AI/LLMs DLSS 5 will be the worst thing to EVER happen to video games


log in or register to remove this ad

Ohhhh. This isn’t a good faith debate. Sorry, I get it now.

Bye.
Mod Note:

If you think someone isn’t debating in good faith, you can choose to disengage, report, ignore, etc., and everything will be fine.

A public accusation of lack of good faith, OTOH, is only going to escalate tensions in the discussion. So let’s not. Thanks!
 


After all, people weren't stuffposting about it en masse with yassified edits because they thought it looked great and merely enhanced the natural features of the characters in question.

Mod Edit: language
Always wondered if that term was forbidden: as an inveterate, er, poster of stuff, I was curious. Better you than me!
 

well a content creator debunked some of the issues.


content creator WhizzDumbPlayz has just released a video analysis showing that most, if not all, of these visual issues were already present in the original games.

In the main Starfield comparison from the beginning of the game, featuring a dialogue scene with Heller and Supervisor Lin at the Argos Extractors Mining Outpost, commenters have pointed out that part of Heller's nostril has grown and that he has some hair where it shouldn't be. The content creator counters that there's strange geometric meshing and lighting interactions even in the original, likely due to visual bugs caused by Bethesda's Creation Engine, which make it look like hair is appearing and disappearing,

Taken from A YouTuber Just Debunked DLSS 5's Geometric Anomalies, Showing That They Were Already in the Games

Not that folks will care all that much anyway.
 

In the main Starfield comparison from the beginning of the game, featuring a dialogue scene with Heller and Supervisor Lin at the Argos Extractors Mining Outpost, commenters have pointed out that part of Heller's nostril has grown and that he has some hair where it shouldn't be. The content creator counters that there's strange geometric meshing and lighting interactions even in the original, likely due to visual bugs caused by Bethesda's Creation Engine, which make it look like hair is appearing and disappearing,
I'd hardly really call this a debunk. If these issues happened just by themselves, folks would have posted them all over the place. Its not like people haven't dragged Starfield enough, its not a popular game and people love to dunk on it for any reason they can. If this was a raw issue with Starfield itself? Folks absolutely would have posted about it elsewhere. But, they haven't and its only come up with the DLSS 5 stuff.

I get blaming Creation Engine and Todd Howard is in vogue, but its abundantly clear the tech behind DLSS 5 makes issues out of what's basically nothing in the original game. This isn't debunking anything, its just trying to shift the blame to the original source. And frankly, I disagree entirely with that.
 

well a content creator debunked some of the issues.


content creator WhizzDumbPlayz has just released a video analysis showing that most, if not all, of these visual issues were already present in the original games.

In the main Starfield comparison from the beginning of the game, featuring a dialogue scene with Heller and Supervisor Lin at the Argos Extractors Mining Outpost, commenters have pointed out that part of Heller's nostril has grown and that he has some hair where it shouldn't be. The content creator counters that there's strange geometric meshing and lighting interactions even in the original, likely due to visual bugs caused by Bethesda's Creation Engine, which make it look like hair is appearing and disappearing,

Taken from A YouTuber Just Debunked DLSS 5's Geometric Anomalies, Showing That They Were Already in the Games

Not that folks will care all that much anyway.
To be fair, Nividia should have better vetted their initial images for weird stuff that their AI tech would inevitably be blamed for.

But yea, assuming these are legit images then most if not all the weird crap wasn’t being caused by DLSS5.
 

I'd hardly really call this a debunk. If these issues happened just by themselves, folks would have posted them all over the place. Its not like people haven't dragged Starfield enough, its not a popular game and people love to dunk on it for any reason they can. If this was a raw issue with Starfield itself? Folks absolutely would have posted about it elsewhere. But, they haven't and its only come up with the DLSS 5 stuff.

