The technology of compression a diffusion model would have to achieve to realistically (not too lossily) store “the training data” would be more valuable than the entirety of the machine learning field right now.
They do not “compress” images.
I dunno. Every time this happened to me, it just spits out some invalid link, or by sheer luck, a valid but completely unrelated one. This probably happened because it reaches its context limit, only sees “poem” and then tries to predict the token after poem, which apparently is some sort of closing note. What I’m trying to argue is that this is just sheer chance, I mean you can only have so many altercations of text.
you can change the TDP of the steam deck, and it yields comparatively very minor performance improvements at 20 Watts of power.
Thanks!
Bit unrelated, but who drew that? The elephant looks sooo cute!!
„Your reactor has been temporarily disabled due to license payment issues. Please consult support@mcaffee.com“
But that won’t happen. Companies have money, and by extension, lobbyists. It doesn’t matter what the general consensus is, they will get their way.
I disagree on the notion that a person that prompted the AI didn’t „make“ the picture. This is the same argument as with digital art, you aren‘t making it, you are simply moving your pen on a screen to create lines and fillings to impress an image. (Also, when it was becoming popular a lot of artists complained that is wasn’t „real art“). To be fair, what someone thinks is art is quite subjective (many people scoff at these random blocks standing around in cities like statues) so it’ll ultimately be up to the lawmakers (that mark my word will lobby to eternity for this to exist) to decide. I respect your opinion, but don’t agree with it. It’s not like you or I can’t enjoy something just because someone else doesn’t.
I disagree on the notion that a person that prompted the AI didn’t „make“ the picture. This is the same argument as with digital art, you aren‘t making it, you are simply moving your pen on a screen to create lines and fillings to impress an image. (Also, when it was becoming popular a lot of artists complained that is wasn’t „real art“). To be fair, what someone thinks is art is quite subjective (many people scoff at these random blocks standing around in cities like statues) so it’ll ultimately be up to the lawmakers (that mark my word will lobby to eternity for this to exist) to decide. I respect your opinion, but don’t agree with it. It’s not like you or I can’t enjoy something just because someone else doesn’t.
I get where you’re coming from about human involvement in AI art. But consider this: the artist isn’t just dropping a prompt and walking away. They’re often curating the dataset, fine-tuning the model, and making tons of decisions that influence the final piece. It’s kind of like a movie director who shapes every scene even if they’re not on camera.
Also, AI art usually isn’t a one-shot deal. Artists go through multiple iterations, making tweaks and changes to get to the final result. Think of it as sculpting, chipping away until it feels right. It takes hundreds if not thousands of different tries with prompts.
And don’t underestimate the prompt. A well-crafted prompt can guide the AI in ways that make the end product unique and meaningful. So while the AI is a tool, the human is still very much the artist here.
When you take a photo, you have a direct hand in making it - when you direct an AI to make art, it is the one making the art, you just choose what it makes.
I understand what you mean, but you’re still directing the Camera; you’re placing it, adjusting the shot, perfecting lighting etc. Isn’t AI art the same? You have a direct hand in making what you want; through prompting, controlnet, Loras and whatever new thing comes along.
Actually, that’s a really good analogy, and it helped me think about this in a different way.
What if the monkey is the camera in this situation, and the training the monkey part is like designing the sensor on the camera. You can copyright the sensor design(AI Model), and the photo taken using the sensor (output), so the same should apply to AI art, shouldn’t it?
I’m not sure that’s the case. For instance, a lot of smaller local models leverage GPT4 to generate synthetic training data, which drastically improves the model’s output quality. The issue comes in when there is no QC on the model’s output. The same applies to Stable Diffusion.
Also in general - summaries are just a strong suit of LLMs right now, and even if the technology doesn’t advance further (which I’m quite skeptical of) it is still an extremely useful tool which will drastically impact so much.
Great explanation, thanks!
Don’t apologize! It’s a really active topic and I’m super interested in other‘s opinions on this!
that seems quite likely.
You are correct. Hollywood will simply change up a couple things and then use the assets.
However, I‘m still undecided about how I think about whether generating AI art should count as Human-generated or not. On one hand, people can spend hours if not days or week perfecting a prompt with different tools like ControlNet, different promptstyles and etc. On the other hand, somebody comes up to midjourney, asks for a picture of a dragon wearing a T-Shirt and immediately gets an image that looks pretty decent. It’s probably not exactly what they wanted, but close enough, right? AI gets you 90% there what you want, and the other 10% is the super-hard part that takes forever. Anyway, sorry for dumping my though process from this comment chain on here xD
edge is chromium based so yes
kid called EU anticompetitive laws: