db0 set up an AI image generator bot both on the Threadverse and Mastodon some time back for anyone to use. All one needs to do is mention it in a comment followed by the text "draw for me" and then prompt text, and it'll respond with some generated images. For example:
@aihorde@lemmy.dbzer0.com draw for me An engraving of a skunk.
Caused it to reply back to me with:
Here are some images matching your request
Prompt: An engraving of a skunk.
Style: flux
The bot has apparently been active for some time and it looks like few people were aware that it existed or used it
I certainly wasn't!
I don't know whether it will work in this community, as this community says that it prohibits most bots from operating here. However, I set up a test thread over here on !test@sh.itjust.works to try it out, where it definitely does work; I was exploring some of how it functions there, and if you're looking for a test place to try it out, that should work!
It farms out the compute work to various people who are donating time on their GPUs via AI Horde.
The FAQ for the bot is here. For those familiar with local image generation, it supports a number of different models.
The default model is Flux, which is, I think, a good choice
that takes English-like sentences describing a picture, and is pretty easy to use without a lot of time reading documentation.
A few notes:
-
The bot disallows NSFW image generation, and if it detects one, it'll impose a one-day tempban on its use to try to make it harder for people searching for loopholes to generate them.
-
There appears to me in my brief testing to be some kind of per-user rate limit. db0 says that he does have a rate limit on Mastodon, but wasn't sure whether he put one on Lemmy, so if you might only be able to generate so many images so quickly.
-
The way one chooses a model is to change the "style" by ending the prompt text with "style: stylename". Some of these styles entail use of a different model; among other things, it's got models specializing in furry images; there's a substantial furry fandom crowd here. There's a list of supported styles here with sample images.
db0 has encouraged people to use it in that test post and in another thread where we were discussing this, says have fun. I wanted to post here to give it some visibility, since I think that a lot of people, like me, have been unaware that has been available. Especially for people on phones or older computers, doing local AI image generation on GPUs really isn't an option, and this lets folks who do have GPUs share them with those folks.
Humbly requesting an explanation for the polarized takes on this. How are we so specifically split 50/50 rejoicing for the utility and cursing the 'slop' at the same? Someone sway me to a side? I'm addicted to reserving judgement
on the one hand, this is an ai horde-based bot. the ai horde is just a bunch of users who are letting you run models on their personal machines, which means this is not "big ai" and doesn't use up massive amounts of resources. it's basically the "best" way of running stable diffusion at small to medium scale.
on the other, this is still using "mainstream" models like flux, which has been trained on copyrighted works without consent and used shitloads of energy to train. unfortunately models trained on only freely available data just can't compete.
lemmy is majority anti-ai, but db0 is a big pro-local-ai hub. i don't think they're pro-big-ai. so what we're getting here is a clash between people who feel like any use of ai is immoral due to the inherent infringement and the energy cost, and people who feel like copyright is a broken system anyway and are trying to tackle the energy thing themselves.
it's a pretty thorny issue with both sides making valid points, and depending on your background you may very well hold all the viewpoints of both sides at the same time.
Both sides having valid points is almost always the case with issues of any complexity. I'm very curious to know why there isn't a sweeping trump card that ultimately deems one side as significantly more ethical than the other
Great analysis tho--very thankful for the excellent breakdown unless you used ai to do it or if that ai is ultimately not justifying the means adequately. No actually I'm thankful regardless but I'm still internally conflicted by the unknown
no matter your stance on the morality of language models, it's just plain rude to use a machine to generate text meant for people. i would never do that. if i didn't take the time to write it, why would you take the time to read it?
I think there may be two exceptions to that rule.
Accessibility. People who may have issues writing long coherent text due the need to use some different input method (think about tetraplegic people for instance). LLM generated text could be of great aid there.
Translation. I do hate forced translation. But it's true that for some people it may be needed. And I think LLM translation models have already surpassed other forms of automatic software translation.
There are always exceptions/outliers to any rule, it's basically playing devils advocate to bring them up, I never care for it, like a conversation about someone murdering someone for funsies and saying "but there are cases where people should be murdered, like the joker from batman"
Just doesn't apply to generative ai, what do they need images explaining the text lol
But these are neither problems of the technology, nor of it being hosted. It's an issue of the person using it, the situation, and the person receiving it, as well as all their values.
Not sure why people are directing their hate against the tools instead of the actual politics and governments not taking the current and potential future ones seriously. Technology and progress are never the problem.
the problem with entirely separating the two is that progress and technology can be made with an ideology in mind.
the current wave of language model development is spearheaded by what basically amounts to a cult of tech-priests, going all-in on reaching AGI as fast as possible because they're fully bought into rokos basilisk. if your product built to collect and present information in context is created by people who want that information to cater to their world view, do you really think that the result is going to be an unbiased view of the world? sure the blueprint for how to make an llm or diffusion model is (probably) unbiased, but when you combine it with data?
as an example, did you know that all the big diffusion models (stable, flux, illustrious etc) use the same version of CLIP, the part responsible for mapping text to features? and that the CLIP part is tailored for and trained on medical information? how might that affect the output? sure you can train your own CLIP, but will you? will anyone?
I see what you mean, but is there any evidence that the models are biased in a way that affirms the world view of the owners? If I understood you correctly? I couldn't find any.
I'm as sceptical of the capitalist fuckwits as you seem to be, but their power seems to me to be more political/capitalist through the idea of AGI, than through the models themselves. Something that simple sequestration could solve. But that's on the government and the voters.
I'm not sure about the point you are trying to make with CLIP. It's not a topic I'm familiar with, but also seems to be more a problem of the usage and people that the technology itself. Naïve usage by people who want to follow trends/the cheapest option/just something that works in any capacity.
For me, the issue lies first in the overhyped marketing which is par on course for basically anything, unfortunately, as well as the fact, that suddenly copyright infringement is fine, if you make enough money off of it and lick powerful boots. If it was completely open for everyone, it wouls be a completely different story IMO.
Also, I do not think that the models were created with the goal of pushing a certain narrative. They lucked into it being popular, completely unexpectedly, and only then the vultures started seeing the opportunity. So we will see how it evolves in that regard, but I don't think this is what we're seeing currently.
sorry, i had to think for a while about this one.
so, this is an interesting point. we know they are biased because we've done fairness reviews, and we know that that bias is in line with the bias of silicon valley as a whole. whether that means the bias is a) intentional, b) coincidentally aligned or c) completely random is impossible to tell. and, frankly, not the interesting part. we know there is bias, and we know it aligns.
whether or not the e/acc people at openai actually share the worldview they espouse is also impossible to tell. it could also be just marketing.
as long as the product is sold as it is today, i believe it reinforces that power.
ESL moment... i don't really understand what you mean by sequestration here. like, limit who is allowed to use it? i feel like that power lies with the individual user, even though regulation definitely can help.
agreed, which is why i as an "abolish copyright law" person am so annoyed to find myself siding with the industry in the cases ongoing against the ai companies. then again, we have "open weight" models that can still be used for the same thing, because the main problem was never copyright itself but the system it exists within.
the purpose of a system is what it does. some people with a certain ideology made a thing capable of "expressing itself", and by virtue of the thing being made by those people it expresses itself in a similar way. whether it is intentional or not doesn't really factor into it, because as long as the people selling it do not see a problem with it it will continue to express itself in that fashion. this connects back to my first point; we know the models have built in bias, and whether the bias was put there deliberately or is in there as a consequence of ingesting biased data (for which the same rule holds) doesn't matter. it's bias all the way down, and not intentionally working against that bias means the status quo will be reinforced.