by programmers. Microsoft could not take their model weights and use it. It's not open source.
The weights aren’t binary. It’s not a compile form, it’s the actual source of the AI model.
The same code used to train the model / generate the weights will not produce the same model weights if run again, it is not deterministic.
Also you will need millions of dollars to retrain, retraining is not the goal. The model weights are the source that was opened. The weights can be used to modify the model.
The weights are typically distributed as a binary blob containing IEEE floating point values, or fixed point values; and these numbers are typically not meant for direct human consumption.
> The same code used to train the model / generate the weights will not produce the same model weights if run again, it is not deterministic.
This is not true, except in contrived situations like when you use PRNGs and you deliberately throw away their seeds.
> Also you will need millions of dollars to retrain
This is besides the point.
Even I feel that "700M monthly active user" gives not-so-open vibes, but frankly many of these "open source" licenses have oddly restrictive requirements as well.
The inference and training software used to run the model are open source, the concrete model -- that is, the thing for which the weights are the object code -- is not.
The concrete model is free-to-use closed source, which is better than an undisclosed blob hiding behind a SaaS service, but still not open source.
It's also good that the inference and training code are open source, even though the training data and configuration is not.
I'm curious what you are drawing issue with, but it's also under the bridge. Any sane reading of this license and the release shows that you have WAY more rights compared to GPT-4 or other fully closed releases. The availability of these weights and code are propping up an open source community around LLMs, which lends oxygen to other even-more-open endeavors.
[1] https://ai.meta.com/llama/license/
Open-source tends to foster monopoly and relies on free labor (see Google, Meta, Linux, etc.)
There is a weird definition of open source that creeps in here.
Llama2 is effectively, proprietary compiled software. GPT-4 is proprietary compiled and hosted as SaaS.
We have yet to see a true open source LLM...
AI/ML is reinventing so many terms it's a bit annoying.
It appears all my favorite books have been scanned by Open AI. I am assuming that they didn’t purchase them.
For AI, "open source" is less about the license and what it allows you to do legally, and more about what the technology allows you to build practically.
If/when Meta gets litigious, this might change in a hurry. Right now, there are bigger risks to AI startups (not Google) than worrying about how they'll choose to enforce the llama license.
Edit: Actually, if the training set is IP and the method to train is IP then maybe the weights are licensable, given that it's trade-secret territory. However, ensuring that the edge-weights are not widely disseminated once shared with a licensee is not straightforward and the infeasibility of such makes it an uphill climb. Not like that stops the lawyers from clawing up the tree, tho.
In this field where the metaphysical benefits (and risks) are so great, you simply cannot shade the truth. It's hard enough filtering everyone else's agenda on matters much less significant.
Am I the only one?
It's like finding out that everyone else has an entirely different concept in their head than the one I thought we agreed on a quarter-century ago.
*E.g., Ctrl+F "obscenity".
Firstly, I will state that I am in full agreement with LeCun on what he states in the video about the need for AI systems such as LLMs to be open [1]. I am not sure if I entirely agree with his very high optimism regarding AI safety, but I have always been far more concerned about current risks and I suspect that in practice we are very much aligned.
[1]: https://yewtu.be/watch?v=EGDG3hgPNp8&t=6525s
Now, let me move to the issue of licensing that I have already talked about at great length in comments on other, older submissions. So let me do something new here and try to put it into a historical context.
Computer science and machine learning in particular has a tradition of openness that goes beyond that of many other fields. What I mean by this is that over the last twenty or so years, there has been a gradual move towards open sharing of research, research outcomes, etc. All under the umbrella of open science and FLOSS. One of the earliest examples I am aware of (there may be others) is how the Journal of Machine Learning Research was established as a direct reaction to the resistance against open access in another leading journal at the time [2]. Similarly, in natural language processing, which my primary field of research, we have an open anthology where the great majority (if not nearly all) of the most important research going all the way back to the 70s is available for everyone to read and use [3]. In AI, we really have pushed back against the big publishers and this is a beautiful thing indeed (one which LeCun has contributed towards at that). So, given the context of this, I believe you can understand the dismay which the field feels when Google Deep Mind publishes papers such as the one on Alpha Go in journals without open access [4]. Yes, getting hold of the papers is not challenging for us, but it clearly goes against the efforts of the great majority of the filed. Thus the dismay.
[2]: https://en.wikipedia.org/wiki/Journal_of_Machine_Learning_Re...
[3]: https://aclanthology.org
[4]: https://www.nature.com/articles/nature24270
Moving now to FLOSS. When I entered the field nearly 15 years ago, it was somewhat of a rarity to find source code attached to papers. Likewise, training data was also not always made available or came under awful (and expensive) licensing agreements such as those imposed by the Linguistic Data Consortium. However, we as researchers released more and more open data and code, and by 2015 or so it was starting to become the norm that both code and data was attached to papers. Back in those days, training a system given the training data took maybe a day or so, thus model weights were less important. In addition, the models were less complex and thus replicating the training procedure was less problematic. Still, we saw more and more papers being published with their weights. In terms of licenses, there was a healthy mix of MIT-like, Apache 2.0, and GPL. The notable exception in the field would be Google Deep Mind, which despite their prominence at best would publish a library or two. This in contrast to say Google "Proper", which gave us say BERT [5]. It is in this context which I and many...