-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
AI can now do hands (Midjourney V5 launch)
- Thread starter vestan
- Start date
- Tech/Gadgets
no, that isn't how it works.
if individual images made by actual artists were discernible in the output of this stuff, you wouldn't have issues with hands and fingers in the first place. it's not trained on images of weird hands!
Quoted for later pages.
Not happy with this development, while inevitable, it will fuck over MANY people. Same as Chat-GPT and all the other bots. Those who wish for this can fuck off. Tools should be helpful, but already I am seeing music artists use AI scammers for their music videos and more. Using styles that people worked hard for to create and differentiate themselves from other artists.
So yeah, no love from me.
yeah i don't see a future in which there aren't countless widely available open-source versions of this sort of thing.
you can regulate use of the output but how far is that going to go? would be interested in hearing what gebru and gupta had to say about this but the tech itself is clearly out of the bottle. it's not magic.
Reminder that there are a lot of artists that post on this forum. If someone can't post respectfully about the issues surrounding AI art, I personally feel they really shouldn't be posting at all. This is peoples livelihoods you're demeaning, people with families and rent to pay. It's a games forum primarily, a medium which itself is full of examples of human artistry.
This includes arguments about democratising art and lines such "artists are very skilled and I respect that but". They're not arguments, they're inherently extremely dismissive.
This includes arguments about democratising art and lines such "artists are very skilled and I respect that but". They're not arguments, they're inherently extremely dismissive.
Inserting doubt into every picture and video that we see on the internet has the potential to be catastrophic,imagine if this AI tech matured right before COVID broke out,all the thousands of AI bat pics in China would become ammo for chuds to spread racism against asian people and it could even affect the vaccine campaign effectiveness,by the time people came out and said "it was AI generated" the damage would've been done,onto the next racist AI gneerated pic.
We can't just cross our arms and say "if you fell for it you were already dumb"
We can't just cross our arms and say "if you fell for it you were already dumb"
what a moronic comparison. Computers are also very dangerous then we shouldnt invent them at first place
To be fair Art_3 brought up a very good example of how the tech could be misused in a catastrophic manner.
Governments are being too slow to regulate art/sound AI generated content. We only last week had a Ghostbusters thread where a videomaker had used an AI generated voice of Ernie Hudson on an article they put out on Youtube. It was very convincing, the only thing really pointing it to being AI was the studio-level quality of the recording and the fact that there was no chance it was an on-record interview.
It's a good reason to be concerned, as Art quite rightly said if this had been around in 2019, 2020 may have been very different for all of us.
i'm a professional writer. chatGPT etc threatens my work but i don't mind people posting about it or getting excited by it.
i fully expect it to turn my profession upside down so i understand the fears. ultimately though it's up to me to understand it and beat it.
V5 as in hands now mostly have 5 fingers ?
Impressive.
Scary.
And boring.
The first few versions of Midjourney produced interesting results: those images being wildly incoherent, resembling dreams, producing images that returned a general visual idea of the prompt had some charm. This is just uninteresting.
Still, how this gigantic theft of copyrighted and sensible material was/is tolerated is beyond me.
Impressive.
Scary.
And boring.
The first few versions of Midjourney produced interesting results: those images being wildly incoherent, resembling dreams, producing images that returned a general visual idea of the prompt had some charm. This is just uninteresting.
Still, how this gigantic theft of copyrighted and sensible material was/is tolerated is beyond me.
I admit this is a abrasive example but i just can't even bring myself to even be impressed by it,i'm just terrified
I'm not saying don't excited about it, perhaps a little less dismissiveness about its implications might go a long way though. It can be disheartening to read, even if some of it is through naivety.
I don't know what writing you do, as a writer - I work in story in animation, so I am back up against the wall in two ways. :D We need to be careful how we discern complementary use of AI and the use of AI as a replacement.
Regardless, I feel you shouldn't be talking about "beating" it, or "genie out of the bottle" as it's the lack of regulation which is basically creating this AI versus creators environment which is incredibly unhealthy. Just my view, at least.
Maxey brings up a very important point. I think it could do six of them and add a knee at a strange angle.
You can still get that, just need to specify it in your prompt. These pictures were specifically created to be as realistic as possible
I can't tell if this post is a parody/satire post with its insane use of hyperbole or not. If it isn't this is ridiculous
i hear you and i think it's important to be real about the serious implications, which the "out of the bottle" talk is part of. i just don't see how this tech could ever be regulated out of existence. you can't trace the output back to the model's input, so how could it ever be enforced (and that's if scraping images/text is even found to be illegal, which itself would be difficult)? AI models are going to be open-sourced and there's no way you could ever regulate against that.
we should be thinking about how to understand and use the tech ourselves, and focus on what we can do that AI will never match.
