as far as I understand, AI used in dlss is a form of generative AI... either way, this is a very dumb discussion because generative AI is fully integrated in Adobe tools nowadays. Anyone working in any creative field is using AI in some form and, as long as you dont remove human input and use it as is, its totally fine.
I wish they rejected this too, this is disgusting. This is not an advancement in technology but a fancy filter. It's fine when I add filters to my game on my pc. But when you integrate filters into the code of the game, IT'S NOT!!
@@JardsonJean Usage of AI in real time to generate stuff on the go is so far the one place where the recent generative AI stuff opens up previously impossible options. You cannot manually upscale every single possible frame in a game without just directly outputting it in a higher quality which is often not possible, and as such this is one of those cases where AI is the only option without any real down sides.
@@JardsonJeanFrom my understanding, I think Nintendo is only doing this to slightly make things easier, not having it do all the work. Like I said, Nintendo said they won't use generative AI like other companies have been using it, which the others just solely want to have it do all the work. Correct me if I'm wrong, but that's what I remember at least.
@IDKisReal2401 its still built in the same principles and probably also pre trained with stolen data, which is unethical. If thats the problem, then it should be looked down upon too. My point being, big companies dont care. They will use it regardless.
The term AI is just inherently useless because there are plenty of different ways to interpret intelligence. Imo it worked best when describing NPCs in videogames because it's literally trying to emulate human intelligence in some sense.
As PC gamer my issue with AI upscaling is that game studios quickly figured out people turn it on anyway so they now use it as a crutch to get lazy with optimization
But it’s a dedicated system with a consistent performance, so it’ll be optimized to switch specs regardless. If it runs well and looks good I don’t see the issue.
On PC it’s a definite bad thing. I can’t even break 80fps on Marvel Rivals without it but I think for the Switch it’ll be a net positive overall since it’s a locked down ecosystem
10000%. They use “AI” as a marketing term when it’s the same “AI” that’s been used in games for decades… like the “AI” that’s used for enemies in games like halo for instance… it’s not actually artificial intelligence at all
Is it marketing actually? Because last time I've checked the almost antibody response some people have comes from poor saps being exposed too much to mad twitter artists' tears
@@_.-. It is. I wasn't even speaking on AI's 'art generation'. Which is basically just a search engine that looks to the internet to generate an image/video response rather than a text response. Not all that impressive, imo.
I mean it's better then the game being unable to do 16:9 because the gamecube nver was able to render the game that way... For some reason like the NES doesn't care or the N64 they just work... Although when youn add moder anti alsing and bump up the ingame rendeer rez it take it rether well but nope not widscreen for some reason...
@@GreenBlueWalkthrough Dolphin can run many games on the cube in true 16/9. Nintendo could also do so with an emulator. They could potentially cap the res at 720P and upscale from it even up to 4K as most gamecube games would look great and run well at that. NES and 2d games in general didn't support any kind of widescreen till like 5 years ago or something like that. It was a huge deal and took dedicated fans many years of work to get working. The same has been done with Dolphin and most 3d Emulators seem to be easier to tell to run in 16/9.
I mean autotune is industry standard now... You don't notice it because auto engeeers actually know how to use it now... Just like snyths amd digital formats... The same exact thing with AI... Which if you want to see a licve profiormance you see a live performance or listen to actul lve performance not one made for marketing or something stupid big bands do now...
AI is what allows stuff like NPCs to function? I feel like people are failing to understand that most games utilize artificial intelligence at a fundamental level.
If you automatically assume AI is bad then it’s time to do some research. This is not a bad use of AI. If they were using AI to generate art in games then we could have a problem there
@ well yeah but thats what I’m saying. Understand the difference. Because there will be people that see AI and Nintendo and assume they’re just using chat gpt to generate Mario kart tracks lol
NPC behaviour in games has historically been called "AI", but it's not the same kind of AI that people generally talk about today. Like, if I said "I'm fighting against the computer" in a fighting game, you'd probably understand they haven't installed like, a tiny actual computer inside your games console. In the same way "computer" is a catch all term for programed behaviour, talking about npcs being "AI" was just another way of saying "programmed". Or if I "call in for support troops" in a strategy game, I'm not having to like, pull out a phone and dial into the game. It's just convenient to use that word to describe it. In short, most games don't use "AI" as its generally discussed today (machine learning, generative ai, etc) to control stuff in their games.
in game dev enemy AI is usually referred to as enemy behaviours or scripting, AI can be used shorthand but the problem with the word AI as it's used today is it has too many meanings. And funnily enough we've not even really come close to the true meaning of the original term of AI, as it's often depicted in Sci Fi, as a man made creation that has intelligence that rivals or exceeds it's creators intelligence
Understand AI upscaling is entirely about letting the actual computer do the dirty work to make the game higher resolution. Generative AI is flatting out "Attempt" at stealing the work humans actually want to do.
