Community MetaSteam | June 2023 - halfway through the year

Status
Not open for further replies.

Mivey

MetaMember
Sep 20, 2018
4,330
12,286
113
I guess Valve is just making sure they are not open for lawsuits due to the use of AI trained on copyrighted data. On the other hand, I wonder if this approach wouldn't simply encourage devs to lie about it. As these systems get more and more convincing, we might quickly reach a point where even other AI system have a hard time telling apart real images from fakes.
 

Lashley

My ho ho hoes 🎅
You mean like the boat quests? They’re pretty boring but they’re short and a lot of the filler quests are spread out between eikon battles so its not really a big deal.
It completely fucks the pacing. They introduce some big story element and how urgent it is, then have you do busywork, that they don't even change the animations for to show it's a side quest
 

Kyougar

No reviews, no Buy
Nov 2, 2018
3,315
11,818
113
I guess Valve is just making sure they are not open for lawsuits due to the use of AI trained on copyrighted data. On the other hand, I wonder if this approach wouldn't simply encourage devs to lie about it. As these systems get more and more convincing, we might quickly reach a point where even other AI system have a hard time telling apart real images from fakes.
I recently read something like "AI art is getting worse because more and more samples the AI uses is from AI itself. Practically AI is inbreeding and developing a Habsburg Jaw.
Dunno how correct that is, but it would be hilarious.
 

Nano

Indie games
Nov 25, 2022
1,078
4,767
113
41
Japan
I don't blame them.. the main models which 99% of the other models are based on are straight up trained on billions of scrapped images from the web.

I guess Valve is just making sure they are not open for lawsuits due to the use of AI trained on copyrighted data. On the other hand, I wonder if this approach wouldn't simply encourage devs to lie about it.
If a developer decides to lie about this im sure they got other red flags.
 

C-Dub

Makoto Niijima Fan Club President
Dec 23, 2018
3,992
11,886
113
I think the possibilities of using AI art is exciting, but right now it’s a copyright minefield, home to all sorts of techbro grifters reeling from the crypto collapse out to get rich quick by sidestepping creative talent.

I fully expect the actual application of AI art will be used in two ways:

1. For writers and designers making concepts, prototypes and mood boards before the artists get involved (or are even hired)

2. Tools for professional artists to train with their own work (in a specific art style which they or their employer owns), which they can then use do asset creation at a scale not possible before, even with the biggest art teams.

But the average developer who thinks they can side-step paying an artist when shipping their game is in for a rude awakening.
 

Durante

I <3 Pixels
Oct 21, 2018
4,082
19,691
113
I personally am not sure if I agree with the intensity of the hate on AI image generation that is common in some parts of the internet. Of course, I also come from a background of already thinking that copyright in general (and copyright by corporations in particular) goes too far, and I fear a further overreach in this particular area.

Obviously the copyright issues will need to be figured out, but (and this is where some people will vehemently disagree or even scream at me) I am not convinced that just because an AI model was trained on a given image (as one in billions) there should be any copyright claim by the author of that image on images generated by the model.

Human artists also look at lots of images and then draw new ones based, among other things, on what they have seen before.

People seem to have this idea that AI image generation works by taking some parts of images wholesale and reproducing them with slight changes, but that's just fundamentally wrong. A very easy way to see this is by noting that the average stable diffusion checkpoint has something like 2 GB of data, while being trained on hundreds of terabytes of images. What I want to illustrate by that is that the impact of any individual image (or even 100s of images) on a given output image is minuscule -- similar to or even less than what could easily happen if an artist had looked at some specific images before working on something new.
 
Last edited:

thekeats1999

MetaMember
Dec 10, 2018
1,462
4,261
113
My problem is it is not the corporations that are getting Fucked over by AI. It’s the already underpaid artists that are getting hit here.

it’s the voice actor who refused to take a job because they wanted her to sign over the rights to use her voice in AI.

the artists that are getting less money or work because a company is using AI artwork in their next mobile game.

I am not inherently against AI, Nvidia useage in their upscaling technology is amazing. I am sure the AI used in artwork could help with the likes of texture work and the more tedious sides of the job.

