I.

Throughout Capital and the Grundrisse, Marx makes clear the profoundly contradictory nature of technology under capitalism and what he calls the “civilising tendency of capital”. Capitalism, it can’t be denied, has seen profound technological advancements in machinery, transport, communication, medicine, etc. Capital, though, did not produce any of these advancements. Human knowledge and labour did. Capital, in search of a larger profit, simply organised this human knowledge and labour to do its scientific work to make and then put into actions these innovations—not because these innovations were good for society in some way, but because they would be profitable. The ‘civilising tendency’ of capital is a by-product, a happy accident, of capital’s endless thirst for profit. But it plays a critical historical role all the same.

A consequence of this is that workers, gradually over time, alongside their increasing exploitation by capital, gain access to more use-values, to more things. Vehicles, televisions, smartphones, videogames, medicine gradually shift over time (as their capitalist owners require more profit) from specialist things accessed only by a few elite, to things that that a much larger group of people are able to access. What was previously a luxury item becomes, over time, perceived as a necessary item all people deserve access to. Of course, the quality of the things we get reduce in quality at the same time as they are made under more standardised and cost-cutting conditions.

Again, none of this is the purpose of capitalism. The purpose is to exploit labour to get rich. But along the way, technology is rapidly advanced and social needs are greatly expanded (albeit certainly not equally for everyone at the same pace). And so the glut of those new things we have available to us also come with other material harms that aren’t resolved because it is not profitable to do so. The extreme climate cost of air travel, while we all benefit from internationally shipping cheap products and going on international holidays, is a perfect example here. Immediately here, hopefully, is visible what we all know to be the profoundly contradictory nature of using any technology under capitalism: we have access to all sorts of incredibly valuable use-values that make our day-to-day lives easier, and almost all of them were made by horrible people with terrible politics that are making the world a worse place.

In this context, the Marxist writer Ernest Mandel in his 1972 book Late Capitalism has a chapter critiquing the rise of ‘consumer culture’ and the constant new stuff that people are encouraged to buy by capital. He is, of course, critical of this consumer culture as a product of capital’s requirement to make more profit. But he also provides the following warning to anti-capitalist activists (across pages 394 to 396 if you’re interested):

“Any rejection of the so-called ‘consumer society’ which moves beyond justified condemnation of the commercialization and dehumanization of consumption by capitalism to attack the historical extension of needs and consumption in general… turns back the clock from scientific to Utopian socialism and from historical materialism to idealism…

For socialists, rejection of capitalist ‘consumer society’ can therefore never imply rejection of the extension and differentiation of needs as a whole, or any return to the primitive natural state of these needs; their aim is necessarily the development of a ‘rich individuality’ for the whole of mankind. In this rational Marxist sense, rejection of capitalist ‘consumer society’ can only mean: rejection of all those forms of consumption and of production which continue to restrict man’s development, making it narrow and one-sided. This rational rejection seeks to reverse the relationship between the production of goods and human labour, which is determined by the commodity form under capitalism, so that henceforth the main goal of economic activity is not the maximum production of things and the maximum private profit for each individual unit of production (factory or company), but the optimum self-activity of the individual person. The production of goods must be subordinated to this goal, which means the elimination of forms of production and labour which damage human health and man’s natural environment, even if they are ‘profitable’ in isolation. At the same time, it must be remembered that man as a material being with material needs cannot achieve the full development of a ‘rich individuality’ through asceticism, self-castigation and artificial self-limitation, but only through the rational development of his consumption, consciously controlled and consciously (i.e., democratically) subordinated to his collective interests.”

These passages had a profound impact on my politics when I first read them several years ago. My interpretation of it is, essentially, that any attempt to move society beyond capitalism cannot hope to ask the workers of that society to accept the fulfilment of a smaller range of needs than they currently enjoy under capitalism. We do not defeat capitalism by “turning back the clock”, by a life of abstinence and “artificial self-limitation” but by moving beyond capitalism and fully developing its “civilising tendency” beyond a mere byproduct of the profit motive into the actual motive force of our collective society focused on the collective good. Mandel goes on (on page 401):

The services which correspond to these [expanded] needs are not exclusively tied to the specific form of capitalist production and exchange, and will not in fact be able to develop fully before the capitalist mode of production has been overthrown. Admittedly, both the commercial nature of these services, which are geared to make a private profit, and their content, will undergo a fundamental change: instead of manipulating and alienating real human needs, they will be subordinated to them.

