Another One Bites the Dust: SWG Edition!

[Malakili]

So, on a completely other topic, this thread has been getting popular again at the same time my friend has been playing a lot of The Repopulation.  I ended up picking up the early access this weekend.  It's very much SWG inspired and I'm having a good time with it so far, but I'm not sure how much of that is dependent on the game resembling SWG and therefore evoking nostalgic feelings.  Thing is janky as hell, has a really complex skill system, and a pretty big game world including city/housing building.

[Merusk]

It is, but it's called Landmark, not EQN. Seems very clear that EQN is dead.

Yeah, I have no doubts EQN is dead with the buyout/ company change. I'm not sure Landmark is going to survive, either, but wasn't sure how old the Smed quote being referenced was.

[Paelos]

Think EA's accountants would disagree with you regarding the importance of turning gaming IP into serial franchises.

ATVI is a better example. They print money off their basically 5-6 franchises.

But things are not perfect there either. Revenues were $175M lower in 2014 than the year before, and about $350M lower than 2011. They are the biggest, baddest kid on the block, raking in $4B a year in sales, and even they are seeing slow declines.

[Gimfain]

All this reminds me,

Does anyone know what Smed meant when he recently(?) said that SWG players would be able to come home soon?

I can only think he meant Landmark (which is a shame)?

H1Z1 was the game that SWG players could return to. The question is whether he was actually serious about the connection, or if it was smed trolling people again.

[schild]

Real programmers work in gaming, they just demand the salary they'd get in enterprise.

There are not many that are that good.

Holy shit I hadn't clicked on this thread in a while. 

[Raph]

Latest SWG nostalgia post is on how the worlbuildng worked, and didn't, and what it enabled.

http://www.raphkoster.com/2015/04/20/swgs-dynamic-world/

[taolurker]

Tell us about Voxels next Uncle Raph.

[Raph]

heh. The step after this was actually using cellular automata for world simulation. Not voxels like Minecraft, but not dissimilar. We got water cycles, rivers flowing, snow accumulation that you could actually push to one side, animals leaving scent and predators tracking, spells that could melt holes in the surface of ice lakes, and a bunch more.

[taolurker]

LOL

I was more wanting to see the Developer Diary on how getting a procedurally generated world with Voxel Farm was going with your new employer.. As much as I'm enjoying the nostalgia, and understanding more of what went into SWG, I would kind of like to see what you're capable of doing to not make Voxel Farm worlds get bloated performance wise (because if Landmark is any indication it's gonns be an "uphill" battle).

[Paelos]

I would think the problem wouldn't be as large simply because Crowfall isn't using Voxels to let people build things from scratch.

[Sophismata]

I want an AI that's better than me so I'm able to improve. Fighting something worse than yourself leads to complacency.

We all think we want this on some level.  Just as long as we can still, you know, sneak up on that guard and sever his throat without him ever spotting us.  Or just as long as we don't aggro anything more than 5 meters away from us.  Or just as long as we can run away, break aggro, and come back and find that same guard still at his post.

We don't actually want AI that is better than us.  You would by definition never win that game.  Imagine playing Demon Souls where all the AI enemies were replaced by Schild, who is probably better than you at that game.  No fucking thanks.

O_o

I completely disagree. I *know* I want AI that's better than me for the very reasons outlined above - it betters you as a player. It's the same in any game of skill - competition is what forces us to improve. As humans we have a capacity to grow that the AI does not. And if you are able to execute on that the victory is far more meaningful.

And in the case where you're not talking about something static - an AI that also learns as it plays you. Then bring it on. I relish the challenge. No-one achieves anything worthwhile by aiming for second place.


Edit: Just to be absolutely clear, the best example of this is a FPS where you can have a theoretically perfect AI. 100% accuracy, always knows where you are. I would be happy to fight that opponent. In order to win I'd have to either progress to the same point, or trick / take advantage of it, and I am comfortable with that.

[Malakili]

Edit: Just to be absolutely clear, the best example of this is a FPS where you can have a theoretically perfect AI. 100% accuracy, always knows where you are. I would be happy to fight that opponent. In order to win I'd have to either progress to the same point, or trick / take advantage of it, and I am comfortable with that.

Quake 3 had this AI, more or less, and it wasn't interesting at all to play against.  Beating AI like that is about the "tricking" it, finding out the one thing that it doesn't quite get, and exploiting that because, as you said, it can't learn.  I mean, I guess if you REALLY like that kind of puzzle that's an ok way to play an FPS, but it's really not very interesting in my opinion.  Compare that style of play with actual top level Quake 3/Quake Live play and it's not even close.

[Sophismata]

Edit: Just to be absolutely clear, the best example of this is a FPS where you can have a theoretically perfect AI. 100% accuracy, always knows where you are. I would be happy to fight that opponent. In order to win I'd have to either progress to the same point, or trick / take advantage of it, and I am comfortable with that.

