[Opinion] Uhm... AI-powered WoW is coming sooner than people think.

[Clone]

My gripe is with the people that are saying they want this. And I don't get why. I grew up with WC3 and early WoW. It has a specific style, why would you want anything else?

Personally I was sold on the style of the cinematics than the in game graphics. I've always saw the latter as just representations of the former in a world where that technology doesn't exist yet.

The style of the cinematics is also much closer to photorealism than it is cartoony.

[arandomuser]

i hate high fidelity games i cant see shit unless i turn my contrast all the way up, i think my vision just sucks but w old timey graphics its alot better

[NineSpine]

Fine tunes and loras alone make the base models usable vs photorealism slop. It’s astonishing how good but unusable base models are at times.

- - - Updated - - -

Oh I know, that’s why I like AI tools as referential and inspirational on the fly generations that help with your normal workflows. Being able to slap your mostly done sketch to see a quick result before you commit too much time doing it for yourself is nice - or changing the entire lighting or background as such. It’s very nice for convenient tuning - I’m sure when it comes to music it can be done similarly. The stuff I’ve seen people create using AI gen tools as a basis but not the one button done thing is amazing. But you’re right a lot of this stuff leans heavily into copyright territory depending on how much you use of it directly in a project or if you’re just using it as inspiration.

That’s also why I love the tools for 3D - not many people want to UVMAP a face again and again, let alone a box or a pipe system. Having AI do this efficiently and better vs wasting all that time is really efficient.

I’ll admit being able to tell ChatGPT to make a script that usually works in very simple ways for all of these tools is nice. I don’t trust it to do anything but basics though.

Didn’t know C was so restrictive but python makes sense - casually using it over the years I never have to worry about it since it’s just globally available in a cmd prompt and calling it.

Very narrow use case gen AI tools can be extremely powerful, and that is likely where the innovation is going to come from. They can increase productivity when used responsibly in those use cases. Those use cases just aren't that sexy and don't attract massive piles of investor cash, so instead we are talking about "WE WILL HIT A BUTTON AND IT WILL MAKE WOW!"

I wouldn't call C restrictive. Quite the opposite actually. It's extremely powerful and fast. That's why it is almost always the solution for systems level programming, like drivers. The restrictions just come from the amount of complexity that comes with that power. Ironically, the syntax of C is really simple, but once you start building out a C code base it quickly can become a monster.

[PenguinChan]

Very narrow use case gen AI tools can be extremely powerful, and that is likely where the innovation is going to come from. They can increase productivity when used responsibly in those use cases. Those use cases just aren't that sexy and don't attract massive piles of investor cash, so instead we are talking about "WE WILL HIT A BUTTON AND IT WILL MAKE WOW!"

I wouldn't call C restrictive. Quite the opposite actually. It's extremely powerful and fast. That's why it is almost always the solution for systems level programming, like drivers. The restrictions just come from the amount of complexity that comes with that power. Ironically, the syntax of C is really simple, but once you start building out a C code base it quickly can become a monster.

That’s exactly why I think these tools are amazing and going to change a lot in the background. I know many corpos and grifters are using it to gain money off of front ends but this isn’t the first nor last time someone has done this off new tech. I mean sure it’s funny to have mickey mouse fighting vader with light sabers but that’s obviously meme material. The rest of the common usage is (obviously) smut and nudity.

But the amount of time I’m going to save by using these tools in the boring aspects of my work or hobby creations is nice. One thing that isn’t so niche and actually quite impressive (but insufferably bad for grunt workers) is how well AI can translate just a couple views from concept art and turn it into a functional usable object jn 3D space.

It’s a far cry from hyper realism at the click of a button (you gotta pay $200+ for those plug-ins / software lmao) but it definitively caught up quickly from only generating depth maps.

One thing I have to ask is that AI rn is switched over to (or maybe something more advanced behind closed doors) what is essentially multi-core LLMs instead of what they were before which was more like ARM in a phone. I’m trying to understand it from a layman’s perspective but I would’ve imagined this’d insanely increase the reliability and function of the AI.

[Xilurm]

There's been some news about AI making interactive videos, where you can control it like a video game. There was one with Minecraft some months ago and recently the Microsoft one.

I wonder how far those will go. I don't want the video game industry to end up dying to something like that. Being able to create whatever video game you want out of a few prompts might sound good but it's gonna end up being so boring.

[NineSpine]

That’s exactly why I think these tools are amazing and going to change a lot in the background. I know many corpos and grifters are using it to gain money off of front ends but this isn’t the first nor last time someone has done this off new tech. I mean sure it’s funny to have mickey mouse fighting vader with light sabers but that’s obviously meme material. The rest of the common usage is (obviously) smut and nudity.

But the amount of time I’m going to save by using these tools in the boring aspects of my work or hobby creations is nice. One thing that isn’t so niche and actually quite impressive (but insufferably bad for grunt workers) is how well AI can translate just a couple views from concept art and turn it into a functional usable object jn 3D space.

It’s a far cry from hyper realism at the click of a button (you gotta pay $200+ for those plug-ins / software lmao) but it definitively caught up quickly from only generating depth maps.

One thing I have to ask is that AI rn is switched over to (or maybe something more advanced behind closed doors) what is essentially multi-core LLMs instead of what they were before which was more like ARM in a phone. I’m trying to understand it from a layman’s perspective but I would’ve imagined this’d insanely increase the reliability and function of the AI.

I'm not quite sure what you mean by "multi-core LLMs".