I get blaming Creation Engine and Todd Howard is in vogue, but its abundantly clear the tech behind DLSS 5 makes issues out of what's basically nothing in the original game. This isn't debunking anything, its just trying to shift the blame to the original source. And frankly, I disagree entirely with that.
I can only agree. The entire video is either based on a faulty understanding of the tech or is deliberately duplicitous in its explanation. It seems a lot closer to the latter. The framing of this "debunking" as DLSS5 not altering existing geometry is a smokescreen and nothing else. This isn't a great start because it does seem like a deliberate attempt to miss the point. Normally I like to give people the benefit of the doubt and I was initially more generous but rewrote this section upon finishing the whole thing. You'll see why. Back to the point though this is just "um ackshually" levels of argumentation that elevates pedantry over the point being made by people who might not have the vocabulary to make that point in technically precise ways. This only serves to avoid the actual issue in favour of an easy target.

DLSS5 doesn't alter the geometry because it's single-frame image generation. It just spits out a new frame based on a single traditionally rendered frame and presents that to you. That doesn't mean it can't spit out frames that appear to have geometric changes. Everyone can see that these things look different. That is, in fact, the selling point of the thing. It's these difference that people dislike. People being unable to express the problems they're seeing in ways that are sufficiently accurate for him doesn't mean they're wrong disliking it. Focusing in on the geometry thing also conveniently ignores all the other problems in how it's changing skin tones, lighting, and the like. And, yes, it's not actually "changing" the raw texture files in the game so no textures are being "changed" but it is changing what's being displayed and in casual conversation I think everyone does actually understand what is meant when people say things like "it changed the colour of her lips".

Unfortunately it goes beyond the useless pedantry. The guy says that DLSS5 isn't making problems that don't already exist, yet he does say DLSS5 makes these things more visible. So it is, in fact, causing problems because making something worse is a problem. It is a problem with DLSS5 under any rational view of it. Making a thing worse makes it worse and I wish I didn't have to say a sentence that obviously true. I genuinely do not understand why people bend over backwards like this when it's an explanation that doesn't hold up to even seconds of consideration. If your dog knocks over the trash and makes a mess of the kitchen you don't get to spread the trash around your house and then blame the dog for the whole mess.

He also, more than once, mentions things that are being talked about as flaws but skips over them entirely. He doesn't do any comparisons for haloing seen in some of the comparisons because it's not something he can explain away. One of the most dramatic issues with DLSS5 that we've seen is how radically it alters the lighting and shadows in a scene and he similarly doesn't address this either. Both of those are things he says that Nvidia could solve by just fixing the problem. "If Nvidia can implement some way to track the source lighting in game so that DLSS5 only calculates the lighting effects from those directions, then that could solve the issue entirely." Yes. Obviously if Nvidia can make this problem vanish then it won't be there. But the problem is there and it's demonstrably something introduced by DLSS5 and he doesn't address this in any substantive fashion. It shows an incredible level of bias on his part and should be a massive red flag for everything else he's talking about. DLSS5 also doesn't function in a way that makes that possible. Which he should well know. He explains how it works in this very video and claims to have watched Daniel Owen's video on the subject which has confirmation from Nvidia that the input DLSS5 gets is a 2D frame and motion vectors. All DLSS5 knows is one frame and the direction of movement things within the frame have. He must understand that this makes any complex light source tracking impossible (and is indeed what raytracing is for) yet he handwaves this away. It's one of the most glaring issues with the tech but he opts to avoid comparisons, and basically says don't worry about it because when it's fixed it'll be okay but he understands that it doesn't work like that and cannot be fixed like he suggests. It sees the frame and motion vectors. It does not see all the light sources that will be illuminating things in that frame. He knows that and still decides to handwave the whole thing.

He has a section explaining that this isn't image generation because that would simply be impossible. According to his "logic" the time it takes for an online image generator to make a single image is much longer than the time it takes to display a single frame. Yet if he understood how this tech and the generative AI he's talking about actually function he'd know that is a deeply misleading statement that conflate very different methods of image generation. But he does actually know that because he explains how it works he just goes on to tell you that this isn't generating images despite objectively generating images. Nvidia will tell you that but he has to claim otherwise to defend it and it's simply a lie. Worse, he knows he's lying. He shows an image with a quote from Jensen Huang saying it involves generative AI. He shows a second image from the same page from Nvidia's site saying it's generating images. Claiming otherwise is closer to corporate propaganda than it is to debunking at this stage and even the corporation isn't lying about it being image generation. No one should be taking anything this person is saying as evidence of anything. He's willing to lie about the basic foundation of the tech to make it sound more palatable. Everyone needs to find a second source for anything being claimed in this video.