If this works for current/previous Switch games then I’m all for it. Would love to see games like Odyssey or Breath of the Wild getting the Ai upscale.
@@_.-. I cant tell if you are directing that at artists or the people with lack of skill who use generative AI. If you are actually talking about artist…that’s loser shit.
Oh my goodness, as a pc gamer its weird to see people worried. DLSS 3.0 is simply amazing. You'll love it everytime it is done well and its easy for devs to use. Also, as a pc gamer, know we are approaching the capacity for power on the current tech for chips and nodes. A brand new type will be needed to see any serious improvement, unless you use AI. See the rtx 4000 and 5000 series as examples as well as PS4 vs PS5
DLSS is much better than FSR, so we're in for a treat on Switch 2. Also, they _just_ updated the DLSS AI model from a "CNN" based model to a "Transformer" model similar to what is used in ChatGPT, and it further increases the visual stability of the image and fixes most "smearing/ghosting" issues. However, I dunno if the Switch 2 can handle the more power intensive model, so they may stick to the CNN model on it early on. Which tbh still looks excellent, and is still currently the best upscaler we have until the new transformer model arrives. Nintendo may opt for their own AI upscaler though.
@@TonyDaExpert Well yes no one should expect AI frame gen on the ampere architecture. But the new DLSS transformer model works on 20 series GPUs, so it's possible.
its actually the same kind of architecture as things like ChatGPT - the transformer architecture. Nvidia announced they were moving from CNN to Transformers at CES and it applies to all RTX. Really two sides of the same coin - generating pixels based on context vs generating text based on context.
In fairness, that's because they used AMD's FSR for the Baldur's Gate footage. FSR does tend to have more of a "smoothing filter" kinda look to it whereas DLSS is usually much closer to looking like native resolution.
FSR is actually less developed than dlss so the switch 2 having the superior dlss is a major step up for performance. I wonder if they are using a derivative of dlss3 or if they figured a way to get the new 4.0 into it. Either way the results are probably gonna impress people. DLSS has impressed me on pc.
Either my phoen screen is only 720p or 1080p is not a big step up because there wasnt anything in that Baldur's Gare example that was "really impressive." Like the rocks and grass are just slgihtly less blurry
Upscaling AI is pretty bad because it makes the games give that plastic looks, gives more glitches than without AI and giving people the boot in the graphics department. Just look at the PS5 Pro. Some of the games are not well optimized.
fsr/dlss is like watching a RUclips video that's stuttering on one of those TV's that has the artificial frame smoothing and blurring on. looks better, but nothing I'd call amazing.
I hope Nintendo doesn't fear the negative annotations with AI thanks to the poor use by some really lazy companies who were looking for easy money. Used right, AI could make the gaming experience skyrocket to new heights. Though the most I can think of, of what is already a norm, is NPC/CPU interactions/control.
Should be noted that DLSS predates this generative AI craze - various forms of AI have been around much, much longer than most people would expect. What most people are calling "AI" today are a couple of very specific forms of AI. In fact, as gamers we ought to remember that we've had various forms of AI in our games for a very long time - often used in single player games for NPCs and to battle against.
So apparently some kind of autocorrect doesn't like what I have to say on this subject. Personally, I think influencers will make a huge deal of this. I worry.
I'm not a fan of having detail that isn't there to begin with. Hopefully it won't be crap or you can even disable it. I prefer low resolution over a baseline filter.
@@ask343 It doesn't look EXACTLY like that. The examples here aren't bad and we can already see this in action in Xenoblade DE, so hopefully my scepticism is for nothing, but still, I hate how some TVs can't let you disable they crappy image smoothering.
@@JagGentlemann idk what kind of TVs you've seen which can't disable image smoothing, also we haven't seen it in Xenoblade DE that uses different upscaling technique. DLSS is MUCH better than all the other upscalers out there, only one I've seen close is intel XeSs
AI upscaling is a totally fine thing to use for most purposes (save for those that push the boundary of generative uses and ESPECIALLY any official documental or evidentiary use, because miss me with your "zoom and enhance" fantasies). It is a careful line to walk, though.
If you want a good example of upscaling AI gone wrong, Immortals of Aveum is it. Great game, tones of graphic issues because of the incompatibility between the engine and the upscaling process.
Other forms of AI, some are OK. The main issue is with generative AI, as it literally steals work, it also copies the lighting, that's why so many images are oversaturated and makes it a big giveaway, a human never made that. Some platforms are going as far to remove a persons copyright, hence why a huge amount of people have left X.
