Um, what about just out and out refusing to let these companies train on your music, art, writing, etc? Placing blocks on internet scrapers isn't that hard to do so shouldn't we be yelling at the services we use to create these blocks? Sure we might not be able to sue the slimeballs that run AI companies but we can sure as hell sue the services that we use to post our work for enabling copyright infringement.
Re: what about just out and out refusing to let these companies train on your music, art, writing, etc? - without a technical means of showing them what the work is how could they achieve this even if they wanted to? As it is, I have a hard time seeing any company or artists wanting to send all 800 companies their music files to show them what to opt out vs keep in. This is why I outlined the path in the piece. Simply put, how can a creator refuse to let a machine do something without telling them or showing them what that something is?
Fascinating insight Benji. Can you talk more about Sureel? Aren’t there a billion ways to convert music into data? How can one data-ized version help protect you in this scenario?
There are countless ways to convert music into data, whether it’s melody extraction, rhythm, lyrics, stems, pitch, or something else entirely. Different AI models use different methods to “data-ize” music, which makes the challenge of protection a tricky one. That’s exactly where Sureel comes in.
Sureel creates a unique signature for each song by incorporating diverse data points, like melody, harmonics, lyrics, and even how an AI might perceive the song. Our approach is designed to work seamlessly with all the cutting-edge AI models out there, whether it’s GANs, Transformers, diffusion models, or autoregressive models. We combine music theory, low-level audio features, forensic analysis, and advanced AI to create a comprehensive representation of each song. This signature remains reliable and robust, even when different models use techniques like pruning, optimizations, or vary in their architecture.
While there may be billions of ways to convert music into data, they all share some fundamental elements, core “ingredients” that Sureel always captures. Much like DNA, which is largely similar across all humans with only subtle changes making each individual unique, these fundamental components of music remain consistent across different AI models. Sureel’s methodology extracts both the shared elements and the unique components of each song, ensuring compatibility with any AI model, no matter the specific method used. By capturing this multidimensional representation, Sureel helps protect your music in a way that’s resilient across all these variations.
I stand corrected, you can put your own metadata on you tube videos, via tags, but, it's not obvious how to do it, unless you know how! I do now, so thank you again for the counselling lol!
Any interest coming on my podcast to discuss this (and a bunch more)? Would love to have a chat.
Let's do it!
Um, what about just out and out refusing to let these companies train on your music, art, writing, etc? Placing blocks on internet scrapers isn't that hard to do so shouldn't we be yelling at the services we use to create these blocks? Sure we might not be able to sue the slimeballs that run AI companies but we can sure as hell sue the services that we use to post our work for enabling copyright infringement.
Re: what about just out and out refusing to let these companies train on your music, art, writing, etc? - without a technical means of showing them what the work is how could they achieve this even if they wanted to? As it is, I have a hard time seeing any company or artists wanting to send all 800 companies their music files to show them what to opt out vs keep in. This is why I outlined the path in the piece. Simply put, how can a creator refuse to let a machine do something without telling them or showing them what that something is?
Thanks for this! I will register later today.
Oh wow. This makes a lot of sense.
Fascinating insight Benji. Can you talk more about Sureel? Aren’t there a billion ways to convert music into data? How can one data-ized version help protect you in this scenario?
From Tamay the CEO - more from him to follow:
There are countless ways to convert music into data, whether it’s melody extraction, rhythm, lyrics, stems, pitch, or something else entirely. Different AI models use different methods to “data-ize” music, which makes the challenge of protection a tricky one. That’s exactly where Sureel comes in.
Sureel creates a unique signature for each song by incorporating diverse data points, like melody, harmonics, lyrics, and even how an AI might perceive the song. Our approach is designed to work seamlessly with all the cutting-edge AI models out there, whether it’s GANs, Transformers, diffusion models, or autoregressive models. We combine music theory, low-level audio features, forensic analysis, and advanced AI to create a comprehensive representation of each song. This signature remains reliable and robust, even when different models use techniques like pruning, optimizations, or vary in their architecture.
While there may be billions of ways to convert music into data, they all share some fundamental elements, core “ingredients” that Sureel always captures. Much like DNA, which is largely similar across all humans with only subtle changes making each individual unique, these fundamental components of music remain consistent across different AI models. Sureel’s methodology extracts both the shared elements and the unique components of each song, ensuring compatibility with any AI model, no matter the specific method used. By capturing this multidimensional representation, Sureel helps protect your music in a way that’s resilient across all these variations.
I stand corrected, you can put your own metadata on you tube videos, via tags, but, it's not obvious how to do it, unless you know how! I do now, so thank you again for the counselling lol!