The last few weeks have not been kind to No Man’s Sky. After years of build-up gained it a massive following powered by ever-ballooning hopes, launch day proved it was not only far shallower than fans had expected, but also kind of boring. The situation has since gone full coke-and-mentos as negative reviews pile up and players call for refunds en masse.
I’m experiencing much of this from a distance, because I never had a personal interest in No Man’s Sky. When asked I’ve said it’s down to motivation – I need specific goals when I’m playing a game, which even the best resource-gathering titles (e.g. Don’t Starve) often lack. But the longer I look at the No Man’s Sky situation, the more I realize the thing that repelled me is something else, something much more intrinsic to the game as a whole: procedural generation. Because that shit doesn’t work.
I don’t mean that in a literal, nuts-and-bolts way. Procedural generation is great at taking a list of components and arranging them according to an algorithm, creating different designs from those pieces ad infinitum. Because it works so well on a mechanical level, the idea of building a game with it has been incredibly popular, working off the promise that it would let players explore an infinite world and discover fascinating things even the game’s developers have never seen. It essentially frees players from developers’ expectations and desires, letting them see more than just what the game’s creators want them to see.
That vision is far sexier than the reality however. While the level of variety in procedurally generated worlds seems dramatic at first – No Man’s Sky strange creatures look wildly, delightfully odd for a while – over time the obvious similarities between different setpieces start to show through. ‘Quadraped that walks like a deer’ becomes a defining trait for dozens of No Man’s Sky’s creatures within hours, and ‘Quadraped that walks like a deer with a tiny head’ and ‘Quadraped that walks like a deer plus tentacles’ quickly start to bleed together. Even worse is when your possible actions in that world are dictated by procedural generation – you can only harvest so many plants, hunt so many animals, and explore so many slightly different places before the whole experience gets utterly dull.
Worlds created by procedural generation inherently lack the unique treasures and sights that make exploring fun, in the real world or a digital one. I can still remember the adventures I had as a kid in the forests near my hometown, walking miles to see hidden waterfalls, finding a secret path to a secluded beach, or marveling at trees that are having none of your nonsense today. What I don’t remember is each rock that looked different from another rock, which is essentially all that procedural generation offers. It cannot, by design, be any more than the sum of its parts.
In contrast, games that have more of a human touch do a much better job of offering up one-of-a-kind experiences, because the human minds behind the scenes understand how to create them. That doesn’t have to mean overblown, show-stopping cutscenes either – years after playing BioShock I remember finding two skeletons embracing on a mattress with a bottle of pills beside them, and I still smile thinking about the singing vortigaunt I discovered hiding in a toxic waste pipe in Half-Life 2.
Those games are more confined and structured than No Man’s Sky, but they make exploration in that space genuinely rewarding. Even in games where story isn’t pivotal, clever additions like that can make the exploration feel worthwhile; climbing Felwinter Peak in Destiny is an exercise in saintly patience, but finding the last resting place of the person who got their first made it feel like I’d discovered an entirely new story to mull over as a reward (and it’s hardly Destiny’s only example).
These things all exist because human developers dreamed them up and put them in place. Even hemmed in by trends and biases, human imagination has more room to create unique additions to the world than an algorithm with a thousand different parts ever could. That isn’t to say there’s no place for procedural generation – open world games and MMO’s would take decades to design otherwise. However, after seeing what becomes of a fully procedural game like No Man’s Sky, there’s a case to be made for its judicious use alongside more targeted, human elements.
Think of The Witcher 3, with hundreds of unique interactions hidden in vast, algorithmically-enhanced environments, or the sprawling but carefully curated lands of World of Warcraft. Procedural generation isn’t the godsend that will let computers replace humans in the game-making process, and we shouldn’t want it to be. Instead, we should treat it like the tool it is, and use it a bit more humbly.