Finally, he also claims that the Oblivion demo has no blurring or smearing, which I have demonstrated it does in this thread, but it's also contradictory to prior parts of the video where he does mention that those sorts of artefacts are present in these example. He explains away the smearing in other sections as being from frame generation but even if that was the case he still can't say it's not present nor can he actually say what it's caused by because of the lack of comparisons made between DLSS5 with and without frame gen. But instead he just proclaims it's simply not there. Because any problem he can't argue about isn't a problem at all.

This whole thing shows an incredible lack of integrity and I think even if you love this tech you can do better than this. Please find someone not willing to lie about it like this. Find the people willing to tell you that it's image generation and that they just like the look of it. No one benefits from this sort of content.
 

I can only agree. The entire video is either based on a faulty understanding of the tech or is deliberately duplicitous in its explanation. It seems a lot closer to the latter. The framing of this "debunking" as DLSS5 not altering existing geometry is a smokescreen and nothing else. This isn't a great start because it does seem like a deliberate attempt to miss the point. Normally I like to give people the benefit of the doubt and I was initially more generous but rewrote this section upon finishing the whole thing. You'll see why. Back to the point though this is just "um ackshually" levels of argumentation that elevates pedantry over the point being made by people who might not have the vocabulary to make that point in technically precise ways. This only serves to avoid the actual issue in favour of an easy target.

DLSS5 doesn't alter the geometry because it's single-frame image generation. It just spits out a new frame based on a single traditionally rendered frame and presents that to you. That doesn't mean it can't spit out frames that appear to have geometric changes. Everyone can see that these things look different. That is, in fact, the selling point of the thing. It's these difference that people dislike. People being unable to express the problems they're seeing in ways that are sufficiently accurate for him doesn't mean they're wrong disliking it. Focusing in on the geometry thing also conveniently ignores all the other problems in how it's changing skin tones, lighting, and the like. And, yes, it's not actually "changing" the raw texture files in the game so no textures are being "changed" but it is changing what's being displayed and in casual conversation I think everyone does actually understand what is meant when people say things like "it changed the colour of her lips".

Unfortunately it goes beyond the useless pedantry. The guy says that DLSS5 isn't making problems that don't already exist, yet he does say DLSS5 makes these things more visible. So it is, in fact, causing problems because making something worse is a problem. It is a problem with DLSS5 under any rational view of it. Making a thing worse makes it worse and I wish I didn't have to say a sentence that obviously true. I genuinely do not understand why people bend over backwards like this when it's an explanation that doesn't hold up to even seconds of consideration. If your dog knocks over the trash and makes a mess of the kitchen you don't get to spread the trash around your house and then blame the dog for the whole mess.

He also, more than once, mentions things that are being talked about as flaws but skips over them entirely. He doesn't do any comparisons for haloing seen in some of the comparisons because it's not something he can explain away. One of the most dramatic issues with DLSS5 that we've seen is how radically it alters the lighting and shadows in a scene and he similarly doesn't address this either. Both of those are things he says that Nvidia could solve by just fixing the problem. "If Nvidia can implement some way to track the source lighting in game so that DLSS5 only calculates the lighting effects from those directions, then that could solve the issue entirely." Yes. Obviously if Nvidia can make this problem vanish then it won't be there. But the problem is there and it's demonstrably something introduced by DLSS5 and he doesn't address this in any substantive fashion. It shows an incredible level of bias on his part and should be a massive red flag for everything else he's talking about. DLSS5 also doesn't function in a way that makes that possible. Which he should well know. He explains how it works in this very video and claims to have watched Daniel Owen's video on the subject which has confirmation from Nvidia that the input DLSS5 gets is a 2D frame and motion vectors. All DLSS5 knows is one frame and the direction of movement things within the frame have. He must understand that this makes any complex light source tracking impossible (and is indeed what raytracing is for) yet he handwaves this away. It's one of the most glaring issues with the tech but he opts to avoid comparisons, and basically says don't worry about it because when it's fixed it'll be okay but he understands that it doesn't work like that and cannot be fixed like he suggests. It sees the frame and motion vectors. It does not see all the light sources that will be illuminating things in that frame. He knows that and still decides to handwave the whole thing.

