No Man's Sky

[Sir T]

Are people that desperate to name a species "Buttmuncher69"?

[schild]

I wasn't but I am now.

[zwohand]

Man. I haven't been this hyped for a game since vanilla WoW

[Reg]

I've been following it for months but I'm not confident that it's going to be any actual fun.

[Sky]

Been following this since the first teaser, expect it to be bland and generic like most 'procedural' games. Also expect adding 'game' will work about as well as it did for  Spore.

[Khaldun]

I am still sure that there is a way to do procedural content right. I think it's about using it in a limited fashion within an otherwise written/created game. Say, to have some NPCs who have procedurally-generated objectives or interests inside a game that's otherwise directly crafted by designers.

[apocrypha]

This launch is going to be very interesting, because it's already gone through the hype/disappointment cycle for a lot of people. 6 months ago the hype was intense, but like all highs it can't last forever, and coupled with the delays has led to a lot of the hype turning to an expectation of disappointment.

I've moved from "buy at launch if initial reviews are good" to "buy 6 months after launch if reviews are still good and continued dev support has been good".

[KallDrexx]

It's pretty hilarious watching all the people so sure that this game is going to be absolutely amazing already get into defensive mode, and talking about how reviewers and youtubers won't be reviewing it "the right way".

[Malakili]

This game is going to be breathtaking and awesome... for one evening at best.

[Khaldun]

I dunno, Elite didn't even last an evening for me, but maybe that's because I could barely figure out the controls and didn't even get to see too much of the awesome.

[01101010]

If I can audio record, "captain's log, stardate 2729.4. Landing on planet Gethafugoph III, I have encountered what appear to be native insects," then I might give this a whirl.

[Sky]

The problem with procedural generation is that it's largely meaningless without hand-crafted content. If you could put in rogue-like 'dungeon' creation, maybe. But in reality it seems to work out to 'this planet has purple trees and the last one had orange trees'. Look for npcs for generic kill/deliver/x quests. Look for obelisks to unlock tech.

Doesn't seem like much to do, outside of busy work and looking at different shuffles of the randomizer.

[Malakili]

Yep. That's precisely why I said it'll be great for about one evening. It'll be fantastic when you don't have a sense of the kinds of things that are in the game. For the first few planets you're likely to find new stuff that feels genuinely different. But you're going to notice the patterns pretty quickly. You'll get a sense for the kinds of things available to the randomizer, and then it's all downhill from there.

So far the only game that's really done a passable job to me is Minecraft.

[Yegolev]

I admit that I don't grok why a new, fresh Minecraft world is so exciting.  I'm not the guy who is going to invent the next big procedural game.

[Sky]

A few mods have really improved both worldgen and explorability. Unfortunately, being home-brew java, you were lucky to try it when it was a thing and not crash. The village system of Millenaire was a start, Better Dungeons added in a ton of (buggy) great content creation/config. I'm always looking for stuff that adds reasons to explore the world, and, you know, find interesting stuff.

[Khaldun]

I agree with all this. But I think procedural content *could* be the secret sauce in an otherwise hand-crafted open-world game setting.

Think about how certain kinds of random spawns in a game like Skyrim sometimes create a special kind of fun--the first time you see a dragon fighting with necromancers while you just watch the whole thing. But when procedural generation is the only thing going on, there's an unfocused, purposeless feeling to it all.

[apocrypha]

Procedural generation can't create any stories that are more than one layer deep. Random fights between random NPCs that don't like each other can be entertaining, sure, but there's nowhere further for them to go.

Compare that to a simple side-quest in The Witcher 3 where something as innocuous as a haunted well can reveal a take of love, betrayal, rape, murder, revenge. It's hard to generate any emotional involvement in a procedurally generated world.

What could work is a combination of hand-crafted content and over-arching story that uses proc.gen to 'fill in the gaps' but that's really just padding. The reason Minecraft works is because of the creative element - the player creates their own 'story' of sorts by virtue of the things they make within the world.

If NMS gives the player any ways in which to change the world it creates, then it'll maybe have some longevity. Otherwise it'll be Steam sale material that'll provide an entertaining distraction for a day or two.

