You’re humanizing the software too much. Comparing software to human behavior is just plain wrong. GPT can’t even reason properly yet. I can’t see this as anything other than a more advanced collage process.
Open used intellectual property without consent of the owners. Major fucked.
If ‘anybody’ does anything similar to tracing, copy&pasting or even sampling a fraction of another person’s imagery or written work, that anybody is violating copyright.
If ‘anybody’ does anything similar to tracing, copy&pasting or even sampling a fraction of another person’s imagery or written work, that anybody is violating copyright.
Ok, but tracing is literally a part of the human learning process. If you trace a work and sell it as your own that’s bad. If you trace a work to learn about the style and let that influence your future works that is what every artist already does.
The artistic process isn’t copyrighted, only the final result. The exact same standards can apply to AI generated work as already do to anything human generated.
i don’t know the specifics of the lawsuit but i imagine this would parallel piracy.
in a way you could say that Open has pirated software directly from multiple intellectual properties. Open has distributed software which emulates skills and knowledge. remember this is a tool, not an individual.
It’s not exactly the same thing, but here’s an article by Kit Walsh, who’s a senior staff attorney at the EFF explains how image generators work within the law. The two aren’t exactly the same, but you can see how the same ideas would apply. The EFF is a digital rights group who most recently won a historic case: border guards now need a warrant to search your phone.
Here are some excerpts:
First, copyright law doesn’t prevent you from making factual observations about a work or copying the facts embodied in a work (this is called the “idea/expression distinction”). Rather, copyright forbids you from copying the work’s creative expression in a way that could substitute for the original, and from making “derivative works” when those works copy too much creative expression from the original.
Second, even if a person makes a copy or a derivative work, the use is not infringing if it is a “fair use.” Whether a use is fair depends on a number of factors, including the purpose of the use, the nature of the original work, how much is used, and potential harm to the market for the original work.
And:
…When an act potentially implicates copyright but is a necessary step in enabling noninfringing uses, it frequently qualifies as a fair use itself. After all, the right to make a noninfringing use of a work is only meaningful if you are also permitted to perform the steps that lead up to that use. Thus, as both an intermediate use and an analytical use, scraping is not likely to violate copyright law.
it does trouble me to think that the creators of stable diffusion could be financially punished. Did they at least try to compensate the artists in anyway?
It “feels” as though it parallels consultation. These creatives are literally paid for their creations. If a software constructs a neural network to emulate intellectual property, does that count as consultation? Could/Should it apply to the software developers or individuals using the software?
From the technical side, I don’t understand how all the red flags aren’t already there. the source material was taken, and now any individual could acquire that exact material or anything “in the spirit of” that material through a single service. Is this a new way to pirate?
stable diffusion is a great opportunity for small businesses. especially in an increasingly anti-small business america (maybe that’s just california?) I’d hate for it become inaccessible to creators that would wield it properly.
as long as creatives retain the ability to sue the bad actors, i’m glad. I personally don’t need Open or whomever is directly responsible for stable diffusion and its training data to be punished.
In the US, fair use lets you use copyrighted material without permission for criticism, research, artistic expression like literature, art, music, satire, and parody. It balances the interests of copyright holders with the public’s right to access and use information. There are rights people can maintain over their work, and there are rights they do not maintain. We are allowed to analyze people’s publically published works, and that’s always been to the benefit of artistic expression. It would be awful for everyone if IP holders could take down any criticism, reverse engineering, or indexes they don’t like. That would be the dream of every corporation, bully, troll, or wannabe autocrat.
The consultation angle is interesting, but I’m not sure applies here. Consultation usually involves a direct and intentional exchange of information and expertise, whereas this is an original analysis of data that doesn’t emulate any specific intellectual property.
I also don’t think this is a new way to pirate, as long as you don’t reproduce the source material. If you wanted to do that, you could just right-click and “save as”. What this does is lower the bar for entry to let people more easily exercise their rights. Like print media vs. internet publication and TV/Radio vs. online content, there will be winners and losers, but if done right, I think this will all be in service of a more decentralized and open media landscape.
Well, given how we’re the ones that developed the models, they are deterministic as we know and can save and reproduce the random weights they are given during training, and we can use a debugger to step through every single step the models makes in learning and “thinking”, yes, we understand them.
We know the input, we can set the model to save the weight in checkpoints during training and can view them any time, and we can see weights of the finished model, and we can see the code.
If what you said about LLMs being completely black box were true, we wouldn’t be able to reproduce models, and each model would be unique.
But we can control every step of the training process, and we can reproduce not just the finished model, but the model at every single step during training.
We created the math, we created the training sets, we created the code and we can see and modify the weights and any other property of the model.
