#🌶|off-topic
1 messages · Page 2 of 1
Yeah we have managed to decouple the two through "technology". Although I would argue, pulling out is not that hard 😄
indeed, evolution is a massive beast - unfathomable in most instances.
We can only scratch the surface of our progenitors
Most of the time, we might find a partial fossil of a partial skeleton, which happened by sheer luck because the creature fell into the right kind of soil before death.
I would love to have general purpose robots when I get older, would be great to be independent.
For sure, the physical job of lifting an elder patient up (to avoid the patient getting bedsores), a robot will be a great asset. The other human nurses can do the jobs that require person-to-person contact. I know they are feeling VERY stressed right now, in my country. One of my best friends is a nurse.
We're literally seeing the automation of our jobs happening in front of our eyes, in real time. Slow, but real.
Lawyers too, will be automated.
perhaps the law won't need to be as intricate and complex, as it is now
Pilots, who have held such a strong position in labor negotiations, are no longer as safe.
At this point, we know the trajectory, but we need a way to make everybody have a meaningful life after the AI has taken over X amount of jobs.
The trajectory is known, AI will take over a lot of jobs. And this is almost like continuing the industrial revolution, like a Moore's Law kind of metric.
So it's not really something new to say.
it will be extremely difficult, unless the process will be gradual enough for the old folks who had mastered the entire skill rendered obsolete to retire
We will have to do it piecemeal, slowly, so that our societies don't collapse.
then younger people will face new skill set requirements, just like what happened after the industrial revolution, then the introduction of computers
I, myself, can't imagine what I would do without a computer, my entire life (including my job) revolves around it
Yes but to me the point that is often missed is the huge amount of value that would be generated as a result. Meaning if we do it right, and manage to distribute the benefits in a relatively sane manner, it could (should) be amazing.
But it's an open question to me whether the general population, average person, can create a sense of meaning without being explicitly needed.
Well, there will always be some humans that will be needed. Especially this early into this.
perhaps we should re-think our framework of equating "being needed" with "quality"
that is, "he's more needed, so he's better"
humans are ultimately not needed anyway, and it doesn't make our lives any less meaningful
Anybody else find listening to a half dozen or more YT channels at the same time increases their productivity? I tried noise generators and nice music, but nothing does it like a good ol' cacaphony.
this discussion about automation is riddiculeus
latent diffusion models don't really automate anything, they also can't directly act in the world
it's just another tool
who writes the prompts and the tools to generate prompts?
who labels the images?
etc.
literally millions of people worked very hard to come up with text or labeling images etc.
Once the tools are made things become much quicker and easier. For example, fiftyone and other tools can utilize ai models to automate the labelling and text very quickly.
Also, diffusion models can, and are, used for automation already.
that's not automation for robots etc.
sure it's automation but so is adding with a computer instead of manually
literally millions of people worked very hard to create computers, which automated a lot of things and made some jobs obsolete... and created other jobs as well
that was the point of the discussion, wasn't it?
no point was the software side (text for training of language models, labels of images)
sure and 10000 people make sure that the power is delivered
etc.
we'll inevitably reach a point where most things requiring human input will be automated, that is, abstracted away, sure, that won't make humans absolutely obsolete, but that can make a lot of jobs obsolete
just like I can automate a lot of things without AI just because I can program a computer
it's not only about latent diffusion models
sure but what's the timeframe
I am sure it's 100+ years
because AI is so damn immature
well, I can generate a nice set of collection cards right now, and I don't need to hire an artist for that
right now
because collection card art doesn't need to be non-generic
I was blown away seeing the performance of program synthesis models in solving competitive programming problems
I really was not expecting that so soon.
well I can generate text with GPT-3 but I can't use it for anything because it's so damn expensive to run
sure the cost will come down
that's not real programming
I think we are just scratching the surface of the potential now though...
In what sense?
problem solving
You mean that it's not as messy as 'real' programming?
basically yes
seems like real programming consists out of 1000's of tasks
not programming in different languages, looking up things on stackoverflow, talking to people about what ideas they want implemented
haha
that doesn't sound right, tbh
why not
unless you're maintaining some bs legacy code
But to me what we are seeing now is just a hint of what might be possible.
or, unless you mean "the entire team" needs to deal with 1000s of tasks
or that the entire project is like that
What is your guyses take on the "chinese room" problem, and what understanding something means and so on?
because yeah, my little game engine might probably go well beyond that in the entire lifetime of the project
just try to enumerate the tasks your doing when thinking about what it should do based on requirements, looking stuff up, updating your knowledge, cutting the specification down to lower level specifications, debugging, reading stuff on stack overflow, coming up with good search queries to Google to solve a particular programming or debugging problem, reasoning about changes without execution in your mind, etc.
it depends on whether you understand it "1000s at the same time" or in the lifetime of the project
because the former is bs
yeah but a really tough competitive programming problem can also contain many layers of deep abstract thinking
as in skills required to program
I mean it depends how you break it down right?
you can group it all together as one skill you call software development
or break it down into each keystroke etc.
good point
I suspect that we overestimate the specialness of human intelligence
I mean as tasks for which one NN may be trained
It does feel like we have something special that cannot be emulated
And don't get me wrong, it is magical and amazing in one sense.
and I guess it's easy to underestimate it even after 70 years of AI research
fair enough, I imagined it as literally having over 1k tasks assigned to you, which would be a sign of development hell
But I suspect it will turn out easier than we expect to reach human level.
Sure but we haven't had the compute before
If our basic hardware technology is not there, there may have been no way of cracking it.
because most stuff in the brain is hidden from introspection
Yeah, I was more speaking to the view that some people hold, that there is some special component in human intelligence, that cannot be emulated by a turing machine basically
so what, now we have the compute to simulate a human brain but the software will not be here in our combined lifetimes
yes the fastest supercomputer has enough flops to simulate all neurons in realtime
Well, then just simulate them?
we don't even know the connectome of a human brain
[citation needed]
wait a second
What do you guys think about consciousness?
epiphenomena that just comes along for the ride with intelligence?
does it have a real causal impact?
information integration?
Frontier has 1.1 ExaFLOPs
can't help but feel it's a bit simplistic
I'm not sure the human brain is the only way to go though.
Mind uploading, also known as whole brain emulation (WBE), is the theoretical futuristic process of scanning a physical structure of the brain accurately enough to create an emulation of the mental state (including long-term memory and "self") and transferring or copying it to a computer in a digital form. The computer would then run a simulatio...
It's just the one path that evolution happened to stumble accross this iteration of it.
I hope so
To me the mind uploading thing seems a little sketchy, I mean we have no idea how consciousness works, but presumably that would just be a functionally equivalent copy of you.
And the original physical you wouldn't actually experience any of that.
Would be my guess.
This simulated version may or may not have a subjective experience, but if it did, it seems logical to assume it would just be like a new consciousness spawned suddenly?
yes
but I would argue that ravens have consciousness too
Sure but some people seem to think that they will be able to live in the computer simulation.
it's conceptually easy to realize
one "just" needs to scan the tissue of a brain with electron microscopes, then reconstructing neurons, axons, synapses with programs, then simulate the neurons on a computer
Sure but that doesn't get around the issue for me.
I wonder how much accuracy would matter there
https://openworm.org/ is related to mind uploading
OpenWorm is an open source project dedicated to creating a virtual C. elegans nematode in a computer.
good name 😄
idk, but I would guess that it's in the nanometer range
synapses are relatively small
I man, inaccuracies are inevitable, the question is, how much would they affect the outcome
I'd still expect the brain to be a fairly chaotic system
I am pretty sure this is unknown
I am pretty sure that any intelligent system has its own sources of emergent behavior
yes, I see no reason why a hypothetical GI system as complex as a human being wouldn't be conscious
it seems to be an emergent feature
maybe neuromorphic hardware closes the energy efficiency gap by a few orders of magnitude https://spinnaker.io/
a lot of processes don't need to be Turing complete at all in a GI system
Although do we really know anything about what consciousness 'seems' like from the outside?
right
I feel like we attribute consciousness to things that are able to trigger emotional responses in us and remind us of ourselves.
But to me it's a total mystery
right
I'm not sure if I'm conscious, tbh
I mean, how do I tell the difference
But to me one weird thing is, it seems that consciousness, meaning just the fact that there is subjective experience, shouldn't have any causal impact on the physical world
intuitively
But if it doesn't, then it seems to me that we wouldn't be talking about it?
except it must have a causal impact
The fact that humans discuss it, is to me a causal impact
yet it seems non-physical in essence somehow
e.g. electromagnetic radiation from your brain because electric impulses jump around
it's possible to detect even, to some degree
We seem to be talking about different things
I just mean, first person subjective experience.
not the things that are correlated with it in the brain.