The smartphone, again, provides I think a useful example. It is valuable and good, socially, for everyone to own a smartphone. But perhaps we don’t need an almost identical new model every year when Apple decides to ensure last year’s model is no longer supported just to ensure enough people are still buying smartphones. A “fundamental change” to the technology of the smartphone after capitalism might be the ability to actually repair it, or the possibility of it still working a decade after you bought it, or it not being produced by slave labour. That’s the distinction between use-value (having a smartphone) and exchange-value (needing to sell as many smartphones as possible). Human ingenuity, not capital, created the smartphone, and we can have smartphones, in some (better) form, without capital.

II.

These passages, and my Marxist reading in recent years, has inevitably challenged my own thinking around videogames and generative AI, both separately and together.  I’ve realised, broadly speaking, I disagree with the sentiment from the popular meme that I want worse-looking games that are shorter and made by developers who are paid more. As an emotional counter-positioning it is clear and admirable but it falls into the abstinence trap. The videogames we have after capitalism won’t all need to be small scale hobbyist projects. We can still have bombastic and large and silly and spectacular triple-a games if enough game developers collectively decide to work together to make such games. Indeed, such games will be easier to make without CEOs and private equity syphoning off resources for their executive salaries that could instead fund the actual game and its workers. I want games of all scales with all sorts of visual styles made by game developers who are paid more and I’m not kidding.

And then, at last, we come to generative AI. Generative AI is a range of technologies made by evil corporations, using mostly stolen intellectual labour, doing extreme environmental and social harm, and putting countless game developers and tech workers (among others) out of work. On top of all this, the claims as to what generative AI can actually do are greatly inflated, or simply falsified, in order to prematurely sell their inevitability to investors, policymakers, and my university’s managers. In such a context, it seems like it must be impossible to talk about generative AI within either Marx’s “civilising tendency of capital” or Mandel’s notion of an expansion of rational needs.

But I’m always deeply suspicious of any attempt to situate a technology outside of history. My whole academic career to date has, in part, been about resisting narratives of videogames as magically totally different from every media form that came before them. I am likewise skeptical that generative AI represents a profound cut from history. That somehow, unlike every previous contradictory technological innovation of the past couple centuries of capitalism, generative AI is just simply ‘evil’. Instead, I would suggest it represents the same profound contradictions of technological innovation under capitalism as all the previous ones, just now on an even larger scale. The potentially useful byproducts are larger, as is the possible profits that are the actual goal and the undeniable harms.

This is not where I talk about what generative AI is actually good for and actually is bad for. I’m not enough of an expert on it for that. But I do think it is worth noting that the discussions we have around generative AI in game development are clearly repeating patterns we have seen with previous technological innovations. That all games that use generative AI assets look soulless and the same is not dissimilar to the anxieties less than a decade ago about ‘Unity games’ and the prominence of store bought assets.  That vibecoding disrespects and trivialises the important creative skills of the programmer is, in form, not dissimilar from previous discourses that dismissed games by people who figured out how to make games without any coding skills or tools. That it’s now too easy to make a game without actually knowing how to make a game is leading to an overproduction crisis that’s making the actually good games difficult to find, is not dissimilar to the anxieties around the Indiepocalype a decade ago by those indie developers who were lucky enough to get their foot in the door first.

None of these comparisons are to say the specific anxieties or critiques around generative AI use are unfounded or even that I disagree with them, and certainly not that they are debates not worth having (they are!). The negative impacts this time around are undeniably on a larger scale. Rather it’s to point out the same patterns: the same contesting over which skills and tools of gamemaking are legitimate and which ones are not legitimate. All of these debates are important and worth having. They also demonstrate that generative AI exists within history, as a continuing spiralling outwards of the contradictions of technology use under capitalism. We had these debates with commercial game engines. We’re having these debates with generative AI. We will have these debates with a new automating technology in another ten years.

All of which is essentially to say that I think any robust critique of generative AI use in game development (or indeed any field) needs to be situated within history and not outside of history. It needs to recognise when its critiques of generative AI are actually just deflected critiques of the capitalist relations of technology and labour. It needs to consider what is a criticism of the ‘technology itself’ and what is a criticism of how this technology is used and shaped by capital to exploit labour and increase profit at any cost—the way that capital has always used every technology available to it. Once this distinction is clarified, then a conversation about the actual potential uses of generative AI (if indeed there are any) can be had.

