In an era where all human knowledge is available to anyone with an internet connection, the truth is often more obscured than we’d hope. The sea of lies and disinformation is one thing, but the greater threat lies in the counterfeiting of humanity. It was bad enough when social media obliterated our social landscape, blinding us into believing it was fostering togetherness. It was soon made worse by rudimentary spoofing technology that created bots to converse with human posters. Worse still was the cryptocurrency craze, vulgarly reducing connection to cold economic concepts based on the unreality of computer “work” equating to value and blockchain “contracts” equating to communities. Now we are knee-deep in an era of the complete erasure of humanity itself, predicated on the lie that computers rapidly calculating enormous math problems can make us obsolete.
Now that the damage is done, many video game developers, not letting any bad idea go to waste, have tentatively embraced this technology to “aid” in the creative process. Even the heretofore untouchable Larian Studios is defiantly defending their use of generative machine learning in the conceptual process, never promising they wouldn’t use it, and never promising that it won’t replace concept artists in the future. This form of counterfeit creation is highly destructive and threatens to pulverize the connections between art and the recipient.
Even hardware manufacturers have given up on public goodwill in pursuit of some non-existent breakthrough in the field, hoping to be among the first to claim market share and receive hyper-profits from our, and ironically, their own, downfall. Cast your mind back to a not-so-distant past when Nvidia, trying to generate excitement for their algorithmic models, showed an unimpressive interactive demo showcasing the “conversations” you could have with brain-dead, voice-compression-addled NPCs residing in the deepest depths of the uncanny valley. It appears that as soon as they realized the failure of this demo, and therefore the failure of this technology, they all but totally left the consumer market. I’m not saying that there is any virtue in the ability to buy the latest and greatest GPUs; that’s a conversation for another day, but Nvidia practically leaving the market behind is highly relevant for two specific reasons.
First, they are the market leader, and one can infer the market’s direction by observing Nvidia and other similarly poised companies. Second, it illustrates in no uncertain terms that the most powerful and wealthy among us are often among the most stupid, as evidenced, among other things, by their blind willingness to hand over any aspect of society to glorified calculators. In the current structure of society, that is very worrying.
Regarding NPC conversations generated by LLMs in RPGs, I admit I was one of many who thought the idea was neat. NPCs could dynamically change their dialogue based on story events and even player exploits? Quest givers could have an infinite hopper of quests that could feel unique? Well, knowing what we know about “large language models,” the algorithms most likely used for such features, we can be sure that it would not have played out as we originally imagined.
I recently came dangerously close to experiencing something like this. I’m playing a little bit of Where Winds Meet, a recently released, free-to-play action RPG with a gorgeous wuxia fantasy setting. The game is beautiful, and the martial arts action is really great, so I have what I need from it. In an interesting twist, the game allows you to have generated conversations with some of the NPCs. In a small village, I found Zhao Dali, an accomplished martial artist who had to stop formal training when his estate closed down. Now dedicated to running caravans, his passion for martial arts remains evident as we meet him practicing Tai Chi forms. Upon talking to him, a chat window opens to allow engagement in a chatbot conversation. I admit that I did not enter into this conversation with the most charitable attitude, but the fact that I didn’t immediately close it down and throw my phone out the window is about all the charitability I can muster.
For academic reasons, I decided to engage. I will relate the conversation I had with him, minimally paraphrased.
ZD: (training listlessly) All this martial arts training doesn’t mean anything because I don’t even have a wife. What do I train for if I don’t even have a family to protect?
Me: I fail to see how that’s relevant. Martial arts practice is about disciplining yourself and protecting the weak, is it not? What does finding a wife have to do with it?
ZD: (stops training, looks at you) You speak the truth, young master, but I’m getting older, and I fear that my time to find a woman to settle down with may pass me by.
Me: You have honed your body into a powerful weapon. Use your strength righteously, and you’ll have a good chance of finding a person who can appreciate your virtues.
ZD: (sighs) Perhaps you are right, young master. I will continue to look after people, and I will hold on to hope that I can achieve a happy domestic life.
Me: Great! Well, I must be going.
ZD: Wait, young master! We’ve only begun talking. Why are you in such a hurry to leave?
Me: Because I’m participating in the tournament.
ZD: A tournament, you say? What kind of tournament? Perhaps I can give you some pointers?
Me: The martial arts tournament that’s currently taking place ten feet away from us. They invited me to participate, so I’m going to.
ZD: A martial arts tournament, you say? Just make sure to focus your mind, remember your training, and surely you’ll emerge victorious!
Me: Thank you for the tips. I must be on my way now.
ZD: Wait, why are you in such a hurry?
Then I quit the conversation. I tried to be polite, but this guy wouldn’t shut up.
I will admit that I was charmed at first. You could tell that writers prompted the chatbot with the loneliness narrative, and it almost made me feel for the guy, but the limitations of imparting actual characterization are obvious, even before you get to the annoying “no, please don’t stop using this chatbot” part, where he did not want to let me say goodbye. It was a disturbingly uncanny conversation, and the fact that the chatbot had to narrate its physical actions was an interesting choice, but one that didn’t work for me. I wanted to connect with him, but he wasn’t a convincing character.
One of the most human aspects of fiction is characterization. Powerful subplots like Magnus and Yumil’s relationship in Ogre Battle 64: Person of Lordly Caliber as they end up on opposite sides of a conflict that sees innocent and immature Yumil descend into madness and corruption, or seeing how the citizens of a blasted Paris in the prologue to Clair Obscur: Expedition 33 cope with the certainty that many of their loved ones will be atomized from existence in mere hours, or the deep, aching melancholy of the of the inextricable connection and forced separation of Link and Saria in The Legend of Zelda: Ocarina of Time are examples of the type of characterizations that these large language models could never effectively replicate, let alone create on their own.
By their very nature, they could never concoct a new character situation or relationship that has not been done before, nor could they imbue a familiar character journey with an author’s unique touch. In other words, while it’s true that artists are inevitably inspired by the art that came before, the large language model algorithms do not receive inspiration and produce a work in nearly the same way. For them, it’s all about input, output, and guessing the next word in a sentence or pixel in an image. It’s made more catastrophic by the fact that these models are trained on real pieces of art. Using LLMs in RPGs is not intelligence, artificial or otherwise; this is not a creative tool, and for goodness’ sake, nobody should be getting any artistic inspiration from it. This is just an iteration of the same machine learning algorithms we’ve had for generations.
There is the pervasive thought out there that “this is here to stay” or “you can’t put the toothpaste back in the tube” or whatever. Where there is a lack of will and imagination, this will definitely prove true. However, creators have a choice: Hold on to their power to create or hand it off to a technology that will only produce regurgitated “content.” I don’t wish to turn a blind eye to the pressures of game development, but this is not the answer. A far better answer that has been proven and vetted again and again by history is the organization and democratic power of the workforce being leveraged for better conditions, pay, and autonomy in the workplace, not the very likely future of further worker disenfranchisement, layoffs, and handing over work to a computer so that companies can extract more profits at the expense of artistic value. This type of computing has its place, but it is far from the profoundly human act of creation. Do not let them fool you into thinking otherwise.
