ShareThis
Dating back to 1988, the Game Developers Conference is an annual event held at the Moscone Center in downtown San Francisco. During the convention, many talks are held on all sorts of different topics, including postmortems on completed projects, retrospectives, and analysis of the industry at large. Yesterday, designer David “Rez” Graham, who has worked on many games over the years such as The Sims 4, and who teaches game design at the Academy of Art University, delivered a talk dubbed Game AI Summit: The Human Cost of Generative AI. If the reaction, a swell of enthusiastic applause and cheers, was anything to go off of, Graham’s words struck a chord with the audience, many of whom (including this writer) headed outside of the presentation chamber to pepper him with questions afterwards.
Before we launch into what Graham said, let’s take a moment to (broadly) discuss what generative AI (artificial intelligence) is, for those who might not know. Generative AI is a new form of artificial intelligence technology that produces new art, writing, music, and more with minimal input from the user. Simply enter in a couple of prompts to specify the desired content or outcome, and the given program will make it. To accomplish this, generative AI software is trained via so-called scrapes of as wide a pool of data for reference as possible. What is this data? That’s where things start to become murky and sticky. The data is… people. Specifically, the creative spark of human beings in the form of their drawings, their writings, their songs, their films—whatever it is that the generative AI might be tasked with reproducing.
In the case of many generative AI programs, these scrapes are done without the permission (or in most cases even the knowledge) of the creators. Scrapes will cover swaths of content from across the world wide web, in some cases thousands and even millions of samples of a given subject, and then take that data to derive new content from. But therein lies the rub: generative AI is, at its core, highly derivative. It’s not making anything truly new, but instead learns from the works of others and uses it to produce a Frankenstein’s monster of hodgepodge assets from across the Internet. Generative AI isn’t capable at this point of going beyond what it’s taught via data harvesting. Which isn’t to say generative AI always produces junk (the AI images in this article are all very serviceable and, sadly, you might not even have known they’re AI if I didn’t tell you), but the results aren’t what’s in question—it’s the ethics.
Everything I’ve described thus far might sound relatively benign—”so generative AI is just a tool for creativity!” some of you are potentially thinking—but the reality is more sinister than a first glance lets on. Generative AI leads to somewhere dark and oppressive. It’s this bleak future that Rez Graham sought to highlight with his talk at GDC. In it, he described how these generative AI makers are circumventing creators, and in some cases willfully so, to pad their libraries of content via data scrapes without any intention of crediting or paying those whose work is being utilized. Someone can spend years perfecting an art style, for instance, and a generative AI maker can swoop in, scrape all of that work, and then use it to produce content incredibly similar, if not identical, to what the original creator makes.
Samus Aran generated via Google Gemini AI.
That certainly doesn’t sound very ethical, but it’s become an everyday occurrence in virtually every creative field out there, including video games. What really rankles, however, isn’t just the use of creators’ work without their permission, but also the, as the talk so astutely describes it, human cost of this technology. One cost is very straightforward: generative AI is a way for greedy companies to produce content (note, not art, which we’ll get to) with fewer workers. In the dream world of these companies, AI could eventually get to the point where it handles every single aspect of the creative process, from production, to marketing, to selling, and everything in between. In an industry hemorrhaging workers every month owing to the constant layoffs taking place, that’s a shady proposition for people who are already struggling to find consistent, stable work.
The other human cost is perhaps a bit bombastic in how it’s worded, but honestly, is there any other way to put it than to say that unrestricted generative AI will absolutely lead to the death of art? Graham used the quote from Jeff Goldblum’s character Dr. Ian Malcom in the film Jurassic Park where he states “Your scientists were so preoccupied with whether they could, they didn’t stop to think if they should.” As Graham so eloquently put it yesterday, what design problem, exactly, is being solved by generative AI? Human beings ideate, they riff off of the works of others, they innovate, and so on, all as part of the creative process. Generative AI functions seemingly as a rebuke on these creative endeavors by suggesting that there’s no point in all of that silly human creativity when an AI can do it in seconds.
My distaste for the use of generative AI might actually outstrip Graham’s, as I don’t even really agree with it as a way of reducing so-called monotony for tasks, like, say, creating multiple cells of animation for a cartoon. I think about it like this: if someone were to animate a leaf falling from a tree, taking hours upon hours to meticulously recreate the motion of a leaf slipping free from a branch and slowly swirling down to the earth, that imagery only has meaning and worth in knowing that a human being had to put in the labor to reproduce that sight on screen. What is special or interesting about an AI getting a prompt from someone to animate a leaf falling and churning out the desired result in two seconds? By this logic we should stop eating delicious meals because a nutrition shake can solve all of our sustenance needs with a couple of quick gulps from a cardboard carton.
Graham stated during the talk that companies producing generative AIs would be less destructive if they at least took into consideration the needs and desires of the creators whose works they so gleefully and greedily gobble up for scraping. That indeed, if creators in the micro were able to seize more control over how these AIs are trained and utilized them in a manner that suits their workflow and needs, that the technology would be more beneficial to the industry. It’s certainly some sound logic and I don’t totally disagree with it, but for me, the solution should be (but sadly won’t) to purge generative AI from creative endeavors as a whole. That consumers demand content made by other human beings and reject these mechanically cobbled together abominations. Unfortunately, with the industry so hellbent on pinching pennies and churning out content, generative AI is more likely to become increasingly present, not less so.
Still, as someone who thinks of himself as an artist, I find generative AI to be distasteful, especially in the predatory way that it’s being trained and utilized. As I mentioned earlier, companies are concerned with making content, not art, these days. With the streaming model, corporations have created an insatiable beast whose big appeal to their audience is always, “look how much stuff there is to watch/listen to/play!” But with ease of access comes a new, persistent, constant demand for new, new, new from users. Binge dozens upon dozens of episodes of a TV show in a matter of days as opposed to years, and suddenly the streaming provider realizes that without that new, new, new to regularly offer, users will simply abandon them and move on to whomever is offering them something fresh.
The end result of the so-called demand generation is an endless need for more content that’s driving the uptick in output of virtual slop from everyone making games, movies, TV shows, and all forms of entertainment in between the past few years. So while some might think “the death of art” is an over-the-top assumption about where generative AI leads humanity, I find it to be perfectly apt, if not a bit on the conservative side. It’s a blight and it has no business in gaming or anywhere else where a person could and should be paid for their unique vision, skill, and hard work. Graham’s talk was refreshing, but I fear that it’s a warning that won’t be heeded.