III.

To get overly speculative for a moment, I can readily imagine a post-capitalist world where a collective society has democratically determined specific ways in which this technology is convenient or useful and worth an allotment of that society’s energy consumption. In this utopia, no one’s IP has been ‘stolen’ because all human knowledge is owned by everyone, and no one needs to protect their ‘own’ intellectual ‘property’ just to be able to obtain food and housing.

Of course, such speculation about the ‘use’ of a post-capitalist generative AI or how it might contribute to ‘rational needs’ is not to deny the extreme damage it is doing right now to individual livelihoods and whole societies for those who do live under capitalism and do need to sell their labour.  But I think it does helpfully demonstrate how technologies and their use by capital can be disaggregated.

IV.

The last month or so has seen several well-respected indie developers publicly decried for disclosing highly limited involvement of generative AI in their game production process. Here I am particularly thinking of Molleindustria’s disclosure for Future? No Thanks! here, and of course House House’s disclosure to the IGF elaborated on Aftermath here. From the get go here I should disclose, too, that I would consider myself pretty good friends with the House House team, and I am named in the credits of Big Walk for playtesting it back in 2023.

Personally, I see both of these disclosures as incredibly thoughtful, critical, and reflective considerations of the contradictions of technology under capitalism. Each exists not because a particular indie darling has betrayed everyone. But rather because these indie developers have done precisely what they are well-respected for: thinking long and critically about the politics of their gamemaking. I think the emotional and puritanical response to each is entirely misplaced and harmful even as it is perhaps understandable.

Molleindustria discusses using AI generated content in the form of deliberately bad voices for two AI characters and, originally, a short music loop that was replaced before the full game was completed. Across the two statements, this use I feel most conflicted with personally as it feels closest to actual content generation. Nonetheless, I appreciated seeing the thought process of each of these creative decisions and the disclosure especially of early work that didn’t make it into the completed game. If only all disclosures were so exhaustive and thoughtful about the entire start-to-finish process.

I’m more interested here, though, in the claims of both Molleindustria and House House of using LLM chatbots sparingly to solve specific technical coding issues by receiving suggested script solutions that were then manually modified and integrated into the game. To conflate this with vibecoding a game is, I think, deeply unhelpful for a critical discussion of generative AI and game development.

In each case, generative AI has had, for the team, a clear use-value in helping them to successfully address a technical problem. Two responses I’ve seen to this in terms of House House especially is that, sure, it helped, but they could have solved the same problem without using an LLM chatbot. Some went and solved the problem themselves in Unity; others noted they could have hired consultants. Both responses are effectively saying that, yes, here generative AI has a real use-value, but it is not worth the social cost to access that use-value.

The first response (I was able to fix this myself) returns to the same cycles of skill hierarchies as previous debates around legitimate ways to make a videogame. I think it’s entirely irrelevant if someone else was able to solve the problem without asking a chatbot anymore than I think it is irrelevant if someone else was able to solve the problem without googling. This developer with this skillset needed to find a solution. An LLM chatbot helped them find that solution by synthesising sources it found.

The second response (they could have hired a consultant) I am more sympathetic to in some ways. But here, too, I come back to the fact that the history of game development, indeed the history of all technological innovation under capitalism, is figuring out ways to do things without having to pay someone else to do it. The rise of Unity meant development teams no longer needed to pay a software engineer to develop an engine for them. The rise of asset stores means development teams had alternative to paying animators and musicians. The rise of digital storefronts means development teams could stop paying for physical copies of their games to be manufactured. Speedtree means you don’t need to pay someone to individually model every tree in your forest. It is still unconvincing to me that instead of solving a problem the easier and fast way, a team should solve it the expensive and slow way.

What Molleindustria and House House have each done is what game development teams operating under capitalism have always done: using the available morally-compromised technologies of capitalism to help the current team overcome problems so that no one else needs to be hired. What they have done differently to most previous teams historically is dwell on how those morally-compromised technologies of capitalism are indeed morally compromised and been transparent about their own complicity in capital’s own reproduction.

V.