He has a section explaining that this isn't image generation because that would simply be impossible. According to his "logic" the time it takes for an online image generator to make a single image is much longer than the time it takes to display a single frame. Yet if he understood how this tech and the generative AI he's talking about actually function he'd know that is a deeply misleading statement that conflate very different methods of image generation. But he does actually know that because he explains how it works he just goes on to tell you that this isn't generating images despite objectively generating images. Nvidia will tell you that but he has to claim otherwise to defend it and it's simply a lie. Worse, he knows he's lying. He shows an image with a quote from Jensen Huang saying it involves generative AI. He shows a second image from the same page from Nvidia's site saying it's generating images. Claiming otherwise is closer to corporate propaganda than it is to debunking at this stage and even the corporation isn't lying about it being image generation. No one should be taking anything this person is saying as evidence of anything. He's willing to lie about the basic foundation of the tech to make it sound more palatable. Everyone needs to find a second source for anything being claimed in this video.

Finally, he also claims that the Oblivion demo has no blurring or smearing, which I have demonstrated it does in this thread, but it's also contradictory to prior parts of the video where he does mention that those sorts of artefacts are present in these example. He explains away the smearing in other sections as being from frame generation but even if that was the case he still can't say it's not present nor can he actually say what it's caused by because of the lack of comparisons made between DLSS5 with and without frame gen. But instead he just proclaims it's simply not there. Because any problem he can't argue about isn't a problem at all.

This whole thing shows an incredible lack of integrity and I think even if you love this tech you can do better than this. Please find someone not willing to lie about it like this. Find the people willing to tell you that it's image generation and that they just like the look of it. No one benefits from this sort of content.
To me the main point, that you seem to have left unaddressed, DLSS5 was turned off and similar oddities still existed. For example, The odd oblivion eyes, they were there even without DLSS5!
 
Last edited:

I can only agree. The entire video is either based on a faulty understanding of the tech or is deliberately duplicitous in its explanation. It seems a lot closer to the latter. The framing of this "debunking" as DLSS5 not altering existing geometry is a smokescreen and nothing else. This isn't a great start because it does seem like a deliberate attempt to miss the point. Normally I like to give people the benefit of the doubt and I was initially more generous but rewrote this section upon finishing the whole thing. You'll see why. Back to the point though this is just "um ackshually" levels of argumentation that elevates pedantry over the point being made by people who might not have the vocabulary to make that point in technically precise ways. This only serves to avoid the actual issue in favour of an easy target.

DLSS5 doesn't alter the geometry because it's single-frame image generation. It just spits out a new frame based on a single traditionally rendered frame and presents that to you. That doesn't mean it can't spit out frames that appear to have geometric changes. Everyone can see that these things look different. That is, in fact, the selling point of the thing. It's these difference that people dislike. People being unable to express the problems they're seeing in ways that are sufficiently accurate for him doesn't mean they're wrong disliking it. Focusing in on the geometry thing also conveniently ignores all the other problems in how it's changing skin tones, lighting, and the like. And, yes, it's not actually "changing" the raw texture files in the game so no textures are being "changed" but it is changing what's being displayed and in casual conversation I think everyone does actually understand what is meant when people say things like "it changed the colour of her lips".

Unfortunately it goes beyond the useless pedantry. The guy says that DLSS5 isn't making problems that don't already exist, yet he does say DLSS5 makes these things more visible. So it is, in fact, causing problems because making something worse is a problem. It is a problem with DLSS5 under any rational view of it. Making a thing worse makes it worse and I wish I didn't have to say a sentence that obviously true. I genuinely do not understand why people bend over backwards like this when it's an explanation that doesn't hold up to even seconds of consideration. If your dog knocks over the trash and makes a mess of the kitchen you don't get to spread the trash around your house and then blame the dog for the whole mess.