Quake 3 had this AI, more or less, and it wasn't interesting at all to play against.  Beating AI like that is about the "tricking" it, finding out the one thing that it doesn't quite get, and exploiting that because, as you said, it can't learn.  I mean, I guess if you REALLY like that kind of puzzle that's an ok way to play an FPS, but it's really not very interesting in my opinion.  Compare that style of play with actual top level Quake 3/Quake Live play and it's not even close.

In general, it's more satisfying to fight against a human opponent because of the emergent tactics and mind games involved. AI in shooters is, in general, very lacklustre compared to a real opponent (as with many other genres).

I actually enjoy the puzzle element of beating that kind of AI, though - but I agree that it ultimately isn't that interesting to fight. We need better AI in general, as this is also the primary complaint I have with many other games. Mass Effect and Dragon Age both suffered from that issue, PoE handles it a bit better (unless you play PoE solo). Forged Alliance on the hardest difficulty had certain tactics the computer just couldn't adapt to.

Hmmm. Maybe what I really want is just more inclusive multiplayer (iterating further on the Souls games' phantoms idea). However I definitely do want pandering in the sense of "an AI that is kind of hard but not too hard so that it is beatable". Make it as hard as you can - the Deep Blue equivalent of that game.

[Raph]

LOL

I was more wanting to see the Developer Diary on how getting a procedurally generated world with Voxel Farm was going with your new employer.. As much as I'm enjoying the nostalgia, and understanding more of what went into SWG, I would kind of like to see what you're capable of doing to not make Voxel Farm worlds get bloated performance wise (because if Landmark is any indication it's gonns be an "uphill" battle).

That would be hard, given that I am self-employed, and doing a bunch of different stuff at once for different people. Consulting on Crowfall is only one of them. :)

[Morat20]

You're gonna have to define "better" in "better AI".

The simplest definition of "better AI" is "AI's that can defeat increasingly skilled people at the task at hand".

It's not hard to make AI that slaughters human opponents in a FPS or RTS. Being better than humans isn't really a hard task. As noted, people complain when fighting unbeatable AI opponents.

In fact, they complain when facing AI opponents using simple tactics because, in general, two or three AI's using basic tactics beats one dude in an FPS.

So what's a "better AI"?

So far "so good you have to figure out an exploit the programmer didn't cover" seems to be your definition, which most people would take as "So OP I had to cheat" and consider that crappy AI.

[Paelos]

Welp I learned we can hand Sophismata a hammer for his nuts and he's a happy camper.

[Senses]

Someone needs to make an MMORPG that uses AI to be all the other people so each person can have their individualized MMO full of people that worship them and their awesome gear.

[Sophismata]

You're gonna have to define "better" in "better AI".

The simplest definition of "better AI" is "AI's that can defeat increasingly skilled people at the task at hand".

It's not hard to make AI that slaughters human opponents in a FPS or RTS. Being better than humans isn't really a hard task. As noted, people complain when fighting unbeatable AI opponents.

In fact, they complain when facing AI opponents using simple tactics because, in general, two or three AI's using basic tactics beats one dude in an FPS.

So what's a "better AI"?

So far "so good you have to figure out an exploit the programmer didn't cover" seems to be your definition, which most people would take as "So OP I had to cheat" and consider that crappy AI.

It's actually quite challenging to make an AI that slaughters human opponents in an RTS if that AI has to obey the same rules that the humans do. In fact that works for a large number of games (FPS being one of the notable exceptions).

For me, ideally it's an extremely good, artificial player. It follows all the same rules and constraints as a normal player, has access to the same resources but is strategically or tactically more competent.

That said - I don't mind the situation where the AI cheats to be practically insurmountable, but I view it as a distinct and separate challenge. Primarily because the skill-set required to default cheating AI is different from the skill-set required to defeat a challenging human opponent. It's not the same game, and is no excuse for a good "real" AI (ie one that follows the same gameplay rules as the player).

[Falconeer]

Hilariously enough, when Smed said that line about SWG fans having a new home soon, he meant H1Z1. No kidding. Google all that if you don't believe me.

[Ironwood]

If you are someone who should not be reading technology patents, you should stop now.

Was this a joke or something that actually happens ?  Why ? 

I'm really confused.

[Merusk]

You're gonna have to define "better" in "better AI".

The simplest definition of "better AI" is "AI's that can defeat increasingly skilled people at the task at hand".

It's not hard to make AI that slaughters human opponents in a FPS or RTS. Being better than humans isn't really a hard task. As noted, people complain when fighting unbeatable AI opponents.

In fact, they complain when facing AI opponents using simple tactics because, in general, two or three AI's using basic tactics beats one dude in an FPS.

So what's a "better AI"?