Is this more like how the Wii U could detect an HDMI or AV port and it could automatically adjust the resolution to your television as well? I don't know. To me, the examples you used, admittedly not DLSRR, look marginally better, before proceeding to show the camera pan, and I'd take the worser image quality if I can still see as I adjust the camera. That's a big downgrade to me. At least it doesn't seem like it's generative, I suppose, but I would not oppose downloading a base game and downloading a high resolution texture pack like in Indiana Jones and the Golden Circle, where I can just retain the 20GB version if the television I'm playing on or even if the handheld is whatever screen it comes as by default, is not as high-end, and I can download the 60GB version later, especially in that example, the 60GB was native 4K and the 20GB was upscaled. Native seems to be almost always better, so I've heard, but I'm also someone who's never had a 4K television, so I've never tested this myself, despite having an Xbox Series X and PlayStation 5 (and I believe Microsoft and Sony have used upscaling-AI for theirs, but that doesn't mean Nintendo would be beneficial to follow suit). I did not download the high resolution pack texture pack for Indiana Jones, which is apparently not even worth it for Series S even on a 4K television. I'd love to be wrong, but for now, I'm very skeptical, and I hope there's an option to disable it or modulate it like with the light sensor (which, to be fair, I keep on, but I still manually adjust constantly when it overcorrects).
I just realized that I wonder if this is how Nintendo is going to keep the next Nintendo Switch maybe 64GB or not much more internally. That 60GB example may be how they want to incorporate a NS Online library with GameCube games perhaps. I see their reasoning, but I think it would be easier to have a larger internal memory. This is doing everything but to avoid the problem which is whatever mindset they're working with in that regard. At that point, I'd rather cloud stream Doom Eternal (basically a current-gen game with patches that runs on current-gen platforms already playable on Switch) on a good internet connection but have a local option so when cloud streaming, I can play a near-4K quality high-end PC rendering and performance to save on memory, even if it's not that good when the PC is optimized well, but I don't want to spend a lot of money on a high-end PC or the resources to optimize it. I'm not tech-savvy, as a reader may tell.
This is not the good kind of AI, upscaling is fine for basic upscalers on pc. But his Nvidia garbagio is not it chief. I don't trust Nvidia with complex upcaling algorithms. To date, they have yet to make a decent upscaling algorithm that doesn't happen to sell their next graphics card. I'd rather be in charge of that myself than let the console or video game decide that for me.
People have to learn to separate AI, from features that simply help advanced the state of things, and features that will replace humans and take jobs and steal from others
No offense to you Felix, but this is more of a criticism of the industry and marketing in general, but DLSS/FSR upscaling is not AI. AI has just become a marketing tag line to sell things because it's a "sexy" word to sell products, but it is not AI just because it is "trained" with data. It's just algorithms doing very specialized things.
This isn't right, this is absolutely generative AI. The pixels are being generated. It's just the kind of stuff that is used to create pixels to fill pixels that weren't destroyed. This is where it differs. it's filling blank pixels not ones that have been vacated because someone's chosen to delete the ones they don't want. PhotoShop will create pixels if you want to delete and object out an image. This isn't what DLSS is doing. Nothing is ever deleted, it's filling blank spaces that open up as the canvas size is enlarged. What people are mostly concerned about these days is content created that would otherwise be created by a human. This displaces a person from doing this work. What PhotoShop does is fine to me. But the reality is that AI today can actually create some semblance of that entire image that is being edited in full just with very simple commands. There isn't a photo to edit, the image is entirely generated from a few descriptive words. And it'll typically generate different options. So not a single image, but multiple ones allowing the user generating the image to select the one they prefer, or possibly refine the image based on options that were previously generated. What we don't want is an AI animation of Mario riding an elephant when this animation wasn't previously created by a human. It creates a homogeneous example instead of something that requires human crafting. Maybe this is good enough to use in the drafting stages of the process of making video games, but it would be generally worse than what could be generated by creative person.
I'm genuinely curious as to how they could have rephrased this in a way that wouldn't be "rage baiting". They didn't imply that this was generative, and they outright stated in the title that it wasn't an egregious use.
I mean, nvidias DLSS is just an algorithm, no matter if Machine Learning was used to improve it... The algorithm itself is fast and dumb. :) But nvidia would never say that since they are an "AI" company... it's basically snake-oil marketing for gamers.
They used FSR 2.0 I think. AMD's (inferior) answer to DLSS, that is more available and not locked to hardware, but unfortunately hasn't caught up in terms of functionality.
I would argue that it potentially could be a problem. A neural network being used for upscaling isnt simply just "making the image larger". We've had non-AI methods of doing that for years, and if youve ever used an emulator before, you're probably familiar with scaling methods such as linear, bilinear, or the pixel art destroyer XBRZ. The problem is this. Unlike previous methods that sample a smaller image and use that sample as a basis to reconstruct the image at a larger resolution, neural network based upscaling typically uses a more optical approach. That is, an algorithm will attempt to "identify" parts of an image, like grass, trees, humans, clothes, etc, and then will hallucinate fake detail not found in the original image to give the appearance of the upscaled image being "more HD" than the original. The only problem is that many of these neural networks get confused when looking at scenes with strange or unique visual elements, such as HUDs and small, fast moving VFX. There are tricks to get around this, such as having the HUD be natively generated at a high res while upscaling more performance-heavy 3D elements, but the overall point is that oftentimes images that have been AI upscaled can look blurry, have incorrect detail, or look messier than images natively generated at that resolution. We'll have to see how exactly Nintendo implements it, but unless their upscaling software is somehow more advanced than everything else the industry is using, i sure hope there is a way to switch it off.