But NFT/Crypto has taught us that a lot of people will get fucked over before protections are in place.
 

ezodagrom

JELLYBEE
Nov 2, 2018
1,888
4,936
113
Portugal
www.youtube.com
I personally am not sure if I agree with the intensity of the hate on AI image generation that is common in some parts of the internet. Of course, I also come from a background of already thinking that copyright in general (and copyright by corporations in particular) goes too far, and I fear a further overreach in this particular area.

Obviously the copyright issues will need to be figured out, but (and this is where some people will vehemently disagree or even scream at me) I am not convinced that just because an AI model was trained on a given image (as one in billions) there should be any copyright claim by the author of that image on images generated by the model.

Human artists also look at lots of images and then draw new ones based, among other things, on what they have seen before.

People seem to have this idea that AI image generation works by taking some parts of images wholesale and reproducing them with slight changes, but that's just fundamentally wrong. A very easy way to see this is by noting that the average stable diffusion checkpoint has something like 2 GB of data, while being trained on hundreds of terabytes of images. What I want to illustrate by that is that the impact of any individual image (or even 100s of images) on a given output image is minuscule -- similar to or even less than what could easily happen if an artist had looked at some specific images before working on something new.
I'm completely against AI art, artists are already very underappreciated, if AI art was accepted and wasn't receiving hate, it would just result in taking away work opportunities from actual artists.

Also human artists basing themselves on things they've seen before is in no way similar to AI art being based on existing art, there's actual human input in the process of creating art.
 
Last edited:

Durante

I <3 Pixels
Oct 21, 2018
4,082
19,691
113
Well, I disagree. I think the situation is similar to portrait artists reacting to photography. Drawing portraits was apparently a huge income generator for probably the majority of artists until the advent of photography, which largely supplanted it outside of niche use cases.

In fact, if we were to consider "taking away work opportunities" as a reason for preventing new technology, then we would have to stop basically everything. Automated production obviously takes away work opportunities in production, automated transport takes away work opportunities in transportation, and large language models will take away work opportunities in office jobs.

In my opinion, reducing the amount of work that has to be done by humans is a good thing. Work is not inherently an opportunity, it is a burden. The problem with these things is not technology reducing the amount of work that needs to be done manually. It's that capitalist economic models completely fail at spreading these benefits to the whole of society.

(Also, I purposefully speak of "AI image generation", not "AI art", since the discussion of what constitutes "art" is a whole different can of worms [which, funnily enough, I think was also discussed for decades with photography], and I don't really have a strong or educated opinion on it)
 

inky

🤔
Apr 17, 2019
999
2,797
93
I already bought Avengers which I don't really plan on playing (I hated it on game pass and straight up quit) and the sale hasn't even started...

I fear for my poor decision making this upcoming sale, but there isn't really anything i want to play right now.

Just impatiently waiting on Starfield
 

QFNS

Plays too many card games
Nov 18, 2018
1,308
3,153
113
My number one gripe with "AI" is that it is not Artificial Intelligence. These art generation tools are large machine learning models, and the ChatGPT like ones are similar but use language models instead. There's no intelligence in them.

They are extremely generative and can produce novel output, but that's not the same thing. The machine has no thought. It isn't even close to capable of the kind of interiority of a human mind and it's capacity is limited on only to what it has seen before. You can't ask a learning model to do some function completely outside of it's corpus and get any type of result that makes sense. Where as a human can reason and come up with an answer a model cannot.

Words matter, and using AI for this type of model is silly and demeaning to intelligence overall IMO. But alas techbros have pushed it too hard now to stop.
 

Kyougar

No reviews, no Buy
Nov 2, 2018
3,315
11,818
113
I'm completely against AI art, artists are already very underappreciated, if AI art was accepted and wasn't receiving hate, it would just result in taking away work opportunities from actual artists.

Also human artists basing themselves on things they've seen before is in no way similar to AI art being based on existing art, there's actual human input in the process of creating art.
Human input is also required to make AI art.
At the lowest level you need an "input engineer" or whatever they are called. Basically the person who defines what words or sentences to use, to get a workable image out of it
And if you don't just want to use that image, you also need a real artist to finish the image to the right specifications or because there were errors in the image or something didn't fit right.