[Malakili]

Going back to Elite, which Khaldun mentioned up the thread a few posts, I will say that some of the factional stuff and news feed does at least create SOME kind involvement. I think Frontier is doing a combination of procedural generation and manual nudging of the "story" in certain directions, which has been effective so far. It seems like thats a game that is really divisive. People either love it or think it's boring after 5 minutes. I'm somewhere between. I do get bored with it every so often, but I find myself wanting to dive back in every so often. It's done a good job filling my space sim needs since it came out.

[Kail]

Procedural generation can't create any stories that are more than one layer deep. Random fights between random NPCs that don't like each other can be entertaining, sure, but there's nowhere further for them to go.

Well, theoretically there is.  Procedurally generated doesn't mean "random" it means that the computer writes it according to certain rules.  If you have decent writers / coders who can put together decent rules for creating a story, then you could theoretically generate a lot of procedural content.  In the same way that you can procedurally generate terrain by modeling erosion and weather patterns, you could hypothetically procedurally generate social situations (meaning, stories) by modeling characters with actual motivations and abilities and so on.

[schild]

blah blah

I'm more hyped for this than nearly anything in the last 5 years.

[Sky]

This just in, schild vaping salts now.

[Khaldun]

I am always excited for someone to do this right because I believe it can be.

But I am always ready for someone to fuck it up because I believe they almost certainly will.

We just named another reason why the latter is likely: because procedural generation depends on the quality of the imagination going into generating the procedural ruleset. If it's a bunch of dumb coders trying to save themselves some trouble or suits trying to save on people making art assets, etc., you end up with a dull failure at best. A great procedural design would have to arise out of a sophisticated and complex vision of the world and character and story generation you're trying to pull off. Anybody remember the random tables in Dungeons and Dragons for generating a quick dungeon or adventure? The worst of them made you something like Daggerfall--the same fucking thing you'd done a million times. The best of them made you something only marginally better than that, *unless* the person interpreting it made something unusual out of it.

[apocrypha]

Well, theoretically there is.  Procedurally generated doesn't mean "random" it means that the computer writes it according to certain rules.  If you have decent writers / coders who can put together decent rules for creating a story, then you could theoretically generate a lot of procedural content.  In the same way that you can procedurally generate terrain by modeling erosion and weather patterns, you could hypothetically procedurally generate social situations (meaning, stories) by modeling characters with actual motivations and abilities and so on.

I'd really love for someone to succeed at that, it could be awesome. I suspect that really good AI may be a missing piece, and AI in games seems to have stagnated over the last 10 years.

I thought of another reason why procedural generation in Minecraft works so much better than in, for example, Elite Dangerous. The scale of the variation in terrain in Minecraft perfectly matches the scale at which you play the game. The shapes of hills, caves, plains directly affects how and what you build, how you interact with mobs, etc. Large builds require actual terraforming, sculpting of those randomly generated areas. The scale of E:D randomisation is meaningless. A system has more or less planets, which are of varying colours with various minerals on them. You still warp from jump-in point to station in exactly the same way, engage with pirates in exactly the same way, hand in missions in exactly the same way. The procedural generation is purely cosmetic.

If planets and systems in NMS vary in such a way that the actual gameplay is affected by it then it'll be better than E:D.

[Hawkbit]

The aesthetics sold me from the start. I've thrown $60 at some really shitty games, so I won't feel too burned if it's bad.

I'm really interested to see how they handle the "quest" stuff; there's an overarching quest at least to get from x to y. I'm also very interested to see how they push this out with only a 6gig master.

Cautiously hyped and I figure I'll get at least a few weeks of fun exploring.

[Sky]

I'd really love for someone to succeed at that, it could be awesome. I suspect that really good AI may be a missing piece, and AI in games seems to have stagnated over the last 10 years.

I've been saying this since the days of Lum's old site. It's not 10 years. It's been a red-headed step child of the gaming industry. Not just 'good' AI, but convincingly inept AI as well. It's a laughable joke, with games just cheating or making damage soaks in place of 'difficulty' or making things a walk through the park in place of 'easy'. No ability to use advanced strategies or make newb mistakes. Just a simple AI that can barely do what it's supposed to do, while all the money is dumped into graphics and management.

