Industry-led Labels & Definitions: The Latest Barrier Making Boycotting Generative AI In Game Design All But Impossible

A recent controversy seems to have hit Everybody's Golf: Hot Shots, which has pulled it and its publisher, Bandai Namco, into the growing public scrutiny involving the use of generative AI in game development. The game itself is reviewing fine, at best, but what has prompted some reviewers to air disclaimers in their assessments is concern that developer Hyde has used generative AI as part of their development process. Now, to Hyde and Namco's defense, they admitted that they used proprietary software like SpeedTree to populate tree and grass assets in this specific game. After doing their own independent investigation, Kotaku conceded that the game's trees and leaves didn't look AI-generated, and instead, just looked "ambiently junky." On top of that, Bandai airs the SpeedTree logo during the game's starting credit sequence, leading up to the game's starting menu. Nonetheless, sites like Nintendo Life published reviews of the game, with snippets raising concerns about Namco's wording regarding the extent of their use of generative AI programs to create graphics. In Nintendo Life's case, they wrote:

One footnote here is that Bandai Namco revealed some of the foliage textures were produced using generative AI. Looking at the results, it's hard to imagine that any added efficiency was worth the loss of cred from using the controversial technology.

But again, none of this is "news" because the Steam page for the game has a note on it, and this note has been up for months:

Not to harp on this point, but SpeedTree is literally one of the first logos that grace your screen. So, what exactly did Bandai's use of generative AI entail? Clicking their link reveals that the game simply used SpeedTree to make a bunch of tree and grass assets for cheap, and they did light editing of those assets before shipping the product. That's a far cry from the generative AI scandal that soured my personal opinion of The Alters. This isn't the first time Bandai has eschewed what most of us would consider the "standard" definition of what generative AI is used for in game design by bad actors. When publishing Taiko no Tatsujin on Steam, Bandai aired almost the same disclaimer about generative AI use once again. When you followed up on their footnotes for details, it turns out Bandai classified using Vocaloids and text-to-speech software as "AI Generated Content."

Now, they're technically correct, but putting Hatsune Miku and text-to-speech software in the same category as ChatGPT or Grok is wild, and I'm going to assume most of you would agree. Some of those tools might fall under a broad umbrella that would group them with ChatGPT, LLMs, or other neural networks, which are displacing programmers, coders, and artists from their hard-earned wages and jobs. Even then, the issue is that we are working with industry-led definitions, rather than government-codified ones, on what qualifies as the use of generative AI in games and art in general. In this case, Bandai's barometer far exceeds that of practically every North American and European developer because no one other than Bandai would air a disclaimer about AI use over SpeedTree or Vocaloids. And if these developer-specific standards are all we have, that's a looming goddamn mess for consumers to navigate.

There are two possible readings of Everybody's Golf Hot Shots. The worst possible scenario is that Speedtree was only used for the foliage graphics, and the game's design and programming teams used an LLM to supply additional graphics and supplemental grass and tree models. While some seem to be convinced that's the case, Bandai doesn't have an LLM or AI engine listed on Everybody's Golf Hot Shots beyond their SpeedTree credit, and if they are already listing Vocaloids and text-to-speech as AI tools, why would they lie about using ChatGPT? The second, and more likely, scenario is that someone at Bandai is nervous about the fluidity of what is and is not considered generative AI by consumers and was overly loose with the rules in defining such tools in this and their other works. That doesn't mean the upper echelons of Bandai Namco are saints or guilt-free of sin or shitty anti-artist tendencies. What I and others do not and cannot know is if someone at the company's highest rungs wants to see AI in game projects, as many other CEOs do elsewhere. If that's the case, there's a chance someone working on this game wanted them to get off their case, so they pointed to SpeedTree as generative AI and got away with it.