Also the human brain is a continuous model(i guess more made of sines and cosines rather than linear units)
yeah, and this experience is computed in your brain
which then radiates energy
so it's impossible to separate the two imo
well, a representation of the experience
but the actual experience, qualia
what the hell is that
I've had like 100s of those discussions already haha
Meaning that, why couldn't we have a physical system that behaves functionally exactly like a human, but didn't experience anything
Yeah I've bothered way too many people who have no interest in talking about this in my lifetime 😄
😆
I can't drop it..
And it's so hard to talk about, there seems to be different categories of people, who just can't understand what the other categories are talking about.
I have something better, sort of related: if your eyes grew a new type of photosensitive cells that could detect, say, near IR, and your brain learned to interpret it, I'm pretty sure you'd see a completely new color, which you can't imagine right now
but then, you'd probably never quite understand your past self why you were unable to imagine that obvious color
it would be part of your consciousness, if it's emergent (and there's probably no reason for it not to be emergent)
even though lots of stuff happens unconsciously, your conscious part needs to understand it
To me, any kind of stipulation of how it would emerge just seems so arbitrary. But then again that goes for the whole universe...
what if it's actually necessary for a complex system that can react and act intelligently to develop consciousness
Like if it's somehow related to information integration, information just feels like a concept that we made up and not like a fundamental of the universe that could be included in some kind of specification
And why is consciousness private?
Why am I me and not someone else?
I'm just rambling now, sorry haha..
if there was a way to merge two brains, my guess would be you and the other brain wouldn't notice that you share consciousness
The issue is there is no way to properly study it because of it being private.
We can see physical phenomena in peoples brains that they report with lossy language to be having different effects.
it's too complex
But that's not observing actual consciousness
just something that appears correlated with it.
I'm not sure if we humans generally are talking about consciousness as the intellectual process of thinking about thinking, or talking about consciousness as our experience of being conscious. I think it's hard for us to separate the experience of consciousness from the process of consciousness. Is it the feeling of consciousness that we actually value, even over and above the consciousness itself?
I think that all animals are conscious to some degree, but how developed are they at perceiving or feeling their own conscious processes?
It seems to me that one of our big advantages as humans is a deeper connection between our consciousness and our subconscious processes. In other words, we seem to have a better ability to control our subconscious processes consciously, or to offload processes that were once conscious into our subconscious through experience. In this way, perhaps the quest for consciousness is a fool's errand, if the real important part is the connection between consciousness and subconscious. Modern research seems to indicate that the subconscious talks to the consciousness through an abstraction layer, and we call that abstraction layer "feelings". We get an impulse or a drive or a feeling from our subconscious, and then our consciousness rationalizes reasons to work towards the goal that the subconscious pushed you towards. You could end up going for a job interview, and rationalizing that decision by thinking that you wanted a job there, when in reality you were driven by your subconscious to get closer to the secretary at the front desk. You may know that you're attracted to the secretary, but you may not know that you wouldn't even be on this course at all if it hadn't been for the subconscious pushing you in that direction using your own feelings. At least, that's my understanding of the current state of research.
or to offload processes that were once conscious into our subconscious through experience
this sounds like the more data you have to acquire and synthesize from different parts of the brain, the more conscious you are, once you've learned which connections to use, you become sort of a robot
Sure I agree with much of that, but it's very tricky to talk about.
Yes there does definitely seem to be some aspect of that. When first learning something it is a very deliberate thing, each action involved, whereas with practice it kind of becomes a subroutine.
But this subroutine might have it's own subjective experience, who knows?
To me there is something intriguing about the idea of connecting consciousness to recursion, but I have no idea why.
The recursive loop that is formed with self reflection.
We seem to have a limited amount of consciousness, seemingly by evolutionary design. The truth is that our subconscious is far far more capable of deep calculations and analysis than our consciousness is. Like, when you have a dream, you have a type of consciousness in that dream, regardless of whether you know you're in a dream or not. But, you're just an avatar in the dream and meanwhile your subconscious has created the entire landscape that you're living this virtual life in. Our consciousness is very weak in that way. We seem to file requests to the subconscious for anything more complex. If I tell you to think of an elephant, you've already pictured an elephant without any conscious thought whatsoever. It might take you a little while to draw an elephant consciously. But, to think of an elephant? Instantaneous.
if consciousness emerges from the acquisition and synthesis of data, not necessarily from automatic computations, then that would actually make sense
In this way, to me, the consciousness actually seems more like a project manager that doesn't require huge amounts of intelligence, but does require knowing where to go to get the intelligence that is needed. Other processes tell you if you're hungry or if you're hurt, and all the manager does is decide what you're going to do about that - and passes those tasks off to some other department so that it can get back to project managing.
I agree with all those points, except the fact that I feel like you're describing the functionality that is carried out by the process that is correlated with consciousness.
Rather than conscious experience as a phenomena in itself.
Associated with may be a better word than correlated with.
Could you imagine a process that performs in the same way, but that it doesn't feel like any to be?
But given the same state, acts in the same way.
In a way, I don't think that the subconscious is simply a computer. I think that it is so vast that it has kind of a form of its own consciousness, but that consciousness is alone, disconnected from the outside world, in its own world in your mind. It's only connection to anywhere else is fairly indirect - it sends emotions out, and it gets, and it gets sensory data back, about the body, and about whether the consciousness is rationalizing in the manner that the subconscious wants it to. That's why people can have their own demons that they aren't always fully aware of. Just regulation between the consciousness and the subconsciousness can be catastrophic for people's personalities methinks.
I do find it very interesting to think about what the part of us that is conscious is doing and what role it is filling, but to me it is a separate question from the one I'm referring to.
Yes, it may be that it's a collective of consciousnesses, but that the ones below the surface are not able to express themselves directly.
if it were to receive sensory data, then it would have info about the external world, so not quite as disconnected
No, I can't imagine that, due to what if said about the current state of research. Feelings ARE the language that the subconscious uses to direct the consciousness, and the consciousness rationalizes some reason that it's doing what is doing, even if the reason is false.
Our consciousness inextricably linked to the subconscious through the process of feeling feelings. So no, you wouldn't be able to emulate the same behavior without emulating both processes. That's because, although you think about yourself as your consciousness, that's just a rationalization. You are also your subconscious. In fact, you're mostly your subconscious, and that's why it's communication through emotion is so vital. It's also why people can get in trouble if they consistently ignore their own emotions. That means they are ignoring the language of the subconscious, and the subconscious starts acting out in other ways to prompt the consciousness to take the desired action.
Yes but the way I see it, feelings still have a physical manifestation as some kind of physical signal?
And it doesn't seem necessary for a feeling to feel like something 🙂
You could have the subconscious send this signal representing anger to the conscious part, it can still receive and act accordingly to this signal, without there being an inner experience.
Just a physical system.
A feeling is an experience. If a feeling is conveyed to the consciousness, that's when we experience it. The subconscious doesn't experience it - it just sends it.
I was just giving this some more thought and it occurs to me that the AI that we're using right now is a lot more like the subconscious. It can do amazing things very quickly, and it has so many abstraction layers you could never track down everything that happened. However, similar to many ai's, if you ask for an elephant, it may have one trunk and 4 legs - maybe. But, that's what the consciousness is for. The subconscious is great at mass data processing, but it is not so great at step-by-step analysis. Step by step analysis requires the brain to slow down a little bit, and that's exactly what the consciousness is, the slow part of the brain. If you really need to make sure that everything's in order your consciousness will shut your subconsciousness down and ignore it until it figures out and verifies whatever it's trying to do. If you're in the middle of a complicated calculation and you're subconscious starts mentioning your penis and that you're hungry, you basically just defer that until you've completed your step-by-step analysis. Rationalization requires step by step analysis and that's the main function of the consciousness.
But, perhaps we're not using the right words in the first place. We generally think that we're not aware of our subconscious, and that's why it's called the subconscious. But, now we know that that's not true. If you tell me that you feel sad, you just told me directly something that your subconscious said to your consciousness and you repeated to me. We even use the word feel in a seemingly inappropriate way a lot, when we say things like, "I feel like Chinese food tonight". It's possible that you saw a Chinese food commercial and so you're making a conscious choice regarding Chinese food. But, sometimes you just feel like a burger. I submit to you that this is in fact not an inappropriate use of the word feel. I think that that is the situation where are the subconscious talks directly to the conscious mind. You don't know what vitamins you need, what hormones are to high or too low... And, you have absolutely no idea what your immune system is doing. But, you know who knows all that and can aggregate large amounts of data quickly? Yep, the subconscious. And I think that that is why we express ourselves in that way by literally saying what we are feeling like eating. Consumption of the right amount of things is crucial for life, and most of it is handled by automatic systems. Those automatic systems work on a mass amount of data coming from all over your body and they boil it down into a directional drive towards a specific food.