He also, more than once, mentions things that are being talked about as flaws but skips over them entirely. He doesn't do any comparisons for haloing seen in some of the comparisons because it's not something he can explain away. One of the most dramatic issues with DLSS5 that we've seen is how radically it alters the lighting and shadows in a scene and he similarly doesn't address this either. Both of those are things he says that Nvidia could solve by just fixing the problem. "If Nvidia can implement some way to track the source lighting in game so that DLSS5 only calculates the lighting effects from those directions, then that could solve the issue entirely." Yes. Obviously if Nvidia can make this problem vanish then it won't be there. But the problem is there and it's demonstrably something introduced by DLSS5 and he doesn't address this in any substantive fashion. It shows an incredible level of bias on his part and should be a massive red flag for everything else he's talking about. DLSS5 also doesn't function in a way that makes that possible. Which he should well know. He explains how it works in this very video and claims to have watched Daniel Owen's video on the subject which has confirmation from Nvidia that the input DLSS5 gets is a 2D frame and motion vectors. All DLSS5 knows is one frame and the direction of movement things within the frame have. He must understand that this makes any complex light source tracking impossible (and is indeed what raytracing is for) yet he handwaves this away. It's one of the most glaring issues with the tech but he opts to avoid comparisons, and basically says don't worry about it because when it's fixed it'll be okay but he understands that it doesn't work like that and cannot be fixed like he suggests. It sees the frame and motion vectors. It does not see all the light sources that will be illuminating things in that frame. He knows that and still decides to handwave the whole thing.

He has a section explaining that this isn't image generation because that would simply be impossible. According to his "logic" the time it takes for an online image generator to make a single image is much longer than the time it takes to display a single frame. Yet if he understood how this tech and the generative AI he's talking about actually function he'd know that is a deeply misleading statement that conflate very different methods of image generation. But he does actually know that because he explains how it works he just goes on to tell you that this isn't generating images despite objectively generating images. Nvidia will tell you that but he has to claim otherwise to defend it and it's simply a lie. Worse, he knows he's lying. He shows an image with a quote from Jensen Huang saying it involves generative AI. He shows a second image from the same page from Nvidia's site saying it's generating images. Claiming otherwise is closer to corporate propaganda than it is to debunking at this stage and even the corporation isn't lying about it being image generation. No one should be taking anything this person is saying as evidence of anything. He's willing to lie about the basic foundation of the tech to make it sound more palatable. Everyone needs to find a second source for anything being claimed in this video.

Finally, he also claims that the Oblivion demo has no blurring or smearing, which I have demonstrated it does in this thread, but it's also contradictory to prior parts of the video where he does mention that those sorts of artefacts are present in these example. He explains away the smearing in other sections as being from frame generation but even if that was the case he still can't say it's not present nor can he actually say what it's caused by because of the lack of comparisons made between DLSS5 with and without frame gen. But instead he just proclaims it's simply not there. Because any problem he can't argue about isn't a problem at all.

This whole thing shows an incredible lack of integrity and I think even if you love this tech you can do better than this. Please find someone not willing to lie about it like this. Find the people willing to tell you that it's image generation and that they just like the look of it. No one benefits from this sort of content.

I didn't watch the video and honestly haven't really paid much attention, other than to chuckle at the hysteria.

But I'm curious about this from a technical perspective. I think what you are describing is this:
1. First, the original mesh of triangles as written by the programmers is still used to model the scene in 3D, in memory, along with all the texture mapping and ray-tracing and whatever other techniques are used today to determine the color of each pixel.
2. The scene is rotated so that the viewers P.O.V. is oriented along the z-axis (maybe this is done earlier; I don't really know) and everything is "collapsed" into 2D
3. DLSS 5 is now used on the resulting 2D image...I'm guessing using the some of the same "adjacent pixel" logic as generative AI.

Thus no "changes to the geometry" in the sense of altering the triangle mesh, but possibly some re-touching that looks as if it could have resulted from geometry changes.

Is that correct?
 

Recent & Upcoming Releases

Remove ads

Top