Look, I understand why you think this. I thought this too when I was first beginning to learn machine learning and data science. But I’ve now been working with machine learning models including neural networks for nearly a decade, and the truth is that is nearly impossible to track the path of an input to a given output in machine learning models other than regression-based models and decision tree-based models.
There is an entire field of data science devoted to explaining how these models arrive at their conclusions. It’s called “explainable AI” or “xAI”, and I have a few papers that I’ve published in exploring the utility of them. The basic explanation for how they work is that we run hundreds of thousands of different models and then do statistical analysis to estimate why the models arrived at their conclusion. It isn’t an exact science, however.
Again, we have the input, we have the math and code that make it work, we have the weights, we have everything.
Would it take a lot of time to backtrack and check why we got a given output to an input? Yes, maybe an inordinate amount of time. But it can be done. It’s only black box because nobody has the time (likely years to decades) to wade through the layers of a finished model to check every node and weight.
The whole thing at its core is mathematics. It’s a series of steps, that can be listed and reviewed each step of the way if we wanted. It’s just that if would take too much time.
If what you said were true, we couldn’t reproduce models. And since we can…
It isn’t an exact science, however.
So if math and computer science isn’t an exact science, what is?
sampling a fraction of another person’s imagery or written work.
So citing is a copyright violation? A scientific discussion on a specific text is a copyright violation? This makes no sense. It would mean your work couldn’t build on anything else, and that’s plain stupid.
Also to your first point about reasoning and advanced collage process: you are right and wrong. Yes an LLM doesn’t have the ability to use all the information a human has or be as precise, therefore it can’t reason the same way a human can. BUT, and that is a huge caveat, the inherit goal of AI and in its simplest form neural networks was to replicate human thinking. If you look at the brain and then at AIs, you will see how close the process is. It’s usually giving the AI an input, the AI tries to give the desired output, them the AI gets told what it should have looked like, and then it backpropagates to reinforce it’s process. This already pretty advanced and human-like (even look at how the brain is made up and then how AI models are made up, it’s basically the same concept).
Now you would be right to say “well in it’s simplest form LLMs like GPT are just predicting which character or word comes next” and you would be partially right. But in that process it incorporates all of the “knowledge” it got from it’s training sessions and a few valuable tricks to improve. The truth is, differences between a human brain and an AI are marginal, and it mostly boils down to efficiency and training time.
And to say that LLMs are just “an advanced collage process” is like saying “a car is just an advanced horse”. You’re not technically wrong but the description is really misleading if you look into the details.
And for details sake, this is what the paper for Llama2 looks like; the latest big LLM from Facebook that is said to be the current standard for LLM development:
You’re humanizing the software too much. Comparing software to human behavior is just plain wrong. GPT can’t even reason properly yet. I can’t see this as anything other than a more advanced collage process.
Open used intellectual property without consent of the owners. Major fucked.
If ‘anybody’ does anything similar to tracing, copy&pasting or even sampling a fraction of another person’s imagery or written work, that anybody is violating copyright.
Ok, but tracing is literally a part of the human learning process. If you trace a work and sell it as your own that’s bad. If you trace a work to learn about the style and let that influence your future works that is what every artist already does.
The artistic process isn’t copyrighted, only the final result. The exact same standards can apply to AI generated work as already do to anything human generated.
i don’t know the specifics of the lawsuit but i imagine this would parallel piracy.
in a way you could say that Open has pirated software directly from multiple intellectual properties. Open has distributed software which emulates skills and knowledge. remember this is a tool, not an individual.
It’s not exactly the same thing, but here’s an article by Kit Walsh, who’s a senior staff attorney at the EFF explains how image generators work within the law. The two aren’t exactly the same, but you can see how the same ideas would apply. The EFF is a digital rights group who most recently won a historic case: border guards now need a warrant to search your phone.
Here are some excerpts:
And:
I’d like to hear your thoughts.
thanks for the sauce. Its very enlightening.
it does trouble me to think that the creators of stable diffusion could be financially punished. Did they at least try to compensate the artists in anyway?
It “feels” as though it parallels consultation. These creatives are literally paid for their creations. If a software constructs a neural network to emulate intellectual property, does that count as consultation? Could/Should it apply to the software developers or individuals using the software?
From the technical side, I don’t understand how all the red flags aren’t already there. the source material was taken, and now any individual could acquire that exact material or anything “in the spirit of” that material through a single service. Is this a new way to pirate?
stable diffusion is a great opportunity for small businesses. especially in an increasingly anti-small business america (maybe that’s just california?) I’d hate for it become inaccessible to creators that would wield it properly.
as long as creatives retain the ability to sue the bad actors, i’m glad. I personally don’t need Open or whomever is directly responsible for stable diffusion and its training data to be punished.