Eating disorders are pretty obviously evidence that this system isn't working in the most optimum way. Or, an indication that the subconscious knows something that we don't. When faced with an imbalance, the body tries to attain stasis. It does that in whatever way seems the most effective, regardless of weight gain or weight loss. The body knows that stability is more important for long-term survival, or at least survival long enough for you to breed and raise young. Evolution doesn't care if you had a fat father or a skinny father, just that you still had one. If stasis isn't achieved quickly, imbalances go out of control very fast. So, the body determines the bare minimums that it needs to attain stasis, even if you have to lose body parts to do it, or get fat, or get skinny. And, it does all of this through telling our consciousness what it feels like eating or gathering.
Sorry for the long message.
Use discount code "Ryan10" to get 10% off at autoblow . com
@hipablo
@cer spence
filmed by: @Brooks Kossover
JOIN THE PATREON FOR EXTRA CONTENT:
https://www.patreon.com/ryantheleader
SOCIALS:
Instagram: https://www.instagram.com/ryantheleader
Twitter: https://twitter.com/ryantheleader
Tiktok: https://www.tiktok.com/@ryantheleader
DISCORD...
#stablediffusion #ai #stabilityai
An interview with Emad Mostaque, founder of Stability AI.
OUTLINE:
0:00 - Intro
1:30 - What is Stability AI?
3:45 - Where does the money come from?
5:20 - Is this the CERN of AI?
6:15 - Who gets access to the resources?
8:00 - What is Stable Diffusion?
11:40 - What if your model produces bad outputs?
14:20 - D...
Does anyone know why faces or elements are multiplied when I want to make the images in higher resolution? and why the promt elements are merged without separating them by words? it's hopeless. gracias!!!
https://www.youtube.com/watch?v=qVqDGF4AuXQ this is great
Wow can you imagine if time ran backwards? I can't!! Hahahaha!!!!!!!
SUBSCRIBE: https://bit.ly/2Jg9jc1
If you wish to support the channel:
https://www.patreon.com/mattcolbo
FOLLOW ME:
Instagram: https://www.instagram.com/mattcolbo
Twitter: https://twitter.com/MattColbo
Thank you to Prospre for supporting the channel!
https://prospre.io
iOS: ...
I guess I guessed it correctly
Because current model works best at 512x512 - one of the next versions will be better, they know how to fix it. They just wanted to focus on 512x512 first.
One idea is to make an image you like, then upscale it later. 🙂 #1003034183716835418
thank you so much!
I don't think he's about "dominating" in the negative sense, but enriching everyone's lives, and helping revolutionize the world we inhabit, yes.
if some french speaking peeps are around there, there's a smallsmall community about AI art
Man, those AI videos would be baller for my Lofi music
@limpid egret @novel beacon @rich mortar @white sun I think a person may be experiencing some injustice https://www.reddit.com/r/StableDiffusion/comments/wnmadz/banned/ If what he says is true, please be more careful for banning people on the spot like that
I haven't banned anyone yet. And mind you, that's even after I've put up with some borderline abusive reactions for simple, perfectly justified TOs.
anyone I've banned was generating naked kids.
What is his discord name
And I think that's like 1 person
Yeah we can't do much if we don't know who this is
I've banned 4 people so far
Intentional or not though, doing that repeatedly is a sure way to a ban
and all of that 4 are relatd to child pornography and religious hate @fast oak
higher ups have been alerted and are looking into this case
If you know the guy discord name please tell us @fast oak
We really need that ban appeals process...
People can rest assured there will be an appeals process, that's something to keep in mind, too
he says JonJetCoaster
Ask if he really needs us to check what this was.
Yeah just looked.. Ian Watkins is who he generated 🤷
Welp
Not good 😬
this is what he said ""The only one I can think of was Ian Watkins, but it was supposed to be the one from Steps, and not the other.""
there are two Ian Watkins
🤷 Kinda just.. suspicious or whatever.
wow
Thank you, that's what our team reported and I was just trying to dig up.
What I'd suggest to them is to wait for an official appeals process and we can decide how warranted the ban is then.
Right now we just have to enforce the rules we have; things are in flux.
When this becomes open source, wouldn't someone be able to generate something like that on their gpus regardless of the laws?
I mean, the laws still apply, if you get caught, that's on you.
Not a lot different from drawing it yourself
It'd be absurd/impossible to enforce the laws of every jurisdiction in the model, it's open source, people would just change it.
I think of it like trying to restrict what you can do with a pencil, forget comparing to other software, just go basic. You can tell people what's legal but you can't outright stop them; nothing here is that different.
It's perhaps harder to get some things that it's not intended to do than for an artist to do it
wtf ☹️
Good reminder we aren't DALL-E, if there's a mistake we'll fix it but people usually get banned for pretty good reasons, and everyone here can find them.
That prompt is disgusting
I'd love a channel for those who like to generate horror imagery, it seems that SD has a great potential to be unleashed here
wtf
Yeep 😬
I mean, I'm pretty sure the result was... underwhelming for the author, but still
and they have the audacity to create a reddit post not knowing what they did wrong. Gosh the internet is a scary place
If not here, assuredly somewhere, that's something I'm really interested in too
I have plans myself to fine tune a model for horror imagery, TBD if I even need to but it'll be fun
oh, that sounds promising
btw, I have an idea, what if the bot used the "spoiler" tag
so you don't have to look at other people's pictures unless you want to
although that would require a lot of clicking :P
I think the hard part is still keeping people in their "zone" with that stuff when it's allowed on the server, like you wouldn't want to spoiler everything everywhere
Because that, exactly
So it may really be the kind of thing that's best with its own little community, a server that may not even attract enough attention to get flagged NSFW
And people who are all there knowing why they're there
@split temple Ok here we are, in the right channel. 😄
This feels cozy
indeedy
if we were to move to a different server with our horror imagery, then how about calling it "Un-stable Diffusion" :D
@split temple I never learned Maya, but a friend of mine who works at MPC London uses Maya to animate (he animated Simba and gazelles racing down the ravine in the new Lion King movie, and the octopus in The Boys s3 among other things)
Apparently they don't use MEL so much any more, but some kind of python scripting stuff
Oh cool, that sounds sweet. I dabble in fx for fun but that industry seems too manic for me.
You guys been playing with tyflow at all?
It is funny that mobile games is turning out to be one of the most stable.
Never heard of tyflow
what mobile games you worked on?
tyson ibele's new amazing max plugin for fx
oh for Adobe Aftereffects?
or 3dsmax?
It's inside of Max, kind of like a cracked out pflow, really amazing stuff you should check it out
That would be awesome! I know I would sure get a lot of use out of that.
I made the cover for my newest song with MidJourney: https://audius.co/sonderghost/sonderghost-get-it-together-vol-1
@swift mural respect
@deft apex mm for what? xD
for this "Russian politics real face series" xD сделайте Соловьева еще
ty, doing my best, and putting on Twitter , tried Simonyan didn't work, need also medvedev but doesn't look to much similar, good idea to try Solovyov maybe it will be good
good luck! that's epic
The Prompt Games for the AI art community is live now! (Everything below is rules if you know how to play this is all that matters)
16 competitors go head to head to see who is the best prompt engineer.
Winner gets:
30 dalle prompts & Entry into all-star tournament to compete for over $100!
You don’t need access to Dalle, Steven will be running all the prompts. The theme of the tournament will be announced once we begin.
To participate join at the time above and say you’d like to join in chat.
https://twitch.tv/stevenalejandrovelez
Generating ART with Artificial Intelligence // Competition for DALL-E 2! !socials !tournament
this must be stable diffusion https://twitter.com/arvalis/status/1558623545374023680
Its called stable diffusion and it does not seem that they have any safeguards in place to protect artists. This sucks lol
3340
180
RJ seems to always be on top of AI and nfts, isnt he?
ah advocating to generate cp to 'try to get SD shut down' twitter folks be crazy
@lofty bolt what do you think of about this?
we have a number of thoughts as you may imagine but there is some interesting responses to that already
the thing that makes me worry is alot of artists on twitter seems to hate AI art, and they will told their audience and AI art can end up like nfts hope it doesn't happen though
Yeah so many people are mad in that comment section
I don't think it can ever end up like nfts, especially since the influx of new people (YouTubers, influencers etc) who can use this to create art and expose new people to it in a positive light are going to FAR outnumber people that complain, and general artist followings
worth a read https://twitter.com/Dan_Jeffries1/status/1555527402419154944?s=20&t=_wQ2O3JKN3CZcVZWQq-10g
In my latest article I explore the fantastic new world of AI art generators like #StableDiffusion and #dalle2
Why they're poised to change the world and why their critics seem to get it so wrong.
