Could Blizzard use AI tech to refine or "enhance" their games i.e. World of Warcraft?

[OwenBurton]

No. AI has little to no right to be employed in creative spaces.

We humans are amazing at many things... but we are most amazing at our creativity. Art, music and story are some of the things that define us as people and cultures. It should remain a flesh and blood thing.

AI used to rework old textures or models can be successful and save some man-hours... but it still requires review. Otherwise, customers will notice any jank when they see it. I can already tell they used AI or some form of automation to retrofit every Non Combat Pet to support the new wearable candle toy from Gundargaz. On some pets, it looks ok, on others the alignment is too far forward, and on some the candle is too far sunken into their model. They likely used an automated process to look at the shared skeleton rigs for these models and extrapolate a space above their 'head' to anchor the candle to. Only on some, the actual pet model is much larger than bones of it and you get weird clipping or alignment as a result. A set of human eyes sees it immediately, but I guess the only eyes that looked this over were the ones signing off on the bot's process log. To quote the peons, "Job's done!"

I kind of agree with this in part. Nothing can and should supplant the unique creativity of the human individual in work or art. But neither I think should we reject modern technologies in enhancing both as long as they are used in balance and with discernment.

Blizzard is already using AI, for at least a year now, but they also claim they are being very careful in implementing and applying it to in-game artwork:
Blizzard Clarifies Use of AI After Patent Surfaces [UPDATED]

[Eurhetemec]

But that's also not my point. My point isn't "all AI is efficient" nor is it "all people are inefficient" or any mix of the two. My point is simply and purely: "intentionally maintaining inefficiencies with no good reason causes systemic problems". Period. The fact that sometimes companies misidentify inefficiencies positively or negatively does not matter for that point. Of course there will be mistakes; these systems aren't and never have been perfect. That doesn't invalidate the operating principles.

To use an analogy: I'm saying "you find someone without a pulse, start doing chest compressions on them" - and your reply is "but sometimes you miss their pulse and actually they have a pulse!". Yeah, okay. SO WHAT? How does that invalidate the previous statement? You are talking about a different step in a different process, and however valid your point may be it has no impact on the validity of the point I actually made.

My point is pretty clear I thought - researchers and professionals have spent decades, close on a century, creating ways to identify and deal with inefficiencies in industrial processes and a shorter time for dealing with them in office processes. Basically no serious research has been done on identifying inefficiencies in creative-oriented processes (and less than we'd like on software-development-related ones), and yet we're seeing replacement of creatives and creative processes with overpaid schlubs armed only with generative AI.

It's unserious and stupid behaviour by people who are literally paid and promoted to know better than to engage in such unserious and stupid behaviour.

Also your analogy is a rather strange choice. This literally is NOT life and death. If humans with no pulse were fine for say, 3 hours (instead of barely being able to make it 3 minutes without brain damage at normal temperatures), people would be much, much, much more careful about starting chest compressions (which can potentially kill someone who does have a normal heartbeat you just failed to detect, as well as causing serious bruising, broken ribs, etc.).

AI is not essential. It is not something we need to rush into. It is something we should be selective and careful about, and frankly wait for it to reach serious standards. Instead supposedly serious people are behaving like children with Halloween candy. There's going to be a lot of puking later.

[Biomega]

Basically no serious research has been done on identifying inefficiencies in creative-oriented processes (and less than we'd like on software-development-related ones), and yet we're seeing replacement of creatives and creative processes with overpaid schlubs armed only with generative AI.

I'm not even sure what you mean by this. These "creative processes" are production processes like any other, determined by cost, output, and quality. All that changes is the evaluative metric for quality, there is no structural difference to other processes.

You're trying to create some kind of distinction here that makes no sense, economically speaking. We're not talking about free artistry here - we're talking about creative work as part of a business operation. For that it really doesn't matter, structurally, if you're someone who screws on toothpaste caps or someone who paints pictures of orcs decapitating humans. The underlying workflow dynamic is practically identical in structure, differing purely in executive detail (mostly the evaluation metric, as I said).

Efficiency is simply maximizing the way the various workflow components stack up against each other. AI doesn't change that, because at the end of the day, there'll be an evaluation of the creative work either way - whether it was made by a human or by an AI doesn't matter for that step in the process. No one is or has ever been suggesting some kind of fire-and-forget model by which we go from AI to consumer with no intervening quality control. That's maybe something to think about in the more distant future when it comes to ad-hoc content generation, but is no factor in the debate at hand. Everything that's made by AI will still first pass through human hands for evaluation just like a human artist's work would.

The efficiency calculation therefore only really applies to the prior workflow: how long does it take and how much does it cost to produce the end result, and how does that end result stack up against the target metric. Assuming a constant target metric, AI is pretty much a straight-up efficiency gain in most cases. Assuming shifting target metrics, it becomes more complicated - you might want to up the volume, for example (because you can). In which case it comes down to the business plan, since you're no longer in a straight comparison.

Also your analogy is a rather strange choice. This literally is NOT life and death.