I'm sure Japansse fans were equally concerned when nintedo off America was getting it/s own Famcom and even made it own games... Like AI are clearly humans andpeople too so I'm tried of people being extremely ractist to them and using us autistics and the standard of why Ais are bad... Like I get it you suck at your job and don't want to replaced as you have no skills at life... But stop being extremely racsist and hatefull about it... Espacly when you use us autsics as the text book example of what AI can do... When you know AIs are nwerotypical like you guys...
Wow Felix, you really baited me with this title. I was fully prepared to give a lecture about different types of AI and its mostly harmless applications and i honestly didn't expect you to single out Generative AI as the single 'bad' application all by yourself. Bravo 👏
![Blox Fruits Dragon Rework Update [Full Stream]](http://i.ytimg.com/vi/EqDAp8udhm0/mqdefault.jpg)
People need to remember that Nintendo outright rejected GENERATIVE AI, not the AI that's been used up to this point.
as far as I understand, AI used in dlss is a form of generative AI... either way, this is a very dumb discussion because generative AI is fully integrated in Adobe tools nowadays. Anyone working in any creative field is using AI in some form and, as long as you dont remove human input and use it as is, its totally fine.
I wish they rejected this too, this is disgusting. This is not an advancement in technology but a fancy filter.
It's fine when I add filters to my game on my pc. But when you integrate filters into the code of the game, IT'S NOT!!
@@JardsonJean Usage of AI in real time to generate stuff on the go is so far the one place where the recent generative AI stuff opens up previously impossible options.
You cannot manually upscale every single possible frame in a game without just directly outputting it in a higher quality which is often not possible, and as such this is one of those cases where AI is the only option without any real down sides.
@@JardsonJeanFrom my understanding, I think Nintendo is only doing this to slightly make things easier, not having it do all the work. Like I said, Nintendo said they won't use generative AI like other companies have been using it, which the others just solely want to have it do all the work.
Correct me if I'm wrong, but that's what I remember at least.
@IDKisReal2401 its still built in the same principles and probably also pre trained with stolen data, which is unethical. If thats the problem, then it should be looked down upon too. My point being, big companies dont care. They will use it regardless.
I hate how corporations have ruined the word AI, which in reality has both good and bad applications.
The term AI is just inherently useless because there are plenty of different ways to interpret intelligence. Imo it worked best when describing NPCs in videogames because it's literally trying to emulate human intelligence in some sense.
"blockchain" and "crypto" as well. Fascinating and useful technology completely ruined by $$$.
As PC gamer my issue with AI upscaling is that game studios quickly figured out people turn it on anyway so they now use it as a crutch to get lazy with optimization
But it’s a dedicated system with a consistent performance, so it’ll be optimized to switch specs regardless. If it runs well and looks good I don’t see the issue.
On PC it’s a definite bad thing. I can’t even break 80fps on Marvel Rivals without it but I think for the Switch it’ll be a net positive overall since it’s a locked down ecosystem
@@mudedmmuh964 Hopefully
@@archonthaaproducer Fair
"AI" is just a buzzword at this point, means almost nothing. Largely, its just software like all before it, marketing is getting out of hand.
10000%. They use “AI” as a marketing term when it’s the same “AI” that’s been used in games for decades… like the “AI” that’s used for enemies in games like halo for instance… it’s not actually artificial intelligence at all
@@V8chumpI mean, that's technically AI too.
Is it marketing actually? Because last time I've checked the almost antibody response some people have comes from poor saps being exposed too much to mad twitter artists' tears
@@_.-. It is. I wasn't even speaking on AI's 'art generation'. Which is basically just a search engine that looks to the internet to generate an image/video response rather than a text response. Not all that impressive, imo.
GenAI is actually fundamentally different because no human made it or can really understands what it does.
This is a good use of AI. It is buggy sometimes but it would be funny if Mario just started crashing out.
"Itsa me, Maaaaaaaaaaaaa-"
(Game crashes)
I mean it's better then the game being unable to do 16:9 because the gamecube nver was able to render the game that way... For some reason like the NES doesn't care or the N64 they just work... Although when youn add moder anti alsing and bump up the ingame rendeer rez it take it rether well but nope not widscreen for some reason...