163
I've already seen videos using midjourney that just casually use ai art as a replacement for any other art and it isn't acknowledged except in the description. I assume more people will take advantage of this tech as time goes on and it is more accessible
that viral twitter post is just infuriating lmao. so much confidence for someone who doesn't know how the ai works
yeah
...Because the world needs yet another thing to be divided upon. Sigh. Why can't I make memes of gandalf fighting a massive hotdog in peace?
Feelings = clicks = … = money
It’ll happen for the rest of your life, no point letting it get to you. One can also find positives in the human organism being able to hold opposites in mind at the same time.
@lofty bolt when your model is finally public it will break the internet even more than dalle mini i want all of those AI art haters get drowned with AI art and their claimed of it just copying, mashing image together, and stealing get dismissed completely, i want that happen, just few more weeks... they seems so mad they might sue stability.ai so it has to be done fast
hmm
where have I seen this before
if the uk bans ai art il think about moving to another country
With almost infinite possibilities of ai art how would they even be able to check of every image is ai generated or not that's my question.
true
I bet they're just trying to keep some sort of control as governments usually do. Although I think they're attempts will only restrict and slow down progress
I think they will build a kind of classifier
Ai beats ai
Have you guys tried CapCut for adding 3D zoom effect to your AI art?
I hope SD or MJ have this feature so we can add more advanced 3D zooms to 2D pics.
Here is a music video by using CapCut.
https://www.youtube.com/watch?v=GGeAZjR19dw
This Music Video is made without coding and high-budget.
Here is what I've done.
- Generate wonderful pics fit for your music
Get your pics by using AI tools, like Midjourney, Stable Diffusion, DALL·E (still waiting for an invite), Parti (upcoming one?) etc.
I've been using Midjourney and Stable Diffusion for making this video. I've learned s...
Missing the watermark
!dream an ultra high definition pastel coloured photograph of a real life unicorn eating and ice cream. Refraction, volumetric lighting iridescence. -n9 -s150 -C15
Whoops 😅
ah, things are gonna get fun over the next few months/years
That would be a hilarious trial to watch. “But your honor, I was doing it to get this AI art tool banned so it wouldn’t put me out of the job! Surely that counts as self-defense?”
I guess the subject has been discussed many times, but wow, the hatred about these tools is out of proportion!
It's not even AI but Machine Learning. I can't imagine the level of intolerance we will reach when a real AI is born.
The S-F culture has accustomed us to see hatred on robots.
I don't count the number of movies where we see a human-like robot being destroyed, often in an inhuman way.
But what I didn't consider is that we would probably hate as much those who use these tools / AI.
When there's a clash over SD or AI, you'll be the one to take the brunt of it.
the proposed legislation would do the opposite of ban ai art and training data. governments occasionally get things right. the world is on the cusp of a new industrial revolution and consultations as written up above can help usher in visionary pragmatism. nobody wants to be left behind the curve🤞
The worst thing is that I have seen people who preach tolerance for their way of being come and complain about the latent Diffusion. It's really a reminder of how small a human being is, a dirty egotist.
latent diffusion is just effective in building structures, there is more to ML than latent diffusion
Interesting.
I wonder if we'll be able to port all this to something other than PyTorch and Python, like C# or C++.
With the hardware nearly evolving as fast as the improvements in this area, I can imagine a lot of great applications.
CPU's are like maybe 50 times slower than GPU, even with CPU-optimizations
In terms of hardware, I was talking about graphics cards
ahhhh
well PyTorch is just a wrapper for Torch which is C++, but I don't have experience in touching it directly
Indeed, it is the overlay of Python libraries that may make the porting difficult. There seems to be a lot of them. At least on Disco Diffusion. Some have probably already done so
We'll see when it will be open sourced.
Okay.
Ethics issues here with economic impacts.
In 40 years, we have no idea where the world of image generation tech will be.
If we aren’t careful with regulation, SD could have the blood of an industry on its hands.
Key word: could
so as i see it, is this will cause a similar transformational shift as digital audio production had on music, and the end result of that was making the art process more accessible to everyone was an overal positive
Ideally yes
But again, it’s a scale issue
DAWs sped things up, let’s say tenfold.
SD speeds things up, as you said, probably around 1000 fold
i disagree with this, on the point that this is happening with or without SD, all SD is doing is making sure corporations cant privately control it
That’s fair enough.
because corporations are doing this anyway, and what they do with it will be far worse than anything independant artists do with it
That’s actually a phenomenal point.
Now, I ask, how long do you think the industry has before its radical change/“death”?
for example, imagine a special app for your favourite superstar, and the ai is locked down so all you can do is make the kind of content they want, and the superstar intentionally builds toxic parasocial relationships with the fans, exploiting peoples passion for free marketing
I wasn’t arguing that, you’re undeniably right about it.
SD is the least of all the evils.
i know, i just figured providing an example of how this may unfold is useful
id say SD is necessary, and impecable with its timing
as no one has had time to get established yet
so it forces this into the light and starts a debate
which has already started
as ive had this conversation this week in multiple discords
Of course, the human art industry will still survive, just as horses did. Only a shell of its former self, but still alive and kicking.
and its good that people are thinking about these things and openly discussing these things
as if we wanna shape the future, now is the time to act
It’s just a matter of how long until the proverbial automobile replaces the proverbial horse
still need people to manage and work on the automobile or tend to the horse
i truly believe long term, this will be an overall positive for artists, due to how it will break class barriers and make creativity more accessible to everyone
machines sometimes even do things wrong, and that's why engineers are needed
a refit of a tire done by an automated process still needs a human to check it's work
the music example is what i keep coming back to
But considerably fewer engineers. An industry is still being completely replaced, either by machines or other humans
It will take out stock artist, but it will never replace original artists, because we'll lalways be able to make things AI can't
Will we though?
How do you know we always will outperform the machine in some way?
Photography replaced illustrators that made etchings for publications for sure, but it didn't wipe out art. On the contrary it expanded it.
And my response to this is simple; what if you could press a button, wait 15 seconds, and have 4 mixed and mastered songs ready?
Instant EP.
Because the data set is based on the things we make. Unless AI starts huffing it's own farts, we're always going to be the source of the data.
The AI has already begun “huffing its own farts” as you put it.
Not this one, but many language models have started doing this and gotten faster as a consequence
In a year’s time, image generation will do the same.
Yes, in order to make AI art, but there will always be a distinction, at least for a very long time.
How long is very long?
A century?
It's not like it will wipe out all artists overnight, have you seen what users are making on here?
My prediction stands at 40 years
I am not saying it will
Even with AI, 99% of the users are not creating amazing pieces, they're just having fun.
I’m saying the human art industry will deteriorate over the next 40 years until it is a mere novelty.
AI will never replace artists, because AI cannot put emotion and feeling into what it makes
you require the human element to add that
This would be compelling if the people who pay others to make art cared about this
this AI will largely be a way for people to get inspired and become artists in their own right
If anything, stock artists will easily adapt to using AI generator as a tool. It still takes art training, the fundamentals, in order to distinguish good imagery from bad.
You will always need to understand composition, value, color, context, art history, communication, in order to use it well. Most people are making Emma Watson fanfiction
For now
All of this is the status quo. I’m talking about 40 years from now, which I will hold as my timeline.
One thing I'll say as an artist, if you are using AI, you need to be transparent about it. Trying to pass it off like you painted it is no good. It would be like trying to pass off a photograph as a painting, you will lose all respect.
I’m talking about how this is still a terrifying prospect considering I want to make music of my own and have it heard by the world
There is nothing in place enforcing this
But remixing, collaging, painting over AI images, that's all fair game.
If there were something in place to enforce this, many of my concerns would be alleviated
Or being an AI purist, like a photographer, and saying hey, I made this with AI, I took this photograph with a camera, etc.
But as it stands, there isn’t, and right now is probably the optimal time for policymakers to start thinking about this.
It doesn't need policy, it's just cultural ethics
As a digital/traditional artist myself, going from traditional to digital was a hard move but it opened up so much more possibility to me and made things much more accessible. I can do CG work or make animation without having to own a $9k program that can only be licensed for a year. I'm working on a piece that I am using Diffusion for reference/help with because google search is not specific enough for me to actually get something close to what I want/need. I want to use SD to help me work on my weaknesses as an artist and thus improve.
SD is just a new tool for me to play with, and so long as we treat it as a tool then I don't know if it's going to ever truly outpace humans, especially as humanity itself gets smarter and more creative. And the thing is, the paid side of the art sphere with commissions and such is a luxury thing, but people pay for artists to make things because they 1. like their style or 2. like how they approach working with the client. If people want to pay you for something, they will pay for it. But with SD people could figure out prompts easier and be like "Hey I'd like something like this but drawn by YOU" because that's how it goes normally just with walls of text as a prompt.
You assume businesses are above lying to their consumers
I mentioned in another channel, but the best way to frame AI art for others to understand is that it's very similar to photography.