The tone of AI 'art' enthusiasts always comes pretty spiteful. A certain incredulity that there are actually people who are paid to do something they love and they they should consider themselves lucky to have ever lived through a period of time were they were able to get away with charging for it.
The art of the future will be brought to you by the least creative, most uninteresting people you can imagine (or can't imagine in a lot of cases I guess, which is half the problem). People with absolutely nothing to say and no vision to share. It's pretty sad to see people so pumped and excited for this future. You'll get what your asking for in the end I suppose.
The art of the future will be brought to you by the least creative, most uninteresting people you can imagine (or can't imagine in a lot of cases I guess, which is half the problem). People with absolutely nothing to say and no vision to share. It's pretty sad to see people so pumped and excited for this future. You'll get what your asking for in the end I suppose.
that link is a bunch of people saying it should be illegal. not saying that EU courts wouldn't find in their favour, but it's very much not settled precedent, and the class-action suit brought in the US was extremely flimsy.
edit: also that doesn't address the rest of the sentence, which is that it's going to be impossible to enforce this stuff even if it is found to be illegal.
I know this thread is about AI drawing art but Youtube's been pushing ChatGPT videos about coding to my recommendations and while it's not as simple as the art stuff since you still need to understand what it's doing, I worry about up and coming devs looking for junior positions lol.
Sorry, I wrote IS but actually meant to say "should be made".
It isn't technically illegal because there isn't a regulation at all, obviously since this is unprecedented.
This is exactly what EGAIR is asking for, their manifesto explains exactly what could and should be done about it.
yeah, all i said was that it'll be difficult to prove the case. i totally get why people think it should be illegal but again, would be hard to enforce against scraping even if it was banned
I've brought this up in a similar thread from a more "pro-AI" perspective and I think it's probably the best analogy there is for this situation. Nuclear bombs aren't really a misuse of the tech because they're fundamentally the same technology.I think it makes more sense to frame AI as a "discovery" rather than an "invention". Machine learning is basically the "oh shit we can split the atom" moment of nuclear research except distributed across many years of people saying "oh shit we can train CPUs".
In both cases, the only difference between the discovery going great and going sideways and hurting a lot of people are a few tweaks of the numbers and how awful the people in charge of them are. The former is a math problem and those can be solved. The latter is a political problem and that's more of an uphill battle imo.
I'd also argue that for both AI and nuclear technology, it's the government's responsibility to directly compensate the relatively small amount people most affected during the tumultuous period of turning the discovery into things that can conceivably help all of society. In a just world, we would distribute all nuclear-powered energy and awesome new AI discoveries (along with the profits from the saved time) to everyone along with a nice bonus for the people providing the training data and dealing with radioactive materials.
Do you have a link to this? I'd like to hear their perspectives and it's kind of hard having any sort of take without a more concrete idea of what the context is for the "out of the bottle" quote.
I was gonna complain about the definition of "owner" until I got to this part:
Now this is something I can get behind. Especially because it upgrades "owners" to "legitimate owners" and there's a huge amount of data where there's a difference between the legal owner and the legitimate owner once copyright law stops being the guiding rule. My problem with the lawsuits in the US are that they're based on US copyright law and US copyright law is already completely fucked in favor of the corporations who will be screwing us over in the future with AI.
In terms of enforceability, it sounds like it would be pretty much limited to medical and other sensitive personal info which is fine by me since you start running into trouble objectively defining "legitimate owners" of data beyond that. If "training rights" function like existing IP rights instead of existing human rights, then this runs right into back the issue of the ability for those rights to be legally bought from people under sketchy terms.
I think it's so interesting that it's the white collar jobs that are going to be threatened first when for a long time many thought it was the blue collar jobs.
I'm going to need someone to take a boat load of acid, start talking, and someone writes it in the prompt. My body is ready to feel fear once again.
this is a really good point. medical data in particular i think is one where you could force any companies in the space to publish their work and sources because the scope of where they're getting it and what they're doing with it is going to be inherently limited, and there's already a high standard of regulation for them to abide by.
i don't think anything like that is ever going to be practical for art generation tools, though.
Yea this is scary.
Also, people saying they dont want to see threads like this anymore... I dont think it is a good idea to ignore news on A.I advancements. I think it is important to stay up to date on how fast this thing is moving. It is best to educate yourself and practice safe surfing when online especially now that a.i can impersonate loved ones. Scammers will be banking on you ignoring this stuff. Dont be the mom or grandpa who doesnt "know how to use a computer at all" of this generation of tech.
I know it's rough to see, but just try to peek into these threads every once in awhile so you arent blindsided one day.
with that being said.....
I do think we need an a.i. tag tbh. But for a different reason.
Also, people saying they dont want to see threads like this anymore... I dont think it is a good idea to ignore news on A.I advancements. I think it is important to stay up to date on how fast this thing is moving. It is best to educate yourself and practice safe surfing when online especially now that a.i can impersonate loved ones. Scammers will be banking on you ignoring this stuff. Dont be the mom or grandpa who doesnt "know how to use a computer at all" of this generation of tech.