@@GreenBlueWalkthrough Dolphin can run many games on the cube in true 16/9. Nintendo could also do so with an emulator. They could potentially cap the res at 720P and upscale from it even up to 4K as most gamecube games would look great and run well at that.
NES and 2d games in general didn't support any kind of widescreen till like 5 years ago or something like that. It was a huge deal and took dedicated fans many years of work to get working. The same has been done with Dolphin and most 3d Emulators seem to be easier to tell to run in 16/9.
AI in games is like autotune in music: It should be used like seasoning in the kitchen. If you use it too much, the endproduct might be a mess.
i feel like that's giving too much credit to ai
That's like a simile within another simile lol
I mean autotune is industry standard now... You don't notice it because auto engeeers actually know how to use it now... Just like snyths amd digital formats... The same exact thing with AI... Which if you want to see a licve profiormance you see a live performance or listen to actul lve performance not one made for marketing or something stupid big bands do now...
AI is what allows stuff like NPCs to function? I feel like people are failing to understand that most games utilize artificial intelligence at a fundamental level.
Keep autotune out of music entirely, everything sounded better before it was invented and singers had to actually, y'know, hit the notes.
If you automatically assume AI is bad then it’s time to do some research. This is not a bad use of AI. If they were using AI to generate art in games then we could have a problem there
You're not wrong, but it's hard to blame people for being leery given the absolute flood of dogshit generative AI out there right now.
@ well yeah but thats what I’m saying. Understand the difference. Because there will be people that see AI and Nintendo and assume they’re just using chat gpt to generate Mario kart tracks lol
I'm very reluctant to assume all AI is bad, but generative-AI has made me reluctant to accept anything new, AI or not.
Don’t most games use AI? Otherwise the enemies will just stand there…
NPC behaviour in games has historically been called "AI", but it's not the same kind of AI that people generally talk about today.
Like, if I said "I'm fighting against the computer" in a fighting game, you'd probably understand they haven't installed like, a tiny actual computer inside your games console.
In the same way "computer" is a catch all term for programed behaviour, talking about npcs being "AI" was just another way of saying "programmed".
Or if I "call in for support troops" in a strategy game, I'm not having to like, pull out a phone and dial into the game. It's just convenient to use that word to describe it.
In short, most games don't use "AI" as its generally discussed today (machine learning, generative ai, etc) to control stuff in their games.
in game dev enemy AI is usually referred to as enemy behaviours or scripting, AI can be used shorthand but the problem with the word AI as it's used today is it has too many meanings. And funnily enough we've not even really come close to the true meaning of the original term of AI, as it's often depicted in Sci Fi, as a man made creation that has intelligence that rivals or exceeds it's creators intelligence
oh no it will use the industry standard rendering pipeline
Understand AI upscaling is entirely about letting the actual computer do the dirty work to make the game higher resolution.
Generative AI is flatting out "Attempt" at stealing the work humans actually want to do.
If this works for current/previous Switch games then I’m all for it. Would love to see games like Odyssey or Breath of the Wild getting the Ai upscale.
Thanks for clarifying Felix, and for standing with writers, artists, VA’s and more against generative AI 😌
Oh no, not the golden egg hens having to get an actual job
@@_.-. I cant tell if you are directing that at artists or the people with lack of skill who use generative AI. If you are actually talking about artist…that’s loser shit.
Oh my goodness, as a pc gamer its weird to see people worried. DLSS 3.0 is simply amazing. You'll love it everytime it is done well and its easy for devs to use.
Also, as a pc gamer, know we are approaching the capacity for power on the current tech for chips and nodes. A brand new type will be needed to see any serious improvement, unless you use AI. See the rtx 4000 and 5000 series as examples as well as PS4 vs PS5
This gives me hope for a much higher quality replacement of the Tegra X1 chip used for docking the current Switch
Upscaling AI is early and causes processing issues. We’ll see how that nets out.
DLSS is much better than FSR, so we're in for a treat on Switch 2.
Also, they _just_ updated the DLSS AI model from a "CNN" based model to a "Transformer" model similar to what is used in ChatGPT, and it further increases the visual stability of the image and fixes most "smearing/ghosting" issues. However, I dunno if the Switch 2 can handle the more power intensive model, so they may stick to the CNN model on it early on. Which tbh still looks excellent, and is still currently the best upscaler we have until the new transformer model arrives.
Nintendo may opt for their own AI upscaler though.
Switch 2 will be mostly based on dlss 2 features with some 3.5 features for raytracing support
@@TonyDaExpert Well yes no one should expect AI frame gen on the ampere architecture. But the new DLSS transformer model works on 20 series GPUs, so it's possible.
its actually the same kind of architecture as things like ChatGPT - the transformer architecture. Nvidia announced they were moving from CNN to Transformers at CES and it applies to all RTX. Really two sides of the same coin - generating pixels based on context vs generating text based on context.
