The other day, a browser game was shared to Hacker News in a Show HN. Within a few hours someone had decided they didn’t like some mechanic in the game and built and shared a new version of the same game, with the mechanic tweaked. It seems nobody found this even remotely problematic.
When almost any programmatic system can be trivially copied, why share?
There is still human labour involved in building things, and for most systems and especially for games, it’s still a lot of the boring parts: determining spec, discovering edge cases you hadn’t thought about, and for games, a ton of playtesting to see if it’s actually even any fun. LLMs are still not that great at knowing what’s fun and they certainly don’t feel all the tiny micro-adjustments you have to make to have your game feel as fun as possible.
Of course, it’s still a lot less labour than ever before. I’ve built (or rather, half-built to validate design) over a dozen games since November - easily more than I’d ever built before in my 15-year career, and I went to college for game development.
(To some extent, I find that games are among the least ennui-inducing things to build with LLMs because a human still has to be firmly in the loop to see if the game is actually fun. And now I really get to focus on the actually interesting stuff - game design - rather than the typing of the code, the endless refactoring when I want to trial some variant, the constant reading of docs of the game engine and so on.)
Anyway, back to the topic at hand. Why share any programmatic system, if it can be trivially copied, tweaked, modified by anyone in a matter of minutes?
Let’s say I’ve been inspired to build a game by the Aubrey-Maturin series. I’ve read 20 books, I’ve thought about how to make it fun for months. I finally sit down with Claude for a few days or weeks building the game, Great Guns. I buy a domain name, I put it on itch.io, whatever. I put a post together detailing the game, my ideas, and I share it on HN. A few hours later, in the comments, the top comment is something like “I didn’t like the way ship movement worked so I built a new version where they turn faster. It only took Claude 35 minutes! Try it here.”, and there are comments like “I like this version of the game more, keep at it!” and “This version is better than OP’s” (verbatim comments from the original post).
To me, pretty infuriating, pretty depressing.
I dunno. Maybe it’s a bit rich to have this reaction when I use LLMs myself all the time. Maybe this is how authors of original works that got hoovered up into LLM training data feel. Maybe this is a fruit of a poisonous tree thing - because of my choice to use LLMs to build things, I have no right to complain when others copy what I built.
I’ve been thinking a lot about the meditation of the artist, and the concept of the ‘aura’ of works by humans as described in The Work of Art in the Age of Mechanical Reproduction by Walter Benjamin.
In Mechanical Reproduction, the ‘aura’ isn’t directly explained. Wikipedia suggests “uniqueness” but I’m also not sure if that’s really accurate either. Here’s how it’s introduced in the original:
The authenticity of a thing is the essence of all that is transmissible from its beginning, ranging from its substantive duration to its testimony to the history which it has experienced. Since the historical testimony rests on the authenticity, the former, too, is jeopardized by reproduction when substantive duration ceases to matter. And what is really jeopardized when the historical testimony is affected is the authority of the object
One might subsume the eliminated element in the term “aura” and go on to say: that which withers in the age of mechanical reproduction is the aura of the work of art. This is a symptomatic process whose significance points beyond the realm of art. One might generalize by saying: the technique of reproduction detaches the reproduced object from the domain of tradition. By making many reproductions it substitutes a plurality of copies for a unique existence. And in permitting the reproduction to meet the beholder or listener in his own particular situation, it reactivates the object reproduced. These two processes lead to a tremendous shattering of tradition which is the obverse of the contemporary crisis and renewal of mankind.
(Benjamin was primarily concerned with film, in Mechanical Reproduction. Now, I would say that film has probably ascended from primarily propaganda tool (as it certainly was in his time) to aura-possessing work - though I think it’s quite possible that work can be both propaganda and aura-possessing.)
With apologies to any actual artists reading this who know much better, to me, the ‘aura’ of a work comes partially from the meditation of the artist - the time and human effort spent in causing the work to exist. Thinking about the work, conceptualising the work, placing the work in context, considering, crafting, procrastinating, making mistakes, fixing mistakes - not to mention all the time spent learning how to create in the first place.
I think that software had a lot of time for the meditation of the artist. So much time spent thinking about the right way to do things, so much time spent refactoring, rewriting, improving, and so on. Can software have an aura, intrinsically, independent of the purpose of the software? Does the aura of the software have an impact on the aura of the purpose of the software?
I suppose aura is partially why I find games so interesting, even in the age of AI. There is still a large amount of human effort involved in game design, finding fun, storytelling, designing mechanics and so on, and I think there’s a lot of time spent away from keyboard musing about those things, and I think that time is a source of ‘aura’.
Perhaps this is why I had the reaction I had upon seeing those comments. Do drive-by LLM remixes (“Rebuild this but with X in place of Y”) have aura, and can you tell the difference? Why share something in the first place if the aura of the work you spent time creating can be stripped away or diluted by online commentators in a matter of moments to suit their taste? And most fundamentally - can work created with LLMs even be considered to have an aura, and, to what degree can LLMs be used before the lack of aura is palpable?
Further reading on this blog: Releasing Software Now and maybe State of the Art, and of course far more substantial and worthwhile would be The Work of Art in the Age of Mechanical Reproduction by Walter Benjamin.