Every time I mention it, nobody from the dev side ever seems to think there's an issue and everything's hunky fucking dory.

[Malakili]

It's also one of the reasons many of us have been attracted to sandbox PvP games, in spite of the fact that they've been consistently pretty terrible.

[Nija]

This game is a win/win for me. Either I'm going to get a lot of enjoyment out of the reviews and the buzz around the game, or I'm going to get enjoyment out of a good game.

Can't go wrong. Proceed.

[Goreschach]

This game is a win/win for me. Either I'm going to get a lot of enjoyment out of the reviews and the buzz around the game, or I'm going to get enjoyment out of a good game.

That is a powerful base on which to build your life, Paul Muad'dib.

[Falconeer]

It's also one of the reasons many of us have been attracted to sandbox PvP games, in spite of the fact that they've been consistently pretty terrible.

This!

[Sophismata]

Every time I mention it, nobody from the dev side ever seems to think there's an issue and everything's hunky fucking dory.

I think the problem is two-fold.
The 90% don't want good AI because they are shit at playing against bad AI.
Good AI doesn't sell or make for flashy screenshots.

Sadly, for me, bad AI kills a game just as quickly as bad gameplay does.

[Kail]

There's also the issue that it's hugely dependent on the actual game.  What "good AI" even means changes a lot from title to title.  In Dota it probably means a program that can react strategically to what the other team is doing and cooperate with it's teammates, in Skyrim it probably means a program that can react like a real person would to someone doing things like running though town naked and throwing cheeses at everybody.  It's not something you can transfer from one program into another, like a new rendering method or chat system.  It's a problem every team has to solve from scratch, almost.

[Sir T]

When "the 10%" say they want good AI, they really mean something that will exclusively chase the guy with the highest toughness and who will ignore the healers. 

[apocrypha]

When "the 10%" say they want good AI, they really mean something that will exclusively chase the guy with the highest toughness and who will ignore the healers. 

Remember WoWs Faction Champions fight in Trial of the Crusader? It wasn't exactly amazing AI, but the NPCs absolutely targeted healers and squishy DPS, and maaaaaan did people bitch about that fight and fuck it up over and over again.

[Sky]

Every time I mention it, nobody from the dev side ever seems to think there's an issue and everything's hunky fucking dory.

I think the problem is two-fold.
The 90% don't want good AI because they are shit at playing against bad AI.
Good AI doesn't sell or make for flashy screenshots.

Sadly, for me, bad AI kills a game just as quickly as bad gameplay does.

You misunderstand what I mean by 'good AI'. It's the most common response I've seen over the years, so you're in good company.

Good AI doesn't just mean killbotX destructo mode, know all the tricks and use them. That's there at the top end for the players who can easily destroy any paltry AI routines and want a challenge. Pawning this off on other customers is the preferred cop-out ("multiplayer"). Not a dig against multiplayer, but folks should play against others because they want to, not because your game's AI sucks.

The other side of good AI is 'good bad AI'. This is for the people that either suck against current primitive AI or (like me) just want to play to relax and not necessarily be challenged (I like AC 3 more for the setting than the crappy combat, for instance). "Good bad AI" makes believable mistakes, as if you're playing against another novice human or even a talented human who could probably play good but doesn't (like me :p). By the latter, I mean an AI that doesn't exploit full knowledge of the map and character build/itemization, but otherwise plays fairly competently. Also to react humanistically: sure, it may make sense to follow that healer you're targeting, but when a dps blasts you, maybe you turn to fight them instead. AI needs to believably make mistakes as a difficulty scale, removing mistake conditions as you increase difficulty.

[Sky]

There's also the issue that it's hugely dependent on the actual game.  What "good AI" even means changes a lot from title to title.  In Dota it probably means a program that can react strategically to what the other team is doing and cooperate with it's teammates, in Skyrim it probably means a program that can react like a real person would to someone doing things like running though town naked and throwing cheeses at everybody.  It's not something you can transfer from one program into another, like a new rendering method or chat system.  It's a problem every team has to solve from scratch, almost.

I agree with this, though probably at a genre level there are enough commonalities to make a start and then tailor it, as you would an off-the-shelf engine. Heck, maybe that's what it needs.