That doesn't matter. You're making the classical mistake of confusing content and mechanism of an analogy. The content of the analogy is wholly irrelevant here. Only the mechanism matters. It could be about clown noses and tropical fruit, wouldn't make a difference. The "life and death" part isn't relevant to the mechanism I'm explaining.

This worries me greatly. Are you sure you're equipped for this kind of debate if you can't even parse an analogy correctly?

[Eurhetemec]

That doesn't matter. You're making the classical mistake of confusing content and mechanism of an analogy. The content of the analogy is wholly irrelevant here. Only the mechanism matters. It could be about clown noses and tropical fruit, wouldn't make a difference. The "life and death" part isn't relevant to the mechanism I'm explaining.

This worries me greatly. Are you sure you're equipped for this kind of debate if you can't even parse an analogy correctly?

I don't think someone who so quickly drops to crude, fake-concern condescension so immediately and desperately is equipped for any kind of debate, frankly. So I think you need to fix your own problems before worrying about those of others!

Further, you're simply embarrassed that you used a very poor analogy, and seeking to place your poor decision-making on to other people. Life and death is absolutely relevant, because the only reason we approach the situation you present as analogous in the way we do is that it's life and death. I am parsing the analogy correctly - you chose a life and death analogy to attempt to impart a sense of urgency and that something must be done. Then you got called on how inaccurate it was and you didn't like it, so you deployed condescension and hand-wringing to try and distract from that. That's on you.

[Biomega]

Further, you're simply embarrassed that you used a very poor analogy

I didn't. It is in fact an excellent analogy, because it's a common scenario of general recommendation with a specific and common failure point.

You simply don't understand the point I'm making. I'm not even disagreeing with most of what you're saying. There absolutely is a need for proper identification of actual inefficiencies and so on. That's not what I'm pointing out as the problem - I'm point out that despite the validity of all that, it doesn't change my original point. That's what that analogy was for. The second part (the objection) is something that is clearly valid and true - it merely doesn't change the first part. Because it's a separate process. That's what you don't seem to understand, because you can't abstract from the analogy to the underlying structure and get caught up in the image. Classic mistake, as I said - just a result of poor parsing and insufficient experience with actual arguments.

We could use any other image to construct the same analogy. "If someone has made a typo, leave a comment on the document" "but sometimes it's not a typo but an intentional misspelling?!" - yes that's true, but it doesn't change the first statement. Nor does it mean you don't try and identify typos correctly. It's merely a different process that does not alter the validity of the first statement. What image you use in the analogy doesn't matter. Only the structure does. That's how analogies operate. When someone says "eggs are to chickens as apples are to apple trees" you don't then go EXCUSE ME are you suggesting eggs are apples?! Hello?! one is an animal and one is a fruit OMG?! and why are we talking about fruit suddenly this was about poultry?! Because that's not the point. The image doesn't matter. Only the mechanism does.

And it's no surprise that you're going frothy over this little detail, since obviously your actual point doesn't have a leg to stand on Not my first rodeo.

[Daronokk]

The only good use of AI currently is to upscale outdated graphics and textures, which Blizzard did a first try with Cataclysm Classic. It holds on to the graphic style and soul and removes tedious work for Blizzard employees to do it all manually. It's a relatively easy fix to apply and benefits all.

But these AI generated pictures serve no purpose

[Eurhetemec]

And it's no surprise that you're going frothy over this little detail, since obviously your actual point doesn't have a leg to stand on Not my first rodeo.

Oh the desperation of stuff like "frothy" for mild criticism! Come on, you're better than this!

I'm not even disagreeing with most of what you're saying.

Indeed, and that's why there wasn't much point in engaging with the rest. It wasn't exactly a massive disagreement, but you decided to get upset and fake caring and engage in a bit of troll-y condescension because I mildly criticised your analogy, which remains dubious.

As for

We could use any other image to construct the same analogy.

Absolutely. Yet you intentionally chose the most urgent and extreme possible one. You did that for a reason and you can't be all "IM THE BIG DEBATE GUY HERE DONT TRY 2 DEBAT ME!!!" and claim otherwise, frankly! It's a classic tactic, as you well know.

[Biomega]

Absolutely. Yet you intentionally chose the most urgent and extreme possible one.

I think I already explained why I chose that one: because it's common and easy to parse. Well, so I thought.

From the beginning I've been quite insistent on the fact that the image doesn't matter. So trying to now get me with AHA YOU CHOSE THAT IMAGE SO IT MATTERS! only demonstrates that you still don't really understand the problem, you're just trying to mask it by turning things around on me so you don't have to defend yourself.

Which is why you're still going on about this, instead of just swallowing the L and trying to get back to making an actual point. If you have one.

It's okay to just admit a miss every now and then. We learn when we fail. I'm abrasive in order to be educational. You'll remember this, I guarantee you. And next time you see an analogy, hopefully you won't make the same blunder.