And tbf, I'm still learning a LOT about this stuff and figuring out my own morals/ethics when it comes to this, but it's exciting and fascinating to think about.
Fair point, and in fact the end of that is much like what I’m planning on doing with SD before it hits the fan.
My point is, I hope policies are implemented (not overly strict ones mind you), and I hope to god your optimism about this is well placed.
Think of what a photographer does, they choose the subject, they choose the composition, the location, lighting, etc. That's exactly what we're doing, we choose the subject, the environment, the lighting, etc.
speaking of, if you havent you should take a look at the EUs AI Act
Oh sweet policymakers are already on it
Similarly, a photographer takes a ton of photos, then goes back to the studio and picks out the best ones, because they don't have complete control over the subject. Same thing with AI art, we generate a ton of images then we pick out the best ones. The editorial process is half of it.
Of course, since it’s the EU, they’ve done a shit job at it
this one specifically targets limiting some obvious nafarious implimentations
its taking a similar enforcement approach as GDPR
info's here if you wanna take a look
The thing I like about SD is that it's in our hands, the artists hands, and it's not going to be gatekept from those who aren't artists initially either and thus THEY get a chance too. We have the chance to grow this smart, we didn't with how digital vfx grew and was immediately squeezed to death by the movie industry, but now we can work at this as normal people who barely make cost of living wages and I think that's interesting
Again, I hope your optimism is well placed
Fraud is fraud, but just because something can be used for fraud, doesn't mean it's not an artistic tool. Just look at photoshop. I could use it to fake a license, but that doesn't exclude it from being an artistic tool.
I understand the worry, but right now, in the MOMENT, I am going to be optimistic. Time changes everything and I completely understand that fear and I am having it myself in places, but we should at least try right?
For right now it’s good. In 5 years or maybe even less, it might not be. We’ll just have to wait and see.
Good hunting.
heres a breakdown of some key things too: https://datainnovation.org/2021/05/the-artificial-intelligence-act-a-quick-explainer/
The liability right now is public opinion. And I think OpenAI and Midjourney have been wise to be cautious, because AI is very vulnerable in this beginning stage, if public opinion turns against it, it could cause real damage.
A lot of these things relate to fraud.
yeah it doesnt really touch much of what creative ai is gonna do
but its also still being drafted iirc
so we shall see what comes of it, if anything
Actually I take that back, not fraud, but deployment of AI in deterministic roles in government
Nothing to do with art generation
For our purposes, it's more about cultural acceptance, public opinion. Creative legitimacy.
Also, I know artist still making daguerreotype and lithographs that are successful. There will always be traditional artists.
I suppose
Does anyone know if the 800M model is fp32? Because I think its better they quantize it to 8bit to save memory.
Made this to demonstrate interpolations as described in this tweet. https://mobile.twitter.com/xsteenbrugge/status/1556099146762919937 - prompts are “Joe Biden exhaling a large smoke cloud, featured on artstation”, “The world on fire, trending on artstation”, and “Barack Obama exhaling a large smoke cloud, featured on artstation”. I made this using the code on the stable diffusion repo. It’s currently using the fallback 400m model but when the SD weights get released it should be able to be used in the same code. Prompt conditioning vectors as well as starting codes were spherical linear interpolated. The results were then ran through topaz Video Enhance AI to smoothen the animation.
Turns out #stablediffusion can do really awesome interpolations between text prompts if you fix the initialization noise and slerp between the prompt conditioning vectors:
437
I had this in mind a few days ago, just from a theoretical stand-point and have seen lots of examples of it today. It’s just as awesome as I could have hoped.. Nice. 👍
In the future with more compute and better optimized models it would be amazing to have a screen on the wall just random walking through the latent space in real-time.
And if you want you can nudge it in different directions with your mind via neural link. 🙃
No need then for people to take the risk associated with ingesting a psychedelic substance in order to at least experience the visual component of it.
Yeah you could have a huge list of prompts (even just have an ai trained specifically to generate SD prompts) and you can just keep interpolating through that. Would be pretty awesome!
I was thinking of how Data (from Star Trek TNG) would actually ever work. Data knows how to paint, so it's probably been trained on a huge trained set of art - the history of all art. It's also capable of learning, so when it's alive/running, it would see new art and continue to train itself on whatever it sees - look at that art and train it into its model.
I guess Data's the ultimate copyright troll / art stealer.
aren't we all? I mean, artists are "inspired" by others.
Exactly. Filmmaker here. Human creativity is imo a weighted dataset of genetics, nurture and experiences. Those define the "taste" of an artist, which is virtually the same as the nuances, advantages and flaws of datasets that models are fed with.
In my humble opinion, we should look at models trained with the dame dataset like humans who grew up in the same surroundings.
Bottom line is: if Results of ai are to be declared virtually the same as human creativity, you either have to assume that human creativity is nothing more than many parameters in a given Input, or that ai is indeed sentient and as "transcended" as a human is.
Of course, the argument against that is that even SD can't tell you how to raise an emu, or how to perfectly bake a cake. Once this kind of general ai is achieved, we'll have very different perspectives on ai imo.
Sorry for my ramblings
Just look at what someone said https://mobile.twitter.com/arvalis/status/1558623545374023680
lol image generation isnt real AI
Clearly 0 idea about AI in general. Maybe trying to understand it instead of attacking with false arguments should be better. For now thanks to this we have to say bye to the Stable Diffusion Living Artist list and we can only see the passed away ones. At least for now. I hope one day will come back
Yeah, I hope so too. you know, these anti-AI artists like this just make me sad.
the problem with this is that its better to know exactly what it is capable of, than to artifically tie our hands behind our backs because of concerns people have
intentionally not gathering knowledge is never a solution, and is always a red flag as to something being wrong
BUT
yesterday i had the same conversation 3 times, in 3 separate places yesterday, of people having concernns about how ai art will disrupt artistry
which is frustrating, as its a distraction from the real conversation we should be having
the real conversation we should be having is what corporations will do with this stuff, and how we can organise to deal with that
and SD is the exact kind of necessary disruption we need to mess with the corporations plans
the alternative is this stuff being closed source and only accessible in the way corporations decide it should be
as creative AI already exists, theres no undoing it
sorry but as ex-pro concept artist on some companies (and currently material artist) I don't buy neither this shit and neither their false arguments:
1- They are not in the right of telling "you can't copy/imitate X artist" while they're been imitating and copying and referencing the old masters not for years... for more than years...
2- They want to monopolize the art creation to just digital drawing/traditional drawing, same happened when first Wacoms appears on the market. Digital was the devil, I know I know...
3- Stopping this is like stop internet to exist. You simply can. Adapt or move aside, simple as that. We all have to adapt to search for jobs, projects, etc. The "formulas" used before didnt get where you want at that time, now is the same just people easily forget
Wikipedia: luddites
This is not a new thing
and on the independent artist level, what this actually means, is removing class barriers, and making it more accessible
which, just like digital music production, will be a huge positive to art as a whole
corporations will use the technology for bad things, but thats a completely separate conversation that we should be having instead
Reality is that AI will get more jobs created than destroyed, period. Creativity will be a more valuable weapon than it's now (prime on technical nowadays for most companies). And for fuck sake, any artist can adapt to this as we will have the same possibilities which is something a lot of us didnt have when started first steps in the industry (at least me). It just makes me mad people thinking of me using AI I lack of foundations and my work value is zero, when they don't have a single clue about my background...
Yup. Internet is horrible and tumor for the society, same as the social media for some people. Reality is that is not. Depends where you look at it. When you can actually benefit in several ways by them... Of course you can use them to non so healthy things for your mental health, etc. But that's another story...
im saying this more from the perspective that in discussing these things, its possible to organise and work towards effective action to deal with this stuff
but the first step has to be people having these useful conversations, instead of getting lost in the liberalism demonising individuals using plastic bag trap
yeah but won't happen. I totally got your point from the beginning. Working as a single one instead of fighting and avoid understanding the principal elements of the AI and what it means.
aka, they should be concerning themselves with what corporations are doing, not making the independent artists life harder
the irritating thing is the places ive had this conversation in, are leftist communities
aka, communities that absolutely should know better
giving workers shit for simply existing in the environment they are in, is not a valid use of time
honestly right now, my biggest concern is how marketing could use this stuff for parasocial exploitation
as it'd be very possible to package these tools in a very specific way that will only benefit the marketing campaign
therefore exploiting peoples passion to get free content
not enough people are aware of what a parasocial relationship is
and it means they are vulnerable to being taken advantage of
well twitter, facebook, etc. already using AI (non artistic ways) for bots, bad messages, see how people reacts to certain news, fights, etc. That's not new and probably you know about it. Best you can do is use you brain in that case when situtations like those appear.
yep, and this stuff neatly fits in with the stuff like snapchat/tiktok filters thats already happening
its the next obvious step down into the hell that is social media
that's why is good to read, cultivate, go out, etc. Sitting in front of a computer (24hr I mean) will make easier people to fall in their "traps" but IMO most people (or at least people with brains) should be easy avoiding all that shit. We will see what future brings to us but for now we can be quite happy with SD and DD and either MJ or DE2
i for one am rather excited for where this transformational shift in art will lead
and hopefully those who are concered about it putting artists out of work will join us in the excitement once it becomes clear that wont happen
i find it ironic artists are crying that they will lose their jobs, and point fingers at the tech, instead of pointing fingers a their bad employers who want to make more profit... (who are the one that will fire them)
much easier to blame this poor AI algo, who can't defend themselves, than to pick a fight with the employer who will fire you anyways. if they do fire you, they simply don't value you, find another job that will value you ?
that would be the same without AI....
automation isn't the same as AI... people are funny
you still need a human to guide the art, it doesnt simply do itself
Car gets invented
carriage drivers : we are gonna lose our jobs (now they became taxis)
*TV gets invented *
Radio hosts : we are gonna lose our jobs ! (still have radio...)