It just looks like they added a blur filter over the pixelated footage :(
In fairness, that's because they used AMD's FSR for the Baldur's Gate footage. FSR does tend to have more of a "smoothing filter" kinda look to it whereas DLSS is usually much closer to looking like native resolution.
FSR is actually less developed than dlss so the switch 2 having the superior dlss is a major step up for performance. I wonder if they are using a derivative of dlss3 or if they figured a way to get the new 4.0 into it. Either way the results are probably gonna impress people. DLSS has impressed me on pc.
... Make sure you watch this video in 1080p to better see what they are talking about.
it makes sense to use DLSS on a handheld console to an extent. Especially if they want 4k support.
I hate the knee jerk reactions AI gets. I have my own reservations on it, but people hear the merest utterance of AI and immediately throw a fit.
Yeah especially when Nintendo has outright said they're not using generative AI
Probably because big tech has made them associate AI with generative AI
Either my phoen screen is only 720p or 1080p is not a big step up because there wasnt anything in that Baldur's Gare example that was "really impressive." Like the rocks and grass are just slgihtly less blurry
Upscaling AI is pretty bad because it makes the games give that plastic looks, gives more glitches than without AI and giving people the boot in the graphics department. Just look at the PS5 Pro. Some of the games are not well optimized.
fsr/dlss is like watching a RUclips video that's stuttering on one of those TV's that has the artificial frame smoothing and blurring on. looks better, but nothing I'd call amazing.
Bro its literally DLSS 😂
U think I can still play my physical switch games or only digital 😢
They will of course use DLSS in switch 2. Dlss is fsr superior to other upscalers
It's only okay if I can turn it off.
Nintendo said they had been using AI for years in areas like enemy actions, but they won't use AI for creative work. AI is a broad concept.
I hope Nintendo doesn't fear the negative annotations with AI thanks to the poor use by some really lazy companies who were looking for easy money. Used right, AI could make the gaming experience skyrocket to new heights. Though the most I can think of, of what is already a norm, is NPC/CPU interactions/control.
Should be noted that DLSS predates this generative AI craze - various forms of AI have been around much, much longer than most people would expect. What most people are calling "AI" today are a couple of very specific forms of AI. In fact, as gamers we ought to remember that we've had various forms of AI in our games for a very long time - often used in single player games for NPCs and to battle against.
So apparently some kind of autocorrect doesn't like what I have to say on this subject.
Personally, I think influencers will make a huge deal of this.
I worry.
I'm not a fan of having detail that isn't there to begin with. Hopefully it won't be crap or you can even disable it. I prefer low resolution over a baseline filter.
It looks exactly the same as if the game were running at higher resolution natively, just with some ghosting
Welcome to 2018 Nintendo fans!
@@ask343 It doesn't look EXACTLY like that. The examples here aren't bad and we can already see this in action in Xenoblade DE, so hopefully my scepticism is for nothing, but still, I hate how some TVs can't let you disable they crappy image smoothering.
@@JagGentlemann idk what kind of TVs you've seen which can't disable image smoothing, also we haven't seen it in Xenoblade DE that uses different upscaling technique. DLSS is MUCH better than all the other upscalers out there, only one I've seen close is intel XeSs
I wasn’t impressed by the difference between the images. There is too much hype in this space.
I think using generative AI in the right way isn't necessary a bad thing
I have never actually seen upscaling like this. Very cool!
Native and upscale (as shown in the examples) look the exact same to me. Am I going insane?
I knew about this, I just wanna know if Switch 1 games will actually do this as well.
Grabbing the popcorn to read the comment section discussing everything under the sun that’s generically tagged as AI
AI upscaling is a totally fine thing to use for most purposes (save for those that push the boundary of generative uses and ESPECIALLY any official documental or evidentiary use, because miss me with your "zoom and enhance" fantasies). It is a careful line to walk, though.
Lets hear it for another round of "Switch 2" leaks year 8.
Interesting. So in short, it sounds like the new console will be using the technology to improve the display and performance of video games.
I'm Coming Switch 2!!! Coming Soon ✌🏻😊✌🏻
If you want a good example of upscaling AI gone wrong, Immortals of Aveum is it. Great game, tones of graphic issues because of the incompatibility between the engine and the upscaling process.
Is great that Nintendo is using AI the correct way, unlike other companies.
nintendo fans discovering what dlss is 💀
These guys are still getting excited over Wii U re-releases…. I don’t know if they can grasp DLSS outside of remembering the acronym
Right? 😂 this is hilarious
I’m a fan and I already know what dlss is.
Can't they just render @ 1080p and use Internet scaling without AI? Don't know how AI could improve things in that scenario.
Other forms of AI, some are OK.
The main issue is with generative AI, as it literally steals work, it also copies the lighting, that's why so many images are oversaturated and makes it a big giveaway, a human never made that.
Some platforms are going as far to remove a persons copyright, hence why a huge amount of people have left X.
