381 karma · joined January 30, 2017
iOS settings => Sounds and Haptics => Drag the Ringtone and Alert Volume slider up higher and turn off change with buttons
The alarm clock volume on iOS is separate from your normal volume. So if it accidentally got lowered then your alarm clock will never make any sound. It can get accidentally lowered if you have change with buttons on so I turn that off. Turning off change with buttons just fixes your alarm and ringer volume, it won’t prevent you from changing your volume normally when listening to music
Disclaimer: Not an ML expert
Stable diffusion has something called an encoder and decoder. What the encoder does is it takes an image, finds it's fundamental characteristics, and then converts it into a data point (for the sake of simplicity we will use a vector even though it doesn't have to be). Let's say the vector <0.2,0.5,0.6> represents a black dog. If you took a similar vector, you would get another picture of a dog (say a white dog). These vectors are contained in what's called a latent space which is just a collection of items where similar concepts are close together.
Stable Diffusion uses this latent space because it's more computationally efficient. So what it does is it starts with a noisy image which is converted into latent space, then it slowly gets rid of noise. It does this entire process on the latent space representation as opposed to the actual image. This means it's more computationally efficient because it doesn't have to store an entire pixel image in memory. Once it finishes getting rid of the noise, it uses the decoder to convert the image back into a pixel image. What you'll notice is that throughout this entire process it's not just retrieving a compressed image from it's training set and then using it. Instead, it's generating the image through de-noising. This de-noising process is guided by it's understanding of different concepts that can be represented in the latent space.
I think where this lawsuit goes wrong is it implies that the latent space is literally storing a copy of every image in the dataset. As far as I am aware, this is not true. Even though the latent space representations of images are dramatically smaller, it's not small enough to fit the entire dataset in a 5gb file. The only thing Stable Diffusion is storing is the algorithm itself for converting to and from latent space and that's just for computational efficiency as mentioned above. I've heard that Stable Diffusion might store some key concepts from the latent space, but I don't know if that's true or not. Either way, it seems unlikely that the entire dataset is being stored in Stable Diffusion. To me, it seems that saying Stable Diffusion is storing the images themselves is like saying GZIP's algorithm is storing the compressed version of every file in existence.
Disclaimer: Not an ML expert and this is just based on my own understanding of how it works. So I could be wrong
This seems like it’s not an accurate description of what diffusion is doing. A diffusion model is not the same as compression. They’re implying that Stable Diffusion is taking the entire dataset and making it smaller then storing it. Instead, it’s just learning patterns about the art and replicating those patterns.
The “compression” they’re referring to is the latent space representation which is how Stable Diffusion avoids having to manipulate large images during computation. I mean you could call it a form of compression, but the actual training images aren’t stored using that latent space in the final model afaik. So it's not compressing every single image and storing it in the model.
This page says there were 5 billion images in the stable diffusion training dataset (albeit that may not be true as I see online it’s closer to the 2 billion mark). A Stable Diffusion model is about 5 gb. 5 gb / 5 billion is 1 byte per image. That’s impossible to fit an image in 1 byte. Obviously the claim about it storing compressed copies of the training data is not true. The size of the file comes from the weights in it, not because it’s storing “compressed copies”. In general, it seems this lawsuit is misrepresenting how Stable Diffusion works on a technical level.
Even if it is the case, it is still an issue because it can be used for abuse. If you wanted to pin abuse on someone you just need to add them to the project doing something illegal. Google tends to be trigger happy about banning so it could affect someone. I bet there are avenues for companies to deal with a situation like this, but it still affects individuals which is a big problem. Especially if your personal account gets banned.
I'd also argue that it's a spam issue for companies (albeit less than it is for individuals). You may not see it by default for companies, but it will still appear in some lists throughout the UI probably.
I think honestly both AWS / GCP should ask for permission to be invited when you're adding someone outside your company.
Imagine you're a malicious actor and want to spam some company's Google Cloud Dashboard. You could just create a 100 different projects and add the company's users to those projects automatically, even though you're not part of that company. Those projects can never be removed or deleted from the dashboard. So every user at the company you're trying to spam can never get rid of it (without Google support intervening maybe).
Or let's say you're a malicious actor trying to get someone in trouble. You could create a Google account and hide all traces of who created it. Then you could just create a project doing something illegal (e.g. hosting illegal content) and then add someone to the project. When the authorities find out about the project, the person you're targeting will get investigated for hosting illegal content.
Edit: Reported it
I tried to look for their past tweets on the internet archive but didn’t find anything. So I can’t say for sure if they were verified beforehand or not.
Also, when it says tailor-made does it mean that there is a human prompt engineer tuning the images for you? Or does it just mean that the AI generated images are custom, but there's no human prompt "tailoring" done for you?
https://us.loopearplugs.com/products/quiet
https://www.flareaudio.com/products/sleeep-pro
https://www.happyearsearplugs.com/ (which someone already mentioned on here)