[korijenkins]

The only good use of AI currently is to upscale outdated graphics and textures, which Blizzard did a first try with Cataclysm Classic. It holds on to the graphic style and soul and removes tedious work for Blizzard employees to do it all manually. It's a relatively easy fix to apply and benefits all.

But these AI generated pictures serve no purpose

I voted other because I also think putting an AI in charge of class balancing would be smart. Actually, they could also be useful in creating the rough zone shape for the environment teams. As it stands, they have to manually create the islands, smooth out all the shorelines, etc. Takes a huge amount of time.

But actual design work? Nah. No way do I want AI designing tier sets for the classes.

[Eurhetemec]

I think I already explained why I chose that one: because it's common and easy to parse. Well, so I thought.

From the beginning I've been quite insistent on the fact that the image doesn't matter. So trying to now get me with AHA YOU CHOSE THAT IMAGE SO IT MATTERS! only demonstrates that you still don't really understand the problem, you're just trying to mask it by turning things around on me so you don't have to defend yourself.

Which is why you're still going on about this, instead of just swallowing the L and trying to get back to making an actual point. If you have one.

It's okay to just admit a miss every now and then. We learn when we fail. I'm abrasive in order to be educational. You'll remember this, I guarantee you. And next time you see an analogy, hopefully you won't make the same blunder.

Oh the irony.

Also "I'm abrasive in order to be educational" on a WoW message board is, um, incredibly special.

- - - Updated - - -

Actually, they could also be useful in creating the rough zone shape for the environment teams.

You don't need AI for that, and indeed I'm not aware of any AI tool focused on that. Procedural generation processes have been doing that kind of thing since the 1980s, literally! I would suspect any AI tool that was focused on that would just be rebrand of a proc gen one. I know one product my company uses rebranded itself with AI in the name even though 95% of the product doesn't use AI nor would benefit from it.

As it stands, they have to manually create the islands, smooth out all the shorelines, etc. Takes a huge amount of time.

Nah. If Blizzard are doing that in 2024 it's because they want to, not because they have to, because they're committed to a certain level of quality and precision. Proc gen could have created the islands, smoothed out their shorelines and so on in 1994, let alone 2024.

[Biomega]

I voted other because I also think putting an AI in charge of class balancing would be smart.

That's probably not needed. Class balancing is rarely a problem of them being unable to get numbers right. It's more an issue of them having different ideas about things than players do.

The goal isn't to have every class's DPS sims be perfectly aligned. It never was. Blizzard has many goals with classes, and numbers are only a part of it; and a fairly small one at that.

Also keep in mind that "class balance" isn't one static thing. It's not like there's an agreed-upon benchmark standard that every class is measures again. There's a gazillion different scenarios and a gazillion different play styles. If you made everything equal on a patchwerk-style raid sim, for example, it could be that suddenly things are super out of whack in dungeons. Or in PvP. Or Delves. Or whatever else. And even in raid you might have one patchwerk fight in a tier, but all the other fights are something else, and suddenly things are borked there.

And that's not even taking into account things like play complexity, which is a big concern for Blizzard - they're fine with some simple-to-play specs performing slightly worse than super complex ones, for example. And just because some Liquid or Echo raider can pull of a perfect rotation with 50 buttons and 3 tiers of decision trees doesn't mean that translates over to Joe Schmoe in their normal-mode PUG - so who do you balance around? If you balance around Joe, the skilled raiders will tear balance to pieces. If you balance around the top echelon, all the casuals will feel like they're wet noodles and quit. And so on.

There is no need for AI here because there aren't any hard, fast lines to optimize around. AI is great when it comes to "do X, Y, and Z in ways A, B, and C"-type instructions, but a lot of the balancing is touchie-feelie guesstimation that the devs probably couldn't even put to specific words on paper, let alone feed into an AI. And it'll be... a while... until we can simply tell the AI "make the game more fun for everyone".

Players like to oversimply "class balance" as if it was simply about matching a graph to a given line, but that's not how it works.

[qwerty123456]

Those videos use AI to generate images and then use other AI programs to animate it slightly to make it look like a short clip from a video they currently can't do real and long cinematics.


If humans weren't greedy I wouldn't have a problem with AI art. Like if Activision were to hire 20 artists to contribute reference material to an AI program that only uses their stuff and stock reference material so it wasn't just leeching shit off the web, and they paid the artists in perpetuity to use their art for the AI to refence from I don't think it would be unethical to use. You could then have Bob working for WoW 3 who wants to add a new race ask the AI to generate him some concept art for big titty Amazonian hippo human hybrids with leech mouths and bat wings within a matter of minutes to show off at the next meeting. If everyone shoots down his race idea it saved an artist hours of work. If people liked the race you could then have the artists and the AI refine the concept until they get it "perfect".

[Dacia Ultan]

Ehh. My primary response to this is that I'm reminded, once again, of how much I dislike the use of AI image generation toys.

[Val the Moofia Boss]

I could see AI being used to help quickly generate terrain if Azeroth was to be remade on a 100x larger scale to adjust for the travel times with skyriding, where currently you can fly between kingdoms in seconds. But I do not want AI generated textures.