So far "so good you have to figure out an exploit the programmer didn't cover" seems to be your definition, which most people would take as "So OP I had to cheat" and consider that crappy AI.

It's actually quite challenging to make an AI that slaughters human opponents in an RTS if that AI has to obey the same rules that the humans do. In fact that works for a large number of games (FPS being one of the notable exceptions).

For me, ideally it's an extremely good, artificial player. It follows all the same rules and constraints as a normal player, has access to the same resources but is strategically or tactically more competent.

That said - I don't mind the situation where the AI cheats to be practically insurmountable, but I view it as a distinct and separate challenge. Primarily because the skill-set required to default cheating AI is different from the skill-set required to defeat a challenging human opponent. It's not the same game, and is no excuse for a good "real" AI (ie one that follows the same gameplay rules as the player).

It's your own arrogance about being a leet player that leads you to think you want this. Computers will always faithfully execute and out-think humans within a defined set of rules. It's what we've built them to do.

Then, when you find that one situation that wasn't planned for suddenly it's ezmode and you'll be back here complaining that it's too easy because you found that singular exploit.

So the AI gets patched, and you have to find that next exploit, and the cycle continues.  You eventually run into the chess AI problem. Nobody takes on expert mode because it beats Grandmasters regularly.  It's much more entertaining to just ignore the game entirely.

Your 'pushes us to be better' standard is flawed because it doesn't matter. Your gaming skills translate into nothing but being better at games. Great if you're competitively gaming, but that's such a small percentage of the market as to be pointless to develop for.

We can build games to teach other skills, sure, but those don't sell as well. People are looking for entertainment, and most want that to be passive entertainment. "I want to turn off my brain" stuff because humans are fucking lazy creatures. Sure the initial high of the challenge is great, but it gets boring and people move on.

The completion % for Dark Souls was 36% and while it's hard, its nowhere near the level of difficulty you're asking for. "Better AI" by your definition takes money out of dev & publisher's hands. It's not happening.

But we've had this argument before. You think it's pointless to not do things at 100% and be the best at it. You go, Ash Ketchum.

[Paelos]

When I say better AI, I mean AI that doesn't have to cheat to maintain competition against the player.

[Malakili]

It seems like the ideal that people want is AI that is functionally indistinguishable from a human opponent of relatively equal skill to them.  But even assuming that is possible, the standard single player experience has usually been 1 player vs. dozens or hundreds of enemies.  That only works when the AI isn't really challenging on an individual level.

[Paelos]

It seems like the ideal that people want is AI that is functionally indistinguishable from a human opponent of relatively equal skill to them.  But even assuming that is possible, the standard single player experience has usually been 1 player vs. dozens or hundreds of enemies.  That only works when the AI isn't really challenging on an individual level.

Not really. I want AI that doesn't make really stupid decisions. Example, the AI in Rome:TW would at times run a siege at one area of a wall with 1000 troops against your 200. This created a bottleneck you could win by stacking the breach, or pinning them with arrows. A better AI would attack two or more points, and stretch your defense to the breaking point. Later AI's in TW do that. The earlier one was just dumb.

Another example AI acting like players and getting hated? Trials of the Crusader in WoW. The Faction Champions encounter. That fight is probably one of the more hated raid fights in WoW's history. Why? Because they made the AI act like PvP players, and essentially showed how ridiculous the Trinity really is when bosses stop being polite and start getting real. Tanks were essentially useless, everyone had to go into pvp mode, and raiders who sucked at positioning with pvp intent were furious. Yet, that AI was pretty well designed. It acted like a player would. It would go after your healers, it would ignore the tanks, CC became very key, so did stuns, etc. And people despised it because it didn't fit into the mold of raiding they were used to. Also because it showed how terrible they were as players.

[Merusk]

Exactly, and the PVP AI was terrible. They toned it down from what it was because it wasn't getting beat.

HIgh-Functioning AI will never miss a cooldown. Never misclick. Always click the correct priority button. You are fucked against it. Give it the ability to learn from mistakes and humans are pointless.

[sam, an eggplant]

Funny how some arguments are eternal. I first had this exact conversation in rec.games.computer.quake in like 1997, and many times since.

Nobody really wants smart AI. They want AI that doesn't blatantly cheat and doesn't make obviously dumb decisions, so when beaten the player's achievement isn't cheapened. That's it.

We play games to win. Constantly losing isn't fun. And yes, that applies to Dark Souls and its ilk too-- these games are tuned to be challenging, with pixel-tight controls rewarding player experience and skill. You improve as you play, and obstacles once seemingly insurmountable are reduced to trivialities. There's little random number generation bullshit. What you do matters. That's why we love them, because we get better, not because the AI is brilliant-- it isn't, at all.