In the US, fair use lets you use copyrighted material without permission for criticism, research, artistic expression like literature, art, music, satire, and parody. It balances the interests of copyright holders with the public’s right to access and use information. There are rights people can maintain over their work, and there are rights they do not maintain. We are allowed to analyze people’s publically published works, and that’s always been to the benefit of artistic expression. It would be awful for everyone if IP holders could take down any criticism, reverse engineering, or indexes they don’t like. That would be the dream of every corporation, bully, troll, or wannabe autocrat.
The consultation angle is interesting, but I’m not sure applies here. Consultation usually involves a direct and intentional exchange of information and expertise, whereas this is an original analysis of data that doesn’t emulate any specific intellectual property.
I also don’t think this is a new way to pirate, as long as you don’t reproduce the source material. If you wanted to do that, you could just right-click and “save as”. What this does is lower the bar for entry to let people more easily exercise their rights. Like print media vs. internet publication and TV/Radio vs. online content, there will be winners and losers, but if done right, I think this will all be in service of a more decentralized and open media landscape.
You’re mystifying and mythologising humans too much. The learning process is very equivalent.
amazing
Well, there still a shit ton we don’t understand about human.
We do, however, understand everything about machine learning.
LOL
We understand less about how LLMs generate a single output than we do about the human brain. You clearly have no experience developing models.
Well, given how we’re the ones that developed the models, they are deterministic as we know and can save and reproduce the random weights they are given during training, and we can use a debugger to step through every single step the models makes in learning and “thinking”, yes, we understand them.
We can not however, do that for the human brain.
You really don’t understand how these models work and you should learn about them before you make statements about them.
Machine learning models are, almost by definition, non-deterministic.
We know the input, we can set the model to save the weight in checkpoints during training and can view them any time, and we can see weights of the finished model, and we can see the code.
If what you said about LLMs being completely black box were true, we wouldn’t be able to reproduce models, and each model would be unique.
But we can control every step of the training process, and we can reproduce not just the finished model, but the model at every single step during training.
We created the math, we created the training sets, we created the code and we can see and modify the weights and any other property of the model.
What exactly do we not understand?
Look, I understand why you think this. I thought this too when I was first beginning to learn machine learning and data science. But I’ve now been working with machine learning models including neural networks for nearly a decade, and the truth is that is nearly impossible to track the path of an input to a given output in machine learning models other than regression-based models and decision tree-based models.
There is an entire field of data science devoted to explaining how these models arrive at their conclusions. It’s called “explainable AI” or “xAI”, and I have a few papers that I’ve published in exploring the utility of them. The basic explanation for how they work is that we run hundreds of thousands of different models and then do statistical analysis to estimate why the models arrived at their conclusion. It isn’t an exact science, however.
Again, we have the input, we have the math and code that make it work, we have the weights, we have everything.
Would it take a lot of time to backtrack and check why we got a given output to an input? Yes, maybe an inordinate amount of time. But it can be done. It’s only black box because nobody has the time (likely years to decades) to wade through the layers of a finished model to check every node and weight.
The whole thing at its core is mathematics. It’s a series of steps, that can be listed and reviewed each step of the way if we wanted. It’s just that if would take too much time.
If what you said were true, we couldn’t reproduce models. And since we can…
So if math and computer science isn’t an exact science, what is?
So citing is a copyright violation? A scientific discussion on a specific text is a copyright violation? This makes no sense. It would mean your work couldn’t build on anything else, and that’s plain stupid.
Also to your first point about reasoning and advanced collage process: you are right and wrong. Yes an LLM doesn’t have the ability to use all the information a human has or be as precise, therefore it can’t reason the same way a human can. BUT, and that is a huge caveat, the inherit goal of AI and in its simplest form neural networks was to replicate human thinking. If you look at the brain and then at AIs, you will see how close the process is. It’s usually giving the AI an input, the AI tries to give the desired output, them the AI gets told what it should have looked like, and then it backpropagates to reinforce it’s process. This already pretty advanced and human-like (even look at how the brain is made up and then how AI models are made up, it’s basically the same concept).
Now you would be right to say “well in it’s simplest form LLMs like GPT are just predicting which character or word comes next” and you would be partially right. But in that process it incorporates all of the “knowledge” it got from it’s training sessions and a few valuable tricks to improve. The truth is, differences between a human brain and an AI are marginal, and it mostly boils down to efficiency and training time.
And to say that LLMs are just “an advanced collage process” is like saying “a car is just an advanced horse”. You’re not technically wrong but the description is really misleading if you look into the details.
And for details sake, this is what the paper for Llama2 looks like; the latest big LLM from Facebook that is said to be the current standard for LLM development:
https://arxiv.org/pdf/2307.09288.pdf