And personally, I don't see the ARTISTIC CREATION difference between AI art and someone heavily using Photoshop and CG imaging. The tools to create images are insane, and in the end, what comes out of it, is just AI art with more steps and more hours put into it.

The art that is hand drawn (or created with mechanical tools and real objects) is real art for me. But that's also just a tiny fraction of imaging art out there.

Anything software-driven has to compete with better tools and the better tools are currently AI-generated art which is created and finished by human artists.
We (mostly) don't code software one line after another, tools are either helping us, or create the code completely, and the human has to make sure everything works together and fix bugs. If we never would have made that step to optimize coding, so many more programmers could have work right now. But that argument would be silly.
 

ezodagrom

JELLYBEE
Nov 2, 2018
1,888
4,936
113
Portugal
www.youtube.com
Well, I disagree. I think the situation is similar to portrait artists reacting to photography. Drawing portraits was apparently a huge income generator for probably the majority of artists until the advent of photography, which largely supplanted it outside of niche use cases.

In fact, if we were to consider "taking away work opportunities" as a reason for preventing new technology, then we would have to stop basically everything. Automated production obviously takes away work opportunities in production, automated transport takes away work opportunities in transportation, and large language models will take away work opportunities in office jobs.

In my opinion, reducing the amount of work that has to be done by humans is a good thing. Work is not inherently an opportunity, it is a burden. The problem with these things is not technology reducing the amount of work that needs to be done manually. It's that capitalist economic models completely fail at spreading these benefits to the whole of society.

(Also, I purposefully speak of "AI image generation", not "AI art", since the discussion of what constitutes "art" is a whole different can of worms [which, funnily enough, I think was also discussed for decades with photography], and I don't really have a strong or educated opinion on it)
It's one thing for new technologies to take away work opportunities when it comes to objective tasks, repetitive tasks, menial tasks, but it's completely different when it comes to subjective things like art, music, and so on, things where the end result is special because of the human input, creativity, inspiration.

This isn't a similar situation to photography replacing portraits. Before the invention of photography, alot of use cases for portraits were to attempt to replicate reality into a picture (something objective), but since until photography this wasn't possible, portraits were the next best thing.
Photography made that use case a reality, but portraits still maintained their subjective niche from an art point of view.
Plus professional photography still requires skill/human input.

What AI image generation is primarily being used for currently (AI art), it's attempting to replace a subjective work, replace artists.
I'm not against tools that reduce the amount of work that an artist has to do, as long as these tools don't completely replace the artist, the human input, and the subjective nature of creating art, and I'm not against AI image generation if it's used for something objective, but I'm completely against its current primary use of replacing artists.

You were talking about AI image generation and not AI art, but the topic before was about AI art, the hate is towards AI art.
 

Arc

MetaMember
Sep 19, 2020
3,028
11,428
113
I know they're mostly doing it to dodge Starfield, but this made my day. Now I'll have time to play both games when they release.
 

Dandy

Bad at Games.
Apr 17, 2019
1,664
4,006
113
Fuck yes! That's very smart of them. Also, now I don't have to choose!

They also posted a big update with all of the classes, subclasses, and races plus the "new" party member, spells, feats, etc.

 

lashman

Dead & Forgotten
Sep 5, 2018
32,404
91,290
113
Hope the game is still in a good state at launch.
i'm sure it will be, they were most likely just holding back the PC version and polishing the PS5 one (which is why the latter is delayed by another week)
 
  • This!
Reactions: Dandy

Avern

MetaMember
May 14, 2020
370
1,239
93
Did this turn out to be any good? I feel like I haven't heard people talk about it anywhere near as much as I did for Nioh.
Not really. Imagine Sekiro, but instead of having a variety of enemy attacks that require different responses, you have one response for everything. Wo Long's deflect beats everything every enemy can do, and it gets stale quickly. There are RPG systems like Nioh (though simpler), but you have little reason to engage with them since learning deflection is a first-order optimal strat.

PC performance was terrible too.
 