Nevertheless, this controversy is another example of why supporting blanket boycotts of all games that use generative AI feels completely impossible, and that's likely been the case for a while. A few years ago, Polygon published an article about Revolution Software using generative AI to upscale their art assets for the new remasters of the classic Broken Sword games. The headline and byline initially got me riled up about wanting to protest the remasters, but as I progressed with the article, I realized the situation was far more nuanced. They weren't asking ChatGPT or a neural network to make a game, but instead running their old code through an AI algorithm to interpolate frames, and still had actual artists fill out the details and correct minor errors. And they were doing that as a studio with fewer than fifty employees and cited multiple examples of how they couldn't afford to hire more artists. They were an independent game studio struggling to make remasters to keep the lights on and were using AI to automate tedious tasks. To some, that's still no excuse to turn to AI in game design. Still, I caution some absolutists to consider that some of Revolution Software's AI use amounted to actions no different from those of artists today using Adobe to autofill backgrounds. And I hate to say this, but most editing software has used AI to up-res old images for decades, years before people even knew OpenAI existed. That's not to excuse techbros from nefarious or shrewd business practices. Game developers have been trying to get away with selling shit that used machine translation instead of professional localizations for over thirty years. My point is that some of these battles have existed far longer than some are willing to admit, and the moral grey areas here have remained unresolved for equally as long.

Before I make my next point, I want to be clear about AI use. Yes, LLM training is massively expensive and power-intensive. The training of these LLMs often exploits the works of authors and artists without compensation. Yes, data centers consume a significant amount of energy. In general, the hype surrounding AI far exceeds what it can actually do to improve everyday life, and this has led to legitimate concerns that AI companies are operating in an economic "bubble." If and when that bubble pops, there will be widespread impacts across the world that negatively impact millions of people. Practically speaking, most AI systems and engines are still prone to hallucinations and can worsen sycophancy. And as a teacher, I can tell you AI, especially Google's new AI-based summaries, which pop up whenever you submit a question into its search engine, has made my life a living nightmare by making entire teaching strategies and activities either trivial or intellectually hazardous because the results are still spouting out fake news and deepfakes.

I want to make all of that absolutely clear before I say that anyone who wants to be a Puritan about zero generative assistance in entertainment should have stopped watching films and television in the early 2000s when everyone switched from real film or animation cels to digital film and Digipaint. Yes, we can express some of that change in art as part of a sliding scale, but cropping an image from a photo to have it exist in a transparent background to make meme generation easier? That utilizes AI-based software, which has been normalized for years. Also relatively normalized are tools like SpeedTree. While there are plenty of absolute lines we can all agree to as no-nos in game design, hence why I and many others were visibly upset with The Alters, there's way more ambiguity than most are willing to admit to as well. For example, last year, VGC and other press publications reported that EA's CEO, Andrew Wilson, said that he felt there was a "real hunger" among developers to use generative AI in game development and that they were confident that development processes could be "positively impacted" by AI. Andrew Wilson is a noted piece of shit, but digging a bit deeper reveals that he grouped stuff like SpeedTree, Adobe Autofill, and free 3D asset libraries with ChatGPT, which is a big WOOF if you ask me. We can't trust anyone wearing a suit to define what generative AI use is and isn't in game design because they're all morons and they don't know what they really want. And to be honest, if I were an EA developer or programmer being voluntold to complete a workplace survey about AI, I won't lie; I'd probably write that I would accept AI in my work if it meant I could convert a 40+ hour workweek into a 37-hour one. I'd say "yes," ethics be damned.

Simply put, it's not controversial to say we can't trust the industry to self-regulate. The recent situation with Namco and Everybody's Golf Hot Shots is yet another entry in that growing list of examples of why. Beyond that, there's no doubt that there are executives leading game studios who want to squeeze out the average artist and programmer, in the guise of cutting costs. Those are the last people we should trust to figure out what does and does not "count" as needing a disclaimer on AI use. But even if everyone in the industry were well-meaning people who respected artists, industry-led disclaimers and tags still wouldn't cut it. Depending on one's definition, generative AI right now is likely involved in the pipeline of countless things, so if we say no to anything generative, there's nothing new to enjoy moving forward. Boycotting everything isn't going to help, and instead of hoping that public pressure will force the toothpaste back in the tube, this is only getting controlled with legislation. Unfortunately, that's unlikely to happen in the US or elsewhere any time soon. So, buckle up and get ready to be confused. Navigating these waters is not getting easier.