Photography gets invented
Illustrators: we gonna lose our jobs ! (still have illustrators, you can't photograph something that don't exist)

we really ought to be more careful when make arguments to assuage artist’s forebodings.
for one, the lack of distinction between professions of necessity (e.g, radiology, surgery) and professions of fulfillment (e.g, painting, drawing, music, fiction) is jarring to a lot of people; they think you’re silly when you compare their concerns of losing the latter to those who lamented losing the former. The humanities are important for human fulfillment, and so many of these people are concerned about bigger things than job displacement. Their worries simply aren’t addressed with that one.
I think this machine learning married to capitalism will result in changes which meaningfully affect our relationship to the humanities, but I think it’s a adaptable and a necessary consequence on the pathway to scanning one’s arm with an in-app camera and identifying cancer.
I’m not sure how to convincingly address the concerns I’m seeing. Smarter minds than mine with a solid machine learning background will need to lay that groundwork
The thing is that I cant really empathize with people telling "dont put this artist study public!" when they (and me included) learned copying and imitating the old masters (as every single pro artist did). There is no a difference. This is against the education per se. They want the next-gens to don't do the same they did cause they will be using a different tool set. Sorry but their "concern" is just a manipulation. Its like they want to still make art in the caverns with sticks.
I can understand people who don't know what AI is to think that is dangerous, etc. but if they're neither wanting to read and understand first how this really works (and after understanding, critic it if you want) what can we expect? It's like explaining chemistry to my house walls.
The stability diffusion community account is back
Thank God.
people pissed about AI art need to remember: the human characteristic is what sells. False Start didnt go for 80 million because of the craftmanship, or even that it looked good. It sold because it captured the artist's intent better than any prompt could've
@plain edge You're running SD locally right now?
How? I didn't even know it was out yet.
The weights are released to some people
Ohhh, damn. Well, enjoy your early access.
I think the weights will be released quite soon to the public though
And you can do whatever you need on the bot here unless you actually want to mess with the code
@hard jewel TEST
anyone know where I can find more information on harmonai?
could you try runinng on your cpu to benchmark?
I tried to change to CPU but there are some other things on GPU automatically so there's an error. I don't have the time to debug
kk thx
discord is so fucking stupid. it timed me out in this server on the mobile version, and i was fine on pc version; but a restart from moile fixed it... i wasl trolled by discord
Have you tried wombo dreams app yet
What I call a hypocrite
I'm so sure extensively means he's already using it for his work or to get inspired. But in public much easier to shit on it
Now I know what to get my family for Christmas
I feel that the copyright issue is not even the real problem here. Create a competitive model trained only on your own data, and you'll probably get the same reactions.
People are scared that they'll lose jobs - somewhat understandable, after all, it does bring some uncertainty - but copyright is just the low-hanging fruit here. Merely a scapegoat.
is it simple to install and run? any idea how long would it take for "-s150 -n9" on a 3080 ti with 12gb?
hey @raw lily . I saw you were streaming the other day. you had a GUI for disco diffusion, I was wondering if you could tell me more about that. I too have a rig and wish to run models locally.
eventually. https://github.com/CompVis/stable-diffusion - this is what it looks like for now
ooo --init-img will be sooo fun
absolutely! now if folks would stop complaining in general (speaking in general) and smack those buttons and keep putting in the work to train the classifier they will release this sooner.
how do you help train the classifier??
by reporting the false positives and NSFW that slip through. keep generating and giving good feedback, this entire beta was to train the model and everyone here is still doing it.
see here #📣|announcements message
but how can you reveal a blurred image to see if it's a false positive?
@outer hare you can check out aave.com, klimadao.finance, any articles by vitalik buterin, the book radical markets by glen weyl, articles about money legos, articles about ENS, podcast called Greenpilled (book by same name) by kevin owocki,
there's a growing ecosystem here - it's not BTC + shitcoin roulette anymore.
i did wonder that, i saw something that was probably innocent blurred but, cant be sure...
need the local model to test it ourselves
that will come in time, this is just like day one after the change. took 9-10 days of training in the beta to get to here.
just keep reporting. you can pretty much tell if a car is blurred is there is no fleshy blob there on the hood. it's probably a false positive.
Around 7 seconds a generation for a single 512x512 image at 50 steps 11 GB 2080 Ti
It's simple enough for someone with some Python experience. Also, there are going to be Colab notebooks which you can just use.
nsfw sonic
First would look like nightmare fuel
what is a prompt for this? I wanna see it
"realistic sonic the hedgehog photorealistic highly detailed fur detailed facial features heraldo ortega, octane render, trending on artstation " -C 10 -n 4
hehe thx
and seed?
I wanna get exactly this
"realistic sonic the hedgehog photorealistic highly detailed fur detailed facial features heraldo ortega, octane render, trending on artstation " -C 10.0 -n 4 -g -S 873624718
-S 873624718`
-S 3937584502
-S 2476037265
-S 3633924840
What is the collab that adjusts eyes in portraits? Anybody know?
Morbius the Hedgehog
Anybody wondering how to check what's behind the blurred curtain, try adding a coma here or there in the prompt along with the seed
!dream "cute, dramatic sky, sun , empty sahara desert , morning , prison far away , gouache matte painting by pixar studio makoto shinkai subtle frazetta shinkai ghibli " -n1 -W 1999 -n 4
🙌🙌If anybody is interested in architecture images follow the insta account @diffusion_architecture also looking for submissions
@pure estuary can you add me back, I was told by someone you have an answer for my question
#1005627862721769512 is the best channel 😤
im having a hard time thinking of an art use cases where artists will actually be replaced. maybe that weird abstract art on math textbook covers but i cant think of anything else
#1005628033945837620 users 
Loneliness, made by AI 🖤
I see concept artists getting possibly devalued if not outright replaced in very specific cases, not because AI generated stuff is comparable, but because it will be cheaper.
you don't need high accuracy until you get to production assets, or reference for 3D modeling. Smaller studios have people doing all three so they will be fine, bigger companies might have dedicated concept art departments that could be affected.
it'll take a few years for any of this to ripple through the industry though. Ai may be fast, but bureaucracy and culture are slow beasts.
larger firms could theoretically replace a team with one or two people who understands prompt engineering and have some photo shop skills.
However, I feel people really underestimate how important understanding the fundamentals of art is. Sure anyone can make stuff with AI but if you don't have a good handle on composition, color theory, anatomy and perspective you'd be hard pressed to "correct" outputs
Trained artists can use AI as a jumping off point and a labor saving tool. everyone else just has to take what they can get from the prompts.
it'll be a niche to fill for sure.
Wallpapers in monitor product photos are going to be ai generated for sure
ah yes those ad ridden HD wallpaper sites will be rejoicing
must suck for those whose dream in life is to make desktop wallpapers
their job just got way easier
here is a false positive #1005627987435192381 message @plain edge
🙈 hahaha very nice emoji shorthand
Gotta go fast
oh yeah good point
This is "Emma is freed from the avocado" by Magrathea on Vimeo, the home for high quality videos and the people who love them.
I have access to dalle 2. If anyone has any requests just dm me and I will try to fulfill them!
Those are already often computer-generated.
yeah now the people who make those have more advanced tools to work with
Math textbooks are an interesting case though, because you don't want a high-quality artistic image. You want something that looks interesting and has teachable mathematical meaning. Simpler algorithms are good for that.
Now machine learning textbooks, those are going to have some interesting cover art.
