The Gare and Elin Translate It Out: keikaku means plan

[Colamisu]

That's not what it means when it comes to game design though. If the enemies are told "go here, do this", that's script. If they're told "FIND COVER AND KILL THIS GUY" and do the rest themselves that's code. And no, today's AI doesn't have to be told that. At all. This hasn't been necessary for a while and with the class setup of modern uber eloquent devkits it's easy as hell to pull off, which is why it's not that uncommon. If you have to manually tell every character what to do, you're making way more work for yourself than you have to as a programmer. Calling it "random AI" is an absolute misnomer and not even how that works. It runs an AI routine that relies on random number based class selection combined with analysis of the battlefield and some other such nonsense which yields an enemy that gives the illusion of thought because it's intelligently unpredictable and reacts to parameters in unique and different ways each time. As opposed to "run right" or "walk towards player". Which is the whole basis of artificial intelligence. And, yes, a more random AI is, by it's very nature, a better AI. Because intelligence demands randomness. If you can predict the outcome without forcing it, it's not an AI, technically speaking.

I wouldn't really say that a more random AI is better. As you mentioned, the reason for randomness in AI is to avoid taking the same path repeatedly. There should always been an underlying logic, though. It may not make sense to you, but it'll be logic. Generally an AI scheme will look at the possible choices (i.e. move to this cover, move to that cover, stay here and shoot, etc.) and assign to each a point value relating to what it considers its goal. For example, move to cover A has a score of +10 because that's the average of all the leaf nodes for the game tree from that point. Stay here and shoot has a score of -5, still using the average, because it has a score of -100 on a directly connected leaf (i.e. the NPC dies), and a lot of low scores on the leaves farther down the tree.

A lot of AI schemes will always follow what it considers the optimal path at any given point. In other words, taking the one with the highest potential victory score. Some will follow path A with probability p1, B with probability p2, etc. with pi summing to 1. Usually, highest scoring path will have the highest probability.

And then you have stuff like the FF7 AI, which has a very simple set of rules it follows. It's really just a big nested if-else. You can look at it on FFWiki.

Gezegond, the scripting you're talking about is coding. Don't argue semantics. The scripting Ray is talking about is more like scripting a play. If you're going to say that in-engine scripts aren't code, then C++ isn't code either. C++ is just shorthand (a script, if you will) for assembly code. Which is, in turn, just a translation of machine code. Therefore, it's only code if it's in machine language. Have fun with that.

[Raype]

IThe innovations in formula 1 is very different from innovations in a city car.

This is what it has to do with. Everything has to do with this one phrase right here.

But they both have innovations, and they usually borrow from each other. Performance in a certain category doesn't have anything to do with innovation IMO.

There is room for development, improvement, and innovation in your car, a formula 1 race car, or a bicycle.

They're at different stages and the type of things that can be innovated and that SHOULD be innovated are wildly different. You'd be more concerned with increasing cooling in the F1 and increasing performance/fuel economy in the vehicle. The bike probably just needs better oil, tires, and shocks. To put them all on the same scale and say "yup, the same. They should all be part of the same thing" seems really really wrong. No street legal car is ever going to win an F1 race. And they don't have to. Because that's not even the point.

Yes, but of wildly different type and they're obviously not going to go to the exact same location via the exact same route. iOS, Console, and PC are all going in different directions with different strengths and weaknesses. And they have different points of improvement, which is why your iPhone5 isn't the same as your slightly aged PC running Win8 with a vidcard made last year and your console from 2007 is also wildly different and in a different place. Which is why each of these have their own "generations" and the like to denote this stuff. PC has never been included in console generation discussions because PC doesn't have generations. It has considerably more incremental updates that can be better summed up by DX versions or Windows revisions. I'm typing this on a laptop that's like 10 times as powerful as the PS3 I use for gaming. This isn't even a gaming laptop. You really can't lump two semi-similar things together just because you see it as convenient. It's an amazingly perfect example of apples and oranges. Except I think it might be more Apples and orange juice or something. I don't even know anymore. I should be asleep. Why the fuck am I still awake? I have work in the morning. I leave in an hour. I haven't slept.

I think I might be losing it. Again, I mean.

I seriously need to get someone to move in here.

Because holy fuck, I literally just wasted the night doing nonsense.

Kill me, please?

[Raype]

I wouldn't really say that a more random AI is better. As you mentioned, the reason for randomness in AI is to avoid taking the same path repeatedly. There should always been an underlying logic, though. It may not make sense to you, but it'll be logic. Generally an AI scheme will look at the possible choices (i.e. move to this cover, move to that cover, stay here and shoot, etc.) and assign to each a point value relating to what it considers its goal. For example, move to cover A has a score of +10 because that's the average of all the leaf nodes for the game tree from that point. Stay here and shoot has a score of -5, still using the average, because it has a score of -100 on a directly connected leaf (i.e. the NPC dies), and a lot of low scores on the leaves farther down the tree.