Mate the moment the usual suspects left Twitter/X to go circlejerk at bluesky user numbers have only been climbing
AI carumba. As long as Nintendo's own use of the AI doesn't cause weird inconsistencies, we're good. So we're good.
We really clickbaiting hard with this one
and it worked 🤭
Have you noticed how the latest games using this technology all run like shit?
I miss Jon
Is this more like how the Wii U could detect an HDMI or AV port and it could automatically adjust the resolution to your television as well?
I don't know. To me, the examples you used, admittedly not DLSRR, look marginally better, before proceeding to show the camera pan, and I'd take the worser image quality if I can still see as I adjust the camera. That's a big downgrade to me.
At least it doesn't seem like it's generative, I suppose, but I would not oppose downloading a base game and downloading a high resolution texture pack like in Indiana Jones and the Golden Circle, where I can just retain the 20GB version if the television I'm playing on or even if the handheld is whatever screen it comes as by default, is not as high-end, and I can download the 60GB version later, especially in that example, the 60GB was native 4K and the 20GB was upscaled. Native seems to be almost always better, so I've heard, but I'm also someone who's never had a 4K television, so I've never tested this myself, despite having an Xbox Series X and PlayStation 5 (and I believe Microsoft and Sony have used upscaling-AI for theirs, but that doesn't mean Nintendo would be beneficial to follow suit). I did not download the high resolution pack texture pack for Indiana Jones, which is apparently not even worth it for Series S even on a 4K television.
I'd love to be wrong, but for now, I'm very skeptical, and I hope there's an option to disable it or modulate it like with the light sensor (which, to be fair, I keep on, but I still manually adjust constantly when it overcorrects).
I just realized that I wonder if this is how Nintendo is going to keep the next Nintendo Switch maybe 64GB or not much more internally. That 60GB example may be how they want to incorporate a NS Online library with GameCube games perhaps. I see their reasoning, but I think it would be easier to have a larger internal memory. This is doing everything but to avoid the problem which is whatever mindset they're working with in that regard. At that point, I'd rather cloud stream Doom Eternal (basically a current-gen game with patches that runs on current-gen platforms already playable on Switch) on a good internet connection but have a local option so when cloud streaming, I can play a near-4K quality high-end PC rendering and performance to save on memory, even if it's not that good when the PC is optimized well, but I don't want to spend a lot of money on a high-end PC or the resources to optimize it. I'm not tech-savvy, as a reader may tell.
It means the new console will be overflooded with AI generated games.
lol you mean what computers have been using for 70 years
LLMs have really muddied the waters in talking about AI.
This is not the good kind of AI, upscaling is fine for basic upscalers on pc.
But his Nvidia garbagio is not it chief. I don't trust Nvidia with complex upcaling algorithms.
To date, they have yet to make a decent upscaling algorithm that doesn't happen to sell their next graphics card.
I'd rather be in charge of that myself than let the console or video game decide that for me.
DLSS is fine. It's used for optimization. As in, AFTER the game has already been completed.
I dont mind ai if it improves not worsens what it touches.
People have to learn to separate AI, from features that simply help advanced the state of things, and features that will replace humans and take jobs and steal from others
I don't care if it takes the blood of 1000 sacrifices as long as it can hit and stay at 60fps in all games.
No offense to you Felix, but this is more of a criticism of the industry and marketing in general, but DLSS/FSR upscaling is not AI. AI has just become a marketing tag line to sell things because it's a "sexy" word to sell products, but it is not AI just because it is "trained" with data. It's just algorithms doing very specialized things.
AI ≠ Generative AI, that much should be clear by now.
This isn't right, this is absolutely generative AI. The pixels are being generated. It's just the kind of stuff that is used to create pixels to fill pixels that weren't destroyed. This is where it differs. it's filling blank pixels not ones that have been vacated because someone's chosen to delete the ones they don't want. PhotoShop will create pixels if you want to delete and object out an image. This isn't what DLSS is doing. Nothing is ever deleted, it's filling blank spaces that open up as the canvas size is enlarged. What people are mostly concerned about these days is content created that would otherwise be created by a human. This displaces a person from doing this work. What PhotoShop does is fine to me. But the reality is that AI today can actually create some semblance of that entire image that is being edited in full just with very simple commands. There isn't a photo to edit, the image is entirely generated from a few descriptive words. And it'll typically generate different options. So not a single image, but multiple ones allowing the user generating the image to select the one they prefer, or possibly refine the image based on options that were previously generated. What we don't want is an AI animation of Mario riding an elephant when this animation wasn't previously created by a human. It creates a homogeneous example instead of something that requires human crafting. Maybe this is good enough to use in the drafting stages of the process of making video games, but it would be generally worse than what could be generated by creative person.
Nice rage baiting.