Hey guys, I'm on the road to 1K, would love some of your help
https://www.instagram.com/misterruffian.ai/
Welcome back to Instagram. Sign in to check out what your friends, family & interests have been capturing & sharing around the world.
well.. just coming here to say i'm hyped. I said FSCK IT and bought a 3090.. i'll be listing my 12gb m40 and probably slaving out the 24gb m40.. if it will even fit in the case. I'd rather keep the anti-sag bar as that's what stops them from dying.. the PCB likes to flex under cooler weight and it breaks the BGA joints at the DDR RAM near the PCI slot and then she toasted.
GO HEAVY YO!!!!!!
damn bro I couldn't generate a 256 image with 3070, maybe l should save up for 3090 too
yeah.. I was suffering from long delays on 512 and also just wanted a faster card for my transfer learning("fine-tuning") needs so just went ahead and did it. put it on the card. I've got more than a few money making repairs/projects that are done this weekend so it'll work out.
@limpid egret 
lmfaoooo
@limpid egret

New episode of the SANDMAN today 😉
@restive root Are there plans for a music model? Like for example I like A sky full of stars by coldplay, i can just write a prompt: sky full of stars epic orchestra cover and get a good result.
Music is way easier than images and is way way less gpu intensive
ah that exist wow
Is there any way today to put an image into an Algo and have it spit back the approximate tags and settings you would need to generate it?
Like a reverse diffusion of sorts...
Dall-e 2 Generation
Stable Diffusion Generation
Is there even a comparison here?
Im sorry, but i value Stable Diffusion more
@restive root wouldn't you agree?

FYI "detailed realistic digital art, illustration of 💋[any animal]" gets you at least one super beautiful portrait 50 of 50 times in DALL-E
dall-e wins this one for me, the images are more artistic
OH SHIT GUYS #1010300608575709234 message
How the hell does dall-e wins for that one
IT IS CLEAR NIGHT AND DAY
:D
Its a matter of adding more descriptors, these look better imo
agree, now SD wins
u shouldn't compare with the same prompt, each of them speaks a different language
i'm a pro photographer & aesthete yo, i'm tellin u eeehehehhehh
Provided to YouTube by CDBaby
Big Cheeseburgers & Good French Fries · Blaze Foley
Sittin' by the Road
℗ 2010 Lost Art Records
Released on: 2010-05-15
Auto-generated by YouTube.
The sound of airpods running out of battery is one of the most annoying sounds in existence... right? 😄
Any good ai art discord servers?
@main mortar
@green scarab yo, check fren req if you have time.. I wanted to chat about *nix pxe boot
FYI https://wiki.archlinux.org/title/diskless_system
Good luck
ahh super. thank you. 🙏
I have a !dream 😢
SD's are good but they don't look natural, they look like oversaturated tourism pamphlet photos
Grass in the bottom left generation is basically radioactive
One message removed from a suspended account.
I'm also interested
random but im streaming violin stuff on twitch, anyone interested in listening to the violin?
hell yeaaaaa
@limpid egret "do you wont me to speak briish with you mayte"
Oyill speak briish with you mayte
Does anyone have a Colab (or other) link for face swap (images, not video)? thanks
anyone here play angband
@faint flume ive got some test renders ive tried to emulate that style if youre interested
though its more 1990s PC than PSX
stuff like these
i had some traditional psx style renders somewhere here
hell yeah love that isometric pre-rendered sprite look
yeah looks fantastic 👏
Def interested in your process for these but we can chat more later
thats one thing ive always been trying to emulate
mhm! np np, these are all blender made
cheers mate

V I B R A T E
in spain it is recommended to sign a contract to have sexual relations in case the woman decides to falsely report you and fuck you up, if that happens, you are guilty until proven otherwise
, thanks to the great ministry of equality. (law made by irene montero of the communist politycal party of podemos)
@fast oak I think it's best to discuss things like this outside of the server.
but is off topic
Not really appropriate though, I'm sure this is something you could discuss in DM's if you feel passionately about it.
ok
Yikes. That's a lot of money for a cell phone plan
Ehm hello

We found a quite place
Interesting, by a mere coincidence, OpenAI is lowering their price, just got an email 
Sorry, it was maybe irrelevant (?), as it didn't include Dall-E
yeah, another #🌶|off-topic to set up camp in
⛺
in this chat anything can be irrelevant
and no its not even irrelevant
lol
very convenient timing
hehe
Radioactive grass :s
How long did it take to generate 9 images?
it depends on the steps, u can have up to 500 steps
so for 50 steps it takes about 3-5 sec per image
for 150 it takes 10-15
speaking at 512x512
havent seen the 1024
Try out the 1024, I haven't installed the colab yet, will do tomorrow
wait it can work up to 1920
interesting
im going to go for 1024x1024 at 50 steps
for some reason it gives an error for anything beyond 640x640, maybe because of ram issue , im pretty sure that would get fixed in the future
but it took about 25 sec at 50 steps to generate an 640x640 image
but 3-5 sec for 512x512 which is interesting lol
just shows you the potential for optimizations and speed
Interesting, hope they fix the 1024x1024 soon, but we can always use an upscaler
except for fine tuned models
:D
What have I missed?
ah seems they are the same model
its actually the same model of the bot :D
@near idol you click on runtime then click on runall
there is also a place where you have to write ur username and token from huggingface
it will show it to u as the collab runs
tested the seeds , the ones from the bot & the one from open s ource v1.4 they are the same
Well Just a sec let me check (thanks for the help)
Are you here? the video I attach showing the same as before. I don't see section to input a username
that moment you try to clean the console but generate an image instead...
This is the correct one?
Ok thank you champion checking it now
A former user here is spreading some sort of 'beware' or accusations about the volunteer mods in the Stable Diffusion server. I only wanted to at least say something before some unnecessary and potential smear campaign gets flung to some innocent mod's way. 👆
Ok so to much error on the way I give up. I will wait for a version that is more user friendly
I think I'm addicted to this AI, I'm not a python programmer and much less an AI developer but this got me really excited to hack around with it.
gotta look for some good python and AI learning resources now lol
This is the first time an open source project get me curious enough that I want to do something with the code they gave lol I'm sure I won't be the only one feeling this way.
Just realized, setting the default seed in the colab to "42" was probably a joke
#nice
Need GUI
is this local?
anyone run that on a 2060 yet? lol
wtf lmao
Nice, so is that local on your machine? What machine/gpu?:)
running on RTX 2080, 8GB, Windows 10
Ok cool. I have an Mac Studio Ultra and I’m planning to try it. As far as I understand, the entire ram on the machine can be used as VRam.
Quick and easy way to build your own particle detector. Full instructions at http://www.symmetrymagazine.org/article/january-2015/how-to-build-your-own-particle-detector
Have fun!
Filmed in CERN's S'Cool Lab
let me link Visions of Chaos again for those looking to run SD on windows with GUI - https://softology.pro/voc.htm -follow the steps outlined here https://softology.pro/tutorials/tensorflow/tensorflow.htm to get you up and running
Softology Visions of Chaos
Softology Visions of Chaos
Now let's build an echo chamber
what does it do?
how big is the total download?
with all the models for the various txt2img tools/scripts I'm at around 400GB
It maximizes the echo you hear
the app itself is only around 300MB
impossible to build easily, right
🧐 it's just an engineering problem, maybe replicate a cavern shape than bounces the noise at the same time, I guess that's how churches do it, or concert halls
Like a mirror of sound
🤔
dood this takes 400gb of disk space? ist this a joke? theres better alternative for getting gui to work
I think I started off around 109GB when messing around with LatentDiffusion and Disco. the more scripts/tools you try the more models it will need
VoC is not specific to Stable Diffusion, it includes all sorts. if you want SD only there will be other alternatives out there
in terms of txt2img alone there's various iterations of Aleph2Image Aphantasia Big CLIP CLIPDraw CLIP-GLaSS DALL-E Deep FuseDream GLID-3 GLID-3-XL GLIDE Illustrip JAX minDALL-E MMC OpenAI PixelDraw Pytti ruDALL-E ruDOLPH Stable Text2Image Velocity VQGAN VQGAN+CLIP
@thin knoll Timed out for using a derogatory word.
throw error?
Shut the fuck up
How???
Are you cruising for a timeout too?
No I just wanna know what he even said
Timeout, 1h, for telling a moderator to 'chill'.
Don't escalate this further.
damn people whats all the fuss
in the midjourney discord
Hey, is there a stable diffusion discord bot?
A long long time ago (~2 weeks ago) you could use the midjourney bot here but staff disabled it when Dreambot was available for beta testing. Now there isn't any bots in this discord at the moment.
We used to have a craiyon bot as well at one point
Wowza, that looks like it got intense
Where has that dude been for 2 weeks?
Well, shit is happening so quickly it feels like beta was months ago haha
dunno if it makes sense to ask here. but, can a pc repair shop sell the pc parts which are being replaced, if they happen to still be in good condition?
i was thinking of the replaced gpu and mobo being sold to get a better cpu. and maybe sell the old one too
i was told "why not just buy a new pc". i would if every part was better for the price i bought this one
i wish my community could have the MJ bot lmao
call you local shops and see what they can do, but they will need to make money from it so I doubt they will be paying much.
you can also sell online.