A lot of AI schemes will always follow what it considers the optimal path at any given point. In other words, taking the one with the highest potential victory score. Some will follow path A with probability p1, B with probability p2, etc. with pi summing to 1. Usually, highest scoring path will have the highest probability.

And then you have stuff like the FF7 AI, which has a very simple set of rules it follows. It's really just a big nested if-else. You can look at it on FFWiki.

Gezegond, the scripting you're talking about is coding. Don't argue semantics. The scripting Ray is talking about is more like scripting a play. If you're going to say that in-engine scripts aren't code, then C++ isn't code either. C++ is just shorthand (a script, if you will) for assembly code. Which are, in turn, just a translation of machine code. Therefore, it's only code if it's in machine language. Have fun with that.

I was probably horribly oversimplifying, but yes. That's a much better way to put it. Although then you have to factor in things like difficulty levels. Lower difficulty levels often throw in a requirement for the AI to select less-than-optimal solutions or it widens the scope of acceptable scores so that the AI is more likely to pick one that's not so great on a whim, or even actively choose bad choices because it wants to die. Easy mode sometimes just has suicidal AI.

[gezegond]

Yes, but if the AI "code" is taken completely out of the equation, then it's a scripted event. Which is what a lot of people did back in the day to make it *look* like there was an advanced AI under the hood when there wasn't. It was just a really dumb enemy that was designed to do something unusual in that one specific instance. It was just reacting to a player created flag and playing out a little cutscene that'll be identical no matter how many billion times you see it. In short, it's a script. Nowadays we see more coded AI that's actually dependent on actual things that are occurring in the game. When one of these AI sees a player, it reacts in numerous different ways. Hence, better AI and better coded AI, as opposed to illusionary scripted AI that's outright told to act a specific way when the player shows up. Or, in other words, it's the reason goombas walk left all the time but advanced enemies in modern mario games might decide to run right or throw out an actual attack depending on numerous factors, some of which are totally random.

Or, to put it another way, telling a lot of units what to do is terrible game design as a level designer and leads to a shitty product. Often players will simply exploit the AI (look at any poorly designed fighting game to see spectacular examples of this). But telling the AI what its options are so it can make varying choices (which may be dependent on such factors as where the player approaches from, what weapon he's using, and the sort of character the player is running) while throwing in some RNG to make the game different every time the player loads us brilliant game design and the sort of thing you're more likely to see nowadays. If SNK bosses were as easy to predict as Shao Khan, nobody would be raging endlessly about how Geese is a big boiling vat of dicks. They'd be huddled in the corner exploiting AI loops, predictable setups, and bugs for easy victory. Instead the AI reads you and does all of that shit to you instead (which is total cheating and kind of not how this is optimally supposed to work, but it works well enough for this illustrating example).

Regardless, "code" and "script" have very specific and simple definitions. I'm not saying AI coding does not exist or that it's bad in modern games. (Why do you always think of me as a radical sort of guy?) I'm saying there's huge room for improvement. Which you already agreed with. So let's stop discussing this, shall we?

Then I should illustrate that those AI you're linking are nothing more than toys and the stuff you're expecting of video game AI is decades off, not to mention it'll take someone years (and millions, if not billions of dollars) to implement even in a terrible fashion. I can pop a nice french pastry out of my toaster, it doesn't make my toaster a french chef. Just because someone made a half assed toy that runs some simulations doesn't mean it actually designs games. It's incredibly far from it in the same way that the moon landing is a far cry from colonizing mars. Hell I think that might be generous. This game designing AI is more accurately a couple of generations off from van Braun even being born.

That link I just threw there for giving an example of things that can be tackled but never have. This is the work of just one guy, with very limited budget, not some game development studio. His work obviously can't be compared to AAA games out there, but that's not my point. I was just putting an example of how a lot of things are "possible". What this guy is doing is "possible", it's not perfect. But nothing is. Everything starts from something small, which might look mundane and stupid. Think of the first computers, or the first cars, or the first games.

Wolfram isn't even technically an AI. It's just a gimmicky psuedo AI thrown on top of a search engine so people can attention whore it out as the next big thing. It's a giant remote control for an existing function. It can't interpret or analyze, and it can easily fuck up royally.

It's not, but that doesn't mean it's useless. You can use it in games. As I was waiting for your reply I thought about implementing something of the like, and it's quite possible. Sure, it's not perfect. It probably won't be the best feature ever. But it's a start. Again, what I'm saying here is that it is possible and there is room for improvement. Wolfram Alpha is not the best AI possible, but it's damn better than a long ass HTML page which has all the wolfram alpha data on it.