  • Sad
Reactions: toxicitizen

FeedMeAStrayCat

When you see me again, it won't be me.
Sep 19, 2018
390
1,408
93
Pennsylvania
Has someone bit the bullet on AEW yet? i want to know if the port is good!
Impressions have been mostly positive that I've seen, negatives mainly comes down to lack of options in Create A Wrestler mode and some overall jank. I really want to pick it up, but alas, I'm fully sucked into Diablo 4 at the moment and I don't see that changing anytime soon. Plus I practically never buy full price games, and well D4 I even paid for the early access shenanigans :wd_happy:
 
  • Like
Reactions: NarohDethan

NarohDethan

There was a fish in the percolator!
Apr 6, 2019
9,210
25,751
113
Impressions have been mostly positive that I've seen, negatives mainly comes down to lack of options in Create A Wrestler mode and some overall jank. I really want to pick it up, but alas, I'm fully sucked into Diablo 4 at the moment and I don't see that changing anytime soon. Plus I practically never buy full price games, and well D4 I even paid for the early access shenanigans :wd_happy:
I’m tempted since it has somewhat decent regional pricing here. But my current setup makes it a bit of a pain to use my computer
 
  • Like
Reactions: FeedMeAStrayCat

Durante

I <3 Pixels
Oct 21, 2018
4,082
19,691
113
My number one gripe with "AI" is that it is not Artificial Intelligence. These art generation tools are large machine learning models, and the ChatGPT like ones are similar but use language models instead. There's no intelligence in them.

They are extremely generative and can produce novel output, but that's not the same thing. The machine has no thought. It isn't even close to capable of the kind of interiority of a human mind and it's capacity is limited on only to what it has seen before. You can't ask a learning model to do some function completely outside of it's corpus and get any type of result that makes sense. Where as a human can reason and come up with an answer a model cannot.

Words matter, and using AI for this type of model is silly and demeaning to intelligence overall IMO. But alas techbros have pushed it too hard now to stop.
I mean, on the one hand I have made this exact argument (and had the same annoyance) about the preponderance of the word "AI" before, and its use for ML.

On the other hand, these days I feel like we increasingly have a bit of a "god in the cracks" situation going on, but for what we consider to constitute "intelligence". People used to think of chess playing ability as an indicator of "intelligence", just up until it turned out that even postively prehistoric (by today's standards) computers could outplay the best human.
Until recently, being able to write a coherent argument given a short prompt, or passing 70%+ of the University bachelor-level exams across all subjects would have seemed "intelligent", but now that LLMs can do that we scoff at that idea.

What is our real metric here?

AI ‘art’ sucks shit and if crypto taught us anything is that letting tech bros dictate new trends is a bad idea
I don't really like that many people put generative AI in the same category as crypto bullshit, and I don't think it's appropriate.
I guess it's because both are things that aren't rendering which are done with GPU, and there might be some overlap in the people who are interested. But those are just incidental similarities.

However, there's a fundamental, massive difference:
Blockchain was a technology in search of a problem from the start, and most things built on top of it (e.g. NFTs) just went further into that direction.
On the other hand, generative AI is actually incredibly useful across a broad swathe of real application areas.
 

ezodagrom

JELLYBEE
Nov 2, 2018
1,888
4,936
113
Portugal
www.youtube.com
Final Fantasy Pixel Remasters:
Incidentally, we’ve talked a lot about the home console release, but will any of these new features and changes be added to the mobile and PC versions?

Naofumi:
I’m sure that people who played the mobile and PC versions first will be surprised that there’s this kind of definitive version coming out. It’s difficult to say for certain at the moment, but of course I’m considering it. Or perhaps I should say that I’ve already started making arrangements.

That being said, making the home console versions was a completely separate process from making the mobile and PC versions, so I don’t know if it will be possible to include every single new feature and change at this point in time.

Mobile devices in particular have different specs that vary from model to model, so I think we will need to do a lot of very in-depth testing on lots of different devices before we can figure out if we will be able to include everything from this release. And another thing to consider is that the background music feature uses an entirely separate additional soundtrack, which takes up memory. For low-spec smartphones, the games are already pushing the limits of the memory while running, so it might be tricky to add new features. On the other hand, the boost features, for instance, just mess with existing numerical values, so they should be relatively easy to implement.

I’m sorry that I can’t say anything more definitive than that. At the moment, all I can say is that I’m working on it, and I’ll do the best that I can. I should be able to give a more detailed answer in the future, so please bear with me for now.
 
Status
Not open for further replies.