[Paelos]

Nobody really wants smart AI. They want AI that doesn't blatantly cheat and doesn't make obviously dumb decisions, so when beaten the player's achievement isn't cheapened. That's it.

Exactly, it's why I found Civilization AI to be so abhorrent. People would brag about beating the game on so and so mode, and I'm thinking, congrats. You found the loopholes around a system that was actively cheating. It's not IMPROVED at higher levels, it simply cheats more.

[Malakili]

I think there is something to be said for AI that could match human-like play at various skill levels.  Something like DOTA or Starcraft would benefit a lot from being able to get the experience of playing with/against other people without needing the actual people. 

[Rendakor]

So you want AI that calls you a fag or goes AFK for whole games?

[HaemishM]

So you want AI that calls you a fag or goes AFK for whole games?

Those are the easy AI to program. We want AI that can really get at the core of shittardery. Things like posting your IP while threatening to find and kill you.

[Viin]

If you are someone who should not be reading technology patents, you should stop now.

Was this a joke or something that actually happens ?  Why ? 

I'm really confused.

Not a joke. Knowingly breaking a patent and accidentally breaking a patent make a big difference in court.

[Malakili]

It seems like the ideal that people want is AI that is functionally indistinguishable from a human opponent of relatively equal skill to them.  But even assuming that is possible, the standard single player experience has usually been 1 player vs. dozens or hundreds of enemies.  That only works when the AI isn't really challenging on an individual level.

Not really. I want AI that doesn't make really stupid decisions. Example, the AI in Rome:TW would at times run a siege at one area of a wall with 1000 troops against your 200. This created a bottleneck you could win by stacking the breach, or pinning them with arrows. A better AI would attack two or more points, and stretch your defense to the breaking point. Later AI's in TW do that. The earlier one was just dumb.

Another example AI acting like players and getting hated? Trials of the Crusader in WoW. The Faction Champions encounter. That fight is probably one of the more hated raid fights in WoW's history. Why? Because they made the AI act like PvP players, and essentially showed how ridiculous the Trinity really is when bosses stop being polite and start getting real. Tanks were essentially useless, everyone had to go into pvp mode, and raiders who sucked at positioning with pvp intent were furious. Yet, that AI was pretty well designed. It acted like a player would. It would go after your healers, it would ignore the tanks, CC became very key, so did stuns, etc. And people despised it because it didn't fit into the mold of raiding they were used to. Also because it showed how terrible they were as players.

That kind of thing is a whole different ball game.  It is barely AI in the sense I was discussing, it's a puzzle that 40 people (or 25 people, whatever) have to solve together.  But more to the point, I suppose what it really means is that different kinds of AI are appropriate for different games and probably different kinds of content within a game.

[Quinton]

To the tune of 3X damages, I think?

The advice I have always received from various legal departments, as a software engineer, is "do not search for or read patents" and let legal worry about them, because anything you do to "help" is likely to just increase liability somehow.

US IP law is teh suck.

[Torinak]

If you are someone who should not be reading technology patents, you should stop now.

Was this a joke or something that actually happens ?  Why ? 

I'm really confused.

In the US, there's the notion of suing for triple damages based on "willful" patent infringement. It's pretty easy to claim, and any evidence that someone (e.g., a random engineer) read an allegedly infringing patent can be enough to make it stick in a patent lawsuit. Every tech company I've ever worked for had a standing policy of telling its employees to not read any patents, ever, unless their lawyers told you to do so.

If one is lucky enough to get deposed in a patent lawsuit, one may have to make a legally-binding statement as to every potentially-relevant patent one is aware of. Needless to say, it looks really bad if one claims to not have read or heard of a given patent but then subpoena'd search results show otherwise.

It's really perverted, as the primary motivation for the creation of the US patent system was to promote disclosure of inventions to advance the state of the art...but becoming aware of another's disclosures can end up costing huge sums of money, so people are encouraged to stay ignorant.

[Paelos]

The point of bringing up Faction Champions is that sometimes when you make the AI better, it really shows how flimsy the underlying mechanics of the game really are. We are used to the trinity because it always existed that way in raiding games from an MMO standpoint. What Faction Champions did was really expose how stupid that is by making a small change to the Boss Behavior. When you remove the taunt mechanics of tanking (which are totally ridiculous when you think about it) the entire thing falls apart.

We forgive stupid AI decisions all the time for the sake of maintaining the gaming fun status quo. Other examples would be aggro radius, leashing, and ridiculous LOS.

When I say I want better AI, I don't necessarily mean an AI that violates the things we passively understand as necessary to have fun (like taunting, leashing, radius). Rather I mean an AI that doesn't clip through walls, or charge down hallways into a meat grinder, or attack a force of 1000 with 200 ill-trained men. Yet those things still happen all the time in games today. Because programmers have become very lazy about AI and don't think it has any value added to their games.