I'm genuinely curious as to how they could have rephrased this in a way that wouldn't be "rage baiting". They didn't imply that this was generative, and they outright stated in the title that it wasn't an egregious use.
I mean, nvidias DLSS is just an algorithm, no matter if Machine Learning was used to improve it... The algorithm itself is fast and dumb. :) But nvidia would never say that since they are an "AI" company... it's basically snake-oil marketing for gamers.
AI should be amazing when it comes to enemies and npcs on games as well. For real
Ai is not some boogey man people do both amazing and horrible things with it.
what would threat interactive are gonna said about it though 😅
Did they really need to patent it?
In 2 years even binary search algorithm would be called an AI
LMFAO
RIGHT? These cavemen we have to share the world with, I swear to fucking GOD
@@_.-. I just fixed Nintendo Life's Commodore 64.. LOL
I mean you don't have to use it, but it helps the game run better
So they're essentially only upscaling imagery to higher resolution. At least it's a perfectly safe use of AI.
I can't see the difference on my phone.
PSA that Nintendo rejected the kind of AI that puts artists out of a job, but other types of AI such as image-upscaling or ChatGPT are fair game.
ChatGPT isnt fair game though, its still generative
Please do not lump chatGPT, which is essentially the written form of generative AI, with a resource saving upscaling technique
"You see, generative Ai is bad because all those people lucky enough to be born into a golden egg hen body have now to get an actual job"
AI has its benefits, but becomes a problem when over-used :-)
But I turned off ai to play regularly.
Didn't Nintendo use this already? Like with Splatoon or something?
They used FSR 2.0 I think. AMD's (inferior) answer to DLSS, that is more available and not locked to hardware, but unfortunately hasn't caught up in terms of functionality.
NOT ALL AI IS THE SAME AI.
AI?!
I would love if they would use a little AI in animal crossing, so the villagers are more fun to interakt with 😅
Night and Day? Uh....alright lol
😄 Nintendo folks aren't used to it. Hahaha..
oh so there gonna sue sony for using the same type of idea
AI upscaling is not a problem, this is a good use of it, GenAI is the real issue, which this tech does not make use of.
Not a fan of this kid let alone the whole channel .
Am I the only one who finds it kinda sus that they're patenting this??
i love AI and chatgpt but this is just bad
Generative AI is not evil
BLASTPHAMI SLANDER!!! HISSSS
*continues video*
oh, this is a good thing
I would argue that it potentially could be a problem. A neural network being used for upscaling isnt simply just "making the image larger". We've had non-AI methods of doing that for years, and if youve ever used an emulator before, you're probably familiar with scaling methods such as linear, bilinear, or the pixel art destroyer XBRZ.
The problem is this. Unlike previous methods that sample a smaller image and use that sample as a basis to reconstruct the image at a larger resolution, neural network based upscaling typically uses a more optical approach. That is, an algorithm will attempt to "identify" parts of an image, like grass, trees, humans, clothes, etc, and then will hallucinate fake detail not found in the original image to give the appearance of the upscaled image being "more HD" than the original. The only problem is that many of these neural networks get confused when looking at scenes with strange or unique visual elements, such as HUDs and small, fast moving VFX. There are tricks to get around this, such as having the HUD be natively generated at a high res while upscaling more performance-heavy 3D elements, but the overall point is that oftentimes images that have been AI upscaled can look blurry, have incorrect detail, or look messier than images natively generated at that resolution.
We'll have to see how exactly Nintendo implements it, but unless their upscaling software is somehow more advanced than everything else the industry is using, i sure hope there is a way to switch it off.
This is why i use MClassic to play Nintendo Switch games. When i see ppl play Mario Kart 8 whide out MClassic i think they playing on Wii U. XD
What the hell is 'eck
AI is amazing. DLSS is what Nintendo needs. No more PC emulation needed to make switch games look good.
Yea this AI is the other one that steals art from Artists is not and Nintendo already said they would never do that but up scaling DLSS AI is good
Nintendo Switch 2 DLSS 4
Oh my God these videos be sooooo pointless
video game patents are not ok
I'm sure Japansse fans were equally concerned when nintedo off America was getting it/s own Famcom and even made it own games... Like AI are clearly humans andpeople too so I'm tried of people being extremely ractist to them and using us autistics and the standard of why Ais are bad... Like I get it you suck at your job and don't want to replaced as you have no skills at life... But stop being extremely racsist and hatefull about it... Espacly when you use us autsics as the text book example of what AI can do... When you know AIs are nwerotypical like you guys...
Wow Felix, you really baited me with this title. I was fully prepared to give a lecture about different types of AI and its mostly harmless applications and i honestly didn't expect you to single out Generative AI as the single 'bad' application all by yourself. Bravo 👏
Video games, all this time, all along, have always been using AI. Why is this news?
Okay, I watched the video and it’s talking about a different AI. Sheesh.