Well I have news for you
si
españa es una mierda
No lo niego es así
gatopan
Can you suggest me alternates to TacoTron2 and Tortouse 8w
What does the scale parameter do?
@limpid egret here's the rest of the body
AQ3D
It had been 2 years i didn't play that game

lmao, took me a moment to realise what this meant lol
Can somebody tell me why everyone on reddit gets butthurt so easily? I realy dont mean to offend anyone and got banned 2 times in 3 weeks....
First time for answering "Whats a cool gaming Name for a dude called Mike (Anything)" - "Mike Dike" 7 Days - Hate Violence , like what. he said anything.
And now on the permanent Nirvana sub after saying, people tend to hate people when they get stupid rich....
Hey folks, this seems like the place to do this
I created and pushed an awesome list of resources for the existing field of image synthesis (aiArt)
A LOT of it is stabilityDiffusion related, prompt engineering tools, guides, tutorials etc etc
Would love feedback, comments and pull requests!
https://twitter.com/altryne/status/1562595772113121281
I created an @awesome__re list for #AIart and #imagesynthesis tools.
You're welcome to check it out, suggest changes, and subscribe to notifications!
https://t.co/sfQmccBVPm
Hey I’m a drop this discord link for Kandinsky-12b if anyone wants in. It is far cruder that SD but can make some great images from English/Cyrillic text https://discord.gg/ghnn7k2HU5
I've been with them for a few months now, they are ok but I don't see any updates in the future.
California is banning ICE Vehicles from being vended, a major drastic measure for their overall goal to terminate all carbon emissions from the state. And i am all for it
In 1964, I was making my first documentary for television, Bluegrass Roots – a one hour special for the forerunner of PBS called National Educational Television – NET. I was 23 years old and I was headed to the mountains of North Carolina, to Asheville, to meet with and film 82 year old Bascom Lamar Lunsford who I had written a letter to, asking...
R.L. Burnside, Mississippi bluesman from the Hill Country, performing Poor Black Mattie in 1984. Harmonica player Johnny Woods can also be seen.
With the removal of Infinity Train from HBO Max, DVDs are being scalped for $600
So there is GUI of stable diffusion that will run on my pc? without a filters and so on... ?
Hey! We've made an image generation & search platform: https://inspirama.io/ and are looking for some feedback!
DM me up if you would like to generate, I'll give you free credits!
Inspire, Discover Create
Rip cant open it on mobile
free credit pleas
Niiice more and more ai generators!! I love this!!
@tiny tinsel lol someone with my same username
Just the name of my cat XD
Are dictators allowed? if not i will delete it
why not?
dont know, i get banned for seemingly everything. yesterday i got banned on the nirvana sub for this pic
I rather ask now
yeah reddit is a hellhole
okay i know im walking the line, but is this okay?
no its not lel
i cant upload
might be better
It was a perfect digital painting of tailor swift with her nipples censored , no artifacts
This but without bikini
lol
worth it
i figured lol
Is there actually a Nsfw place? I have fire stuff
not here
Where?
This server doesn't allow NSFW there are people who made their own servers though
https://www.youtube.com/watch?v=3Z9fMg1v-D0 many SD users running it on their own computer
This is what happens when too much eggplants ! BONK!
The music is from Sexi Parodius OST: Song Speed Bath!
Hey, does someone know how to see the seed that was used in Google colab :")
Lmao. Guess they didn't find that amusing.
@rugged warren there's a new trailer of atomic heart
get banned on a open minded sub reddit for a open minded opinion is they way of the reddit.
That used work before now it doesn’t
bitch I'm in
Wishlist Atomic Heart 👉
https://store.steampowered.com/app/668580/Atomic_Heart/?utm_campaign=AtomicHeartCombatTrailer&utm_source=YouTube&utm_medium=textdescription
A system failure at Soviet Facility № 3826 leads machinery to rebel against the people. You are Major P-3, Private Agent, and your task is to minimize the consequences of this 'sy...
This looks like you have your CFG Scale pretty high. I would consider dialing it back a little.
What’s a good psychobilly band
the ai has spoken, this is the best meme
maybe one day we will be blessed enough to understand it
Student loan forgiveness undermines one of our military’s greatest recruitment tools at a time of dangerously low enlistments.
19718
3281
Thought that was satire lol
Mask off
What kind of name is Bethesda
MOND!
Good. Nobody should be forced to become a contract killer just to make ends meet.
funny moustache
search, Pool of Bethesda.
Wait what
It means something like house of mercy
it's from Hebrew, it's a pool in Jerusalem and is mentioned in the bible.
Great name for a suburb
I think there are some places in the US with that name so yeah XD
Anyone else find themselves listening to Adema again for some reason? Just me? Hahaha
My friend’s hosting a Prompt Engineering tournament in about 50 minutes! It starts at the top of the hour.
All you gotta do is click this link when it’s time and type into chat that you’d like to join.
All I know ov is Maryland yea
starting soon!
With the great news with India here is some awesome mixing from Sicksick with Indian roots
Sickick's Official SickMix Part 4 ft. Sean Paul, Eminem, Lil Pump, Tyga, Post Malone, Bia, Gwen Stefani, Trinidad Cardona, Kanye West, Farruko, David Morales, Yard Club, Ed Sheeran, Daft Punk, Eiffel 65, Steve Void, Gorillaz & More
🎵 Latest Spotify upload: https://spoti.fi/3IZVjBa
👕 Merch: https://bit.ly/sickickmerch
😷 Spread the Sickness 😷
👾...
PS If you didn't know, this kid has an amazing voice. Look up his originals.
So apparently, Discord is allegedly IP-banning people's accounts because they changed their avatars.
Might be because a lot of people are generating new avatars with SD and putting them on their profiles at this time.
for context: discord has been falsely ip banning people for changing their pfp or anything on their profile, DO NOT change your profile picture for the time being and spread this around.
https://t.co/i0nsAn7fp3
https://t.co/L9h0ykLWz9
https://t.co/32VuC9EI8I
6648
5538
Oh fun. Thanks for the heads up.
What a bunch of twats. They allow a bot that eats up egregious amounts of bandwidth and server storage to exist but then IP ban people for changing their profile image? Unforgivable.
Edit: I rest my case. See messages below.
Although I do wonder how reliable this is, since it is from Twitter.
The Midjourney and SD bots, presumably?
Yeah, basically. Moreso the Midjourney bot, since it has been going for months and seems like it is not going anywhere anytime soon.
I suppose I may be jumping the gun though. It is hypothetically possible this is uninentional. Though it seems like a very odd thing to have happen by accident.
Turns out it's apparently due to a bug in the latest update to the official desktop client, causing the client to effectively spam Discord's servers with packets.
Yeah. The web client seems to be unaffected.
(And it's a good thing I've been using the web client for at least the past month by now.)
The client is so buggy. HTML/JS apps packaged as desktop software always suck. Just as bad as the same crap on android. Browsers disquised as apps.
Good to know. I rest my case against them.
Another good example of why you should NEVER take any claim, allegation or really anything at all that you read on Twitter to be true until you find ample evidence to support it.
Agreed. When I made the @everyone ping on my servers, I specifically mentioned that the source was iffy.
I'm surprised I have not seen a electron app for SD yet.
Right. Or anvil like with MindsEye
how are you getting resolutions like this?
mine all look like clay people
I guess it’s the prompt 🙂 I would suggest that you visit https://lexica.art search for pics you like and Analyse the prompts for it!
@restive root OpenAI Dall-e 2 server really muted me for 3 hours for mentioning that i know a person that has 85 dall-e 2 accounts 🤣 🤣 🤣
they are hopeless
waow!
how many members does that server have
still dont have dalle access
this is next level psychological warfare
stability people are blocked!
joke for obvious reasons
about 40k members in total
lol
still dont have it
waow
1275 credits a month to generate thats over 160$ of free credits every month
and that is not counting the increase of emails being accepted
its so weird for dall-e 2 and their policies
it shit bricks honestly
nothing is like stable diffusion's openess :(
ya think?
these things grow exponentially not linearly.
people are waking up about Stable Diffusion in a rate unlike anything the world has ever seen
you saw the reddit's page growth? the best for the entire month
hmm. Stability also had an issue with people signing up for multiple accounts for dreamstudio.
it seems that there are selfish people everywhere.
holy shit Ariel is offline, what is this sorcery 
holy shit GPT-3 does Homer and Marge talks to a tee
I asked for a morally suspicious group
and well, it doesn't get more shady than those guys
lmao
I read that with their voices in my mind and it was great XD
Pretty cool results.
beautiful, thanks for sharing