I know it's rough to see, but just try to peek into these threads every once in awhile so you arent blindsided one day.
with that being said.....
I do think we need an a.i. tag tbh. But for a different reason.
Last edited:
I go to meetings and have had to write reports on the use of AI among other things (specifically related to coding, and dev productivity). Most threads here are celebrating it and it's gross, hence why I want a filter.
It's difficult to enforce most laws. Are you saying we shouldn't have those either?
Nobody is saying there won't be any more data scraping once laws are passed. There are still murders even though murder is illegal. But that shouldn't stop us from trying to course correct the current, deeply problematic trajectory of AI development.
I'd sign up for that. Also all of the "genie out of the bottle" comments that paint AI as some sort of inevitability that no human or law can change the trajectory of. It's such bullshit.
There's a high standard *in theory*. Maybe this is the real doomer take, but I'm coming more to the conclusion lately that regulation kinda just doesn't work. Companies made a big show of following GDPR at first, but in practice they have a huge number of ways of getting around them:
- Lobby lawmakers to water down the legislation before it passes
- Support right wingers who will dismantle the regulatory agencies once they're established
- Bribe the regulators and use other less direct methods of regulatory capture to not get investigated
- Cut off service to the country with the regulation entirely if they're a small enough market (Belgian lootboxes for example)
- Use manipulation to get people to consent to waiving their protections
- Lie to regulators faces until they get slapped with a fine
- Eat the fine because it's pocket change in the grand scheme of things
And all of that is especially true in the US. We don't even have the original GDPR, much less GDPR 2: AI Edition. If we ever get to a point in US where passing a legitimately well-crafted GDPR 1+2 bundle seems politically viable, we might as well go all in and really transform society for the better in some seriously radical ways. GDPR also doesn't address the issue of intelligence agencies generally having carte blanche to ignore privacy laws if they say the magic phrase "national security"
UBI (which is not really radical enough imo) would be a lot more direct and involve a hell of a lot less paperwork than trying to deal with the impact or art generation tools via copyright. At best, you'd end up with some kind of insanely convoluted royalty system like Spotify that could only be used on the shittier, online-only tools. That's a whole other angle to this that hasn't really been brought up: tracking how bad actors use AI tools requires well... tracking their data. The possible implications of data mining people's prompts and questions for any of these things are staggering, especially if they sell the bots as friendly and trustworthy.
That's one of the main reason I try to stick to offline tools whenever possible tbh. There's potential downsides to other people using the same tools offline (and maybe training them offline too), but I think we can deal with the repercussions of that just like we can deal with the repercussions of everyone having access to encrypted chat apps. The alternative is that everyone's data goes unencrypted through a few centralized services and that's plainly the worse option to me.
well, last i checked you couldn't murder someone with algorithms on an airgapped computer and leave no trace, but that's how AI models already work
I think it's important to separate the theoretical underpinnings of these algorithms and the specific ways that they are trained, deployed, and monetized. I think the former are what were inevitable as long as humans continue to study the practical applications of math. We have a huge amount of control over the latter though.
Continuing the nuclear analogy, the genie is out of the bottle there because we will forever live in a world where it's possible for someone with enough physics knowledge and money can dig radioactive rocks out of the ground and make a dirty bomb with them. That specific issue is still a driver of international politics to this day. I expect AI will be a concern for humanity for just as long if not longer, but that doesn't mean that I think humanity was doomed to eventually drop a nuke on itself. That was all America's fault.
AI "art" is rife with problems
But at the same time, I enjoy seeing pictures of Carl Winslow as a Gundam pilot
But at the same time, I enjoy seeing pictures of Carl Winslow as a Gundam pilot
We really shouldn't. We should only use spice to increase our own -human- computational power.
This is quite the move forward, for better or worse. It's definitely still not perfect, Indian lady is missing a pinky, there several cases of stubby fingers or sausage fingers or just oddly smooth/blurry texture on them, but it can much more easily pass an at-a-glance test or even mild scrutiny.
Makes me wonder if they just told it "give humans 5 fingers on hands unless otherwise prompted" or something to that effect.
Makes me wonder if they just told it "give humans 5 fingers on hands unless otherwise prompted" or something to that effect.
All I know is I'll be using this in due time.
View: https://twitter.com/kortizart/status/1636136818914762752?s=20
View: https://twitter.com/kortizart/status/1636136818914762752?s=20
Truly written by someone that has no idea what AI is and where it's used. You yourself are already a user of AI without even knowing it: the phone you use, the car you drive, the TV you watch movies and shows on, all of those already use AI tech.
Also, AI is proving to be an incredibly valuable tool for doctors to diagnose medical issues.