[PenguinChan]

I'm not quite sure what you mean by "multi-core LLMs".

As in more modern LLMs no longer compute in a single instruction type of way and instead do many instructions at the same time. I know many in the past were working on the former architecture but eventually made it to the latter. I thought that type of advancement would’ve made a bigger difference.

[NineSpine]

As in more modern LLMs no longer compute in a single instruction type of way and instead do many instructions at the same time. I know many in the past were working on the former architecture but eventually made it to the latter. I thought that type of advancement would’ve made a bigger difference.

They are trying a few different methods to get around the exponential complexity problem, and nothing is showing promise of providing large jumps. Attempts to train models further are hampered by exponential complexity and the fact that AI assets keep getting pulled in and corrupting models.

[PenguinChan]

They are trying a few different methods to get around the exponential complexity problem, and nothing is showing promise of providing large jumps. Attempts to train models further are hampered by exponential complexity and the fact that AI assets keep getting pulled in and corrupting models.

Wouldn’t it make sense then to create an LLM that’s dedicated to analyzing and purging pure / mostly AI assets? I guess this is easier for imagery and audio models than pure text ones.

[NineSpine]

Wouldn’t it make sense then to create an LLM that’s dedicated to analyzing and purging pure / mostly AI assets? I guess this is easier for imagery and audio models than pure text ones.

There are systems to detect if content is AI, but they have high overhead and aren't great.

There's also a problem of running out of content.

[Obbi87]

You would think the WoW community hates AI, but then you see videos like this getting over 1,000 likes in under five hours, and it's like... yep, people love it, and it's coming.


[check out 0:40]

In short: an AI will soon start running around in Azeroth for a million (simulated) years, memorizing every asset, every composition, and every pixel from all angles. Based on all that data, Blizzard will develop an AI model that procedurally generates new worlds and adventures in Azeroth. Designers will still guide the AI to ensure it stays true to the WoW experience that players know and love.

Buckle up, boys and girls. This is less than 10 years away. If you love MMOs combined with some RPG adventures—you know, MMORPGs—then let me tell you, you're in for some magical experiences. Those who dislike AI now will become its biggest fans the moment they see it in action.


A couple clarifications:

→ "Procedurally generated" does not mean completely random—developers will still have programmatic control over the world and story.

→ You'll have the choice between classic WoW graphics or complete photorealism. It will be the Classic+ experience you've always dreamed of, playable with your friends in whatever style you like.

→ Playing games 10 years from now will feel like stepping into a movie, where the developers (or the player) decide what happens next, while everything you know remains consistent. Your Fireball spell will always look and feel exactly the same, no matter where or how you use it.

Love this, "Trust me bro its coming"

[LilSaihah]

unironically looks worse than DW:O

Love this, "Trust me bro its coming"

my uncle works at nintendo and he said its true

[KrayZ33]

Personally I was sold on the style of the cinematics than the in game graphics. I've always saw the latter as just representations of the former in a world where that technology doesn't exist yet.

The style of the cinematics is also much closer to photorealism than it is cartoony.

Same.
WC3 cutscenes is the reason I played the game.
Everything else was "the graphics are bad because graphics are bad in strategy games"

If I could play WoW in SpaceMarine2 (or something like that) graphics , I probably would.
Although the game would most likely not be tab-targeting anymore and a different game. Which is fine too because tab-targeting is old derelict shit and a new game shouldn't have it.

[Zodiark]

Of course posting something like this on any WOW forums or fan-sites is going to get a lot of heat. Most players have been with the game a long time and messing with, or even HINTING at messing with their pre-conceptions of the game and or it's nostalgia is going to go over like a bucket of hot lead. That being said I feel like with the rising cost of games especially MMORPG's that utilizing AI to help flesh out content might absolutely be a good thing. So long as the Devs retain control of the game and it's overall look and feel to ensure it has that WOW vibe. Then I don't see what the problem is personally. I always wanted more DEPTH to the game, more ways to play and live in a certain area if you wanted. Like why should I have to keep moving on to a different theme-park or fantasy with each expansion?

If I want to RP as a defender of the Barrens I should be able to do so. However right now if you find a zone/expansion you like then you have to basically create headcannon for why you are remaining here when the rest of the player base has essentially moved on. Sure you could level lock your character but other than dailies once you finish all the quests and level up all the rep there isn't much else meaningful to do in each area anymore. However maybe with AI generated content there COULD be more dynamic quests added on a regular enough basis to allow for protracted gameplay wherever you want.

I mentioned something like this in another post I did a while ago. Just make it so there is more mega-servers so it can condense the players a bit more so there is a larger pool of people on each server/realm/zone(s).

People might argue that not enough people would be spread out enough through each zone/area to make the game feel populated while also keeping enough players in the newer zones/content to also make that bit feel meaningful. However people have basically come to terms with the way the game is currently. I believe though that if you build it they will come. So making it a thing so you could play through each area indefinitely would open the option and opportunity to play the game in a different way so maybe with bigger servers players WOULD branch out more.

In a real living world like the one we live in real life, there are tons of people living lives in all parts of the globe. You can travel there if you wish and implant yourself into that society and become part of it whenever you wish. To me it's a cool feeling knowing that people everywhere at any given time are actively out and about doing different things and engaging with things and places the world over. However in WoW it always felt like once an area is enjoyed then it basically becomes dead and lifeless. Even if I enjoy playing alone it's nice to see other players out and about. Problem is that developers can never stay ahead of players consuming content but if AI is used maybe it can.