The point is that ML can generate works in bulk outpacing humans by a ridiculous rate. And that it explicitly is meant to cover expression, but ML models don’t actually express anything, they just emit statistical averages of the input.
You don’t want a company with obviously way too few human employees to have created all of their works to be able to go look for similar art from others and threaten copyright lawsuits. By forcing humans to be involved in the process of creation you strongly limit the ability of such legal trolls to hurt other creators. Such copyright trolls ALREADY exists prior to ML, but extending copyright to unsupervised ML would superpower their lawsuits. They just have to spam various websites with some samples and then pretend everybody copied them.
This precedence is just what it should be. The reference to photos is completely correct. The creation of a sequence of bits isn’t in itself protected, it’s the selection of inputs in which the creativity lies that then carries over into protection of the output. Photos can be copyrighted because a human express something in their choice of motive. A surveillance camera for example don’t automatically give its operator copyright!
And it still doesn’t prevent you from using ML in the creation of things and claiming copyright, it just requires you to be the one directing the process instead of leaving it unsupervised.
Don’t want to be that guy, but they absolutely don’t “emit the statistical averages of the inputs”. Otherwise they would create a single, most likely unicolor, image
Do I have to explain with math how my high level abstract reply applies?
Most generative ML rely on probabilities. The averages are over multidimensional complex data structures representing patterns extracted from the inputs. Like average faces when you prompt it for a face (try training it on different sets of faces and look at how the output differs, you really do see it retain averages of the patterns in the input such as average skin color and haircuts). I wasn’t talking about linear arithmetic averages.
You still misunderstand my use of “average”. I am once again not talking simple averages over simple arithmetic numbers.
Look at the outputs of models trained on majority white faces vs diverse faces. If you still don’t understand what I mean by averages then I guess this conversation is hopeless
Yes there’s noise in the process.
But that noise is applied in very specific ways, it still fundamentally tries to output what the training algorithm indicated you would expect from it given the prompt, staying in a general neighborhood preserving the syntax / structure / patterns in the input training data related to the keywords in your prompt. Ask for a face without more details and you get a face looking like the average face in the input, usually white in most models, western conventional haircuts, etc, because that’s representative of its inputs, an average over the extracted structure in the inputs. The noise just tweaks some selection of representative features and their exact properties. It is still close enough to average that I feel it is fair to call it average, because it so rarely output extremes (other than when the model just breaks down and produce nonsense).
That’s a bias, not an average. Similar to human biases. Models’ biases are derived from humans’ biases in the training data.
Humans have biases for a male doctor and female nurse, models learn that bias unless someone intervene to identify and remove the cultural (very human) bias from the training data set
You misunderstood again. The model isn’t creating the bias when it is trained on biased data. It just gives a representative output of its input. The average of many outputs will resemble the average of its inputs.
The point is that ML can generate works in bulk outpacing humans by a ridiculous rate. And that it explicitly is meant to cover expression, but ML models don’t actually express anything, they just emit statistical averages of the input.
You don’t want a company with obviously way too few human employees to have created all of their works to be able to go look for similar art from others and threaten copyright lawsuits. By forcing humans to be involved in the process of creation you strongly limit the ability of such legal trolls to hurt other creators. Such copyright trolls ALREADY exists prior to ML, but extending copyright to unsupervised ML would superpower their lawsuits. They just have to spam various websites with some samples and then pretend everybody copied them.
This precedence is just what it should be. The reference to photos is completely correct. The creation of a sequence of bits isn’t in itself protected, it’s the selection of inputs in which the creativity lies that then carries over into protection of the output. Photos can be copyrighted because a human express something in their choice of motive. A surveillance camera for example don’t automatically give its operator copyright!
And it still doesn’t prevent you from using ML in the creation of things and claiming copyright, it just requires you to be the one directing the process instead of leaving it unsupervised.
Don’t want to be that guy, but they absolutely don’t “emit the statistical averages of the inputs”. Otherwise they would create a single, most likely unicolor, image
Ok then, weighted averages.
Neither. Going from a prompt to an image is a stochastic non-linear transformation based on billions of parameters.
Your brain also performs stochastic non-linear transformations of inputs. Just in a different way.
Do I have to explain with math how my high level abstract reply applies?
Most generative ML rely on probabilities. The averages are over multidimensional complex data structures representing patterns extracted from the inputs. Like average faces when you prompt it for a face (try training it on different sets of faces and look at how the output differs, you really do see it retain averages of the patterns in the input such as average skin color and haircuts). I wasn’t talking about linear arithmetic averages.
My comment is that they simply are not averages, that’s it.
As a simpler example, it is like saying that a polynomial plus some noise is an average… It’s simply not.
The stochastic and non linear parts are the reason these models create original images, unless overfitted.
If it was a weighted average you’d have identical, smoothed, most likely non sensical images for identical prompts.
And this is not the case.
That’s all my comment.
You still misunderstand my use of “average”. I am once again not talking simple averages over simple arithmetic numbers.
Look at the outputs of models trained on majority white faces vs diverse faces. If you still don’t understand what I mean by averages then I guess this conversation is hopeless
Yes there’s noise in the process.
But that noise is applied in very specific ways, it still fundamentally tries to output what the training algorithm indicated you would expect from it given the prompt, staying in a general neighborhood preserving the syntax / structure / patterns in the input training data related to the keywords in your prompt. Ask for a face without more details and you get a face looking like the average face in the input, usually white in most models, western conventional haircuts, etc, because that’s representative of its inputs, an average over the extracted structure in the inputs. The noise just tweaks some selection of representative features and their exact properties. It is still close enough to average that I feel it is fair to call it average, because it so rarely output extremes (other than when the model just breaks down and produce nonsense).
That’s a bias, not an average. Similar to human biases. Models’ biases are derived from humans’ biases in the training data.
Humans have biases for a male doctor and female nurse, models learn that bias unless someone intervene to identify and remove the cultural (very human) bias from the training data set
You misunderstood again. The model isn’t creating the bias when it is trained on biased data. It just gives a representative output of its input. The average of many outputs will resemble the average of its inputs.