So what’s my point here? Any generative AI use is fine if you are open about it when someone eventually asks? That is certainly not something I believe. I’m deeply critical of generative AI and, I think like many of us, I still have that moment of shock and sadness when a colleague casually mentions asking ChatGPT something. I would still not voluntarily choose to play a game that has actually AI-generated content. Rather, my discomfort with how each situation has unfolded is that I think if the only generative AI disclosure statement that is considered morally acceptable is “No generative AI was used in making this game” then there’s no point in having any generative AI disclosure statements at all. Even ignoring the fact that such a situation inevitably leads to lying and dishonesty, even among honest developers this leads to drastically different responses and different interpretations of what it means to ‘use’ generative AI ‘in making’ a game. I don’t doubt for a second that countless beloved games with disclosure claims of no AI use have done precisely what House House and Molleindustria have done in terms of turning to an AI chatbot with a query. I don’t think those teams are lying; I think they are simply not being as self-critical and open with themselves or their communities as House House and Molleindustria are. It’s worth stressing that both House House and Molleindustria answered Steam’s specific disclosure requirements correctly. Both became an issue publicly when they chose to disclose more than was strictly required by the IGF’s much more expansive (and from the sounds of it vaguer) disclosure requirements as detailed by Mike Cook here.

You could argue that this is a great argument for more expansive disclosure requirements. I don’t disagree. Though, I would instead say it’s a great argument for more precise disclosure requirements that foster more consistent responses and reflection.

As a university lecturer who greatly resists any use of generative AI in my teaching content or used by my students, I quickly learned that a unit AI policy of ’No AI is to be used in this course’ is deeply counter-productive as much as I would indeed like that to be the case. The good students trying their best panic when the university library automatically provides them an AI summary of the book they looked up. The less attentive students still use AI and just hide it from me, making it impossible to discuss their education or how to use it critically. Other students just simply have no idea when they have or haven’t used a generative AI tool in the first place. Instead, by asking students to disclose how they use different software and being clear that, sure, you can use a chatbot to find your sources if you really must but you better not use it to write your essay draft, allows me as a teacher to have more robust and honest discussions with my students that ensures that even if they use a generative AI tool, they are at least learning something from this class.

I would suggest generative AI ‘use’ in game development urgently requires its own disaggregation of different kinds of use and just which of these are considered an illegitimate way to approach making a videogame. When is AI being used to solve a real problem, and when is it being used to make more profit? It’s not up to me or for anyone else to individually decide where that line is. It’s also not necessarily inappropriate for that line to indeed be ‘all generative AI use’. Indeed, it’s probably critical for a vanguard of people to indeed have that view and to be pulling in that direction. For my part, though, if we are to have honest discussions and transparency and reflection about how this capitalist-complicit technology actually is and actually is not useful for game developers alongside all the other capitalist-complicit technologies that actually are and actually are not useful for game developers, that requires being able to have a more nuanced conversation about what labour-saving activities are acceptable for what reasons and at what cost. This requires some effort to seperate out the use-value of generative AI (the real use-value, not the ones pedalled by the capitalists trying to sell them) from what is the particular ways generative AI is developed and harnessed by capital to produce private profit. It requires, I think, disclosure requirements from platforms and awards that perhaps are more expansive but that certainly are more specific in just what is meant by ‘using’ generative AI—ideally in ways that encourages developers not simply to fess up but to do what Molleindustria and House House did, which was to reflect on what indeed was the ways in which generative AI related to their projects.

Once we do that, I would hope it appears misdirected to use our time attacking individual developers who acknowledge in detail their own complicity in the capitalist system we all exist within instead of the capital relations and indeed the capitalists that are actually responsible for the exploitation of both workers and the planet beyond any ‘rational need’.

I have an essay over at Overland on the Arenanet firings. I don’t focus on the (boring, pointless) argument as to whether or not Price was ‘right’ to be mildly rude to some streamer on the internet, but instead focus on the gendered labour issues of how gamers always side with the corporations and bosses against the workers, and how this clashes with increased expectations that developers put work into their social media presence.

Lots of other good publications on this issue, too. Polygon spoke to Price directly about the firings. Game Workers Unity put out a blunt statement. This Verge piece was some good early reporting that helped get things moving. Up until and during Gamergate, the videogame press was never very good at dealing adequately with labour issues and worker rights in the videogame industry. That’s because ‘the enthusiast press’ has historically existed to serve consumers, not workers. It’s been a lot of hard work to see this change, and it’s come too late to protect a lot of people, but it’s good to see all the same.