There's a reason Microsoft didn't showcase the voice commands of the XBone onstage (it was a slideshow). Because it doesn't work. Oh sure, it'll probably work much of the time and all that. Unless you have an accent. Or a speech issue. Or are in a noisy room. Or happen to be slightly too far from the mic. Or happen to speak too fast/slow for its liking. Or you use a different voice command than the one hard coded into it. Or saturn isn't properly aligned with neptune. Or any of the billion things that happens to get in the way of this stuff. And therein lies the problem. It'll work for certain people just fine but everyone else gets fucked. I'm a moderately accented Canadian with an off tone voice and a tendency to speak fast. Voice recognition software hates me with an unbridled passion. God help anyone from Jamaica or Portugal. Hell the Kinect had serious trouble picking up black people on its camera. We can't even do object recognition without serious glitches getting in the way. And command based voice recognition (the only one even remotely viable right now), doesn't even work right all the time. Full on voice recognition and interpretation of a functional capacity is so distant we might not see it within our lifetime. The closest we might get is lots of codespeak phrases (XBOX ON and the like) and that's just replacing your controller with voice commands. Which is already being done with Kinect (Mass Effect 3 had it, as I recall). You're just using your voice instead of moving your thumb a little. It's the wii-mote thing all over again.

I replied before that I don't mean voice commands. I mean written text. As in "you write English text with keyboard". That's something I can say is possible and I could probably do it myself in a game if I had the time and resources.

Voice Commands are another thing entirely. They're another program that gets sound as input and creates text as output. The text then can be used if it is any good. If I were designing a game, I would just create the text part. If by the time it was done there was any good voice recognition software I would buy it (If I could afford it) and implement it in my game. If not, I would just make it text based. I could make the voice recognition optional, so people who don't like it could easily turn it off and just use keyboard.

Voice recognition itself, I don't understand it much, so I won't talk about it. But I'm sure they'll get better as time passes.

[Zaladane]

I'M NOT READING ALL THAT CRAP!!!!!!

[Zaladane]

[gezegond]

Gezegond, the scripting you're talking about is coding. Don't argue semantics. The scripting Ray is talking about is more like scripting a play. If you're going to say that in-engine scripts aren't code, then C++ isn't code either. C++ is just shorthand (a script, if you will) for assembly code. Which is, in turn, just a translation of machine code. Therefore, it's only code if it's in machine language. Have fun with that.

Yes. If for some reason the AI developer is using in-engine scripts for main AI coding that is true. But usually, the AI itself is part of the engine, and the scripts were there for the level designer for specific actions. (At least, that's how it was in Half-Life. I can't talk about Uncharted 3 because there's no source code afaik) Still, cover locations are always done by level designers.

I don't know, but I just feel that you are just renaming things. Alright c++ is not "code", it's programming language. And assembly code is assembly code. And scripts are scripts. I know what you mean when you say "scripted level" etc. but I just can't accept that as the right definition for "script"s.

And anyway, I don't want to argue about this. I don't really care what you want to call what. I'm not really sure how we got into this topic, we were discussing Gen7 in general...

[gezegond]

This is what it has to do with. Everything has to do with this one phrase right here.

They're at different stages and the type of things that can be innovated and that SHOULD be innovated are wildly different. You'd be more concerned with increasing cooling in the F1 and increasing performance/fuel economy in the vehicle. The bike probably just needs better oil, tires, and shocks. To put them all on the same scale and say "yup, the same. They should all be part of the same thing" seems really really wrong. No street legal car is ever going to win an F1 race. And they don't have to. Because that's not even the point.

Yes, but of wildly different type and they're obviously not going to go to the exact same location via the exact same route. iOS, Console, and PC are all going in different directions with different strengths and weaknesses. And they have different points of improvement, which is why your iPhone5 isn't the same as your slightly aged PC running Win8 with a vidcard made last year and your console from 2007 is also wildly different and in a different place. Which is why each of these have their own "generations" and the like to denote this stuff. PC has never been included in console generation discussions because PC doesn't have generations. It has considerably more incremental updates that can be better summed up by DX versions or Windows revisions. I'm typing this on a laptop that's like 10 times as powerful as the PS3 I use for gaming. This isn't even a gaming laptop. You really can't lump two semi-similar things together just because you see it as convenient. It's an amazingly perfect example of apples and oranges.

Well, I see what you mean here. I must think about this more...

Except I think it might be more Apples and orange juice or something. I don't even know anymore. I should be asleep. Why the fuck am I still awake? I have work in the morning. I leave in an hour. I haven't slept.

I think I might be losing it. Again, I mean.

I seriously need to get someone to move in here.

Because holy fuck, I literally just wasted the night doing nonsense.

Kill me, please?

Time zones, man. I didn't know you were staying up. I accept that everything I said is wrong, now you can sleep.

[gezegond]

I'M NOT READING ALL THAT CRAP!!!!!!

YOU SHOULDN'T.

Raype goes off topic too much. and then goes more off topic in the off topic. And then Colamisu joins him to defend him in the off topic of the off topic.

Anyways, it's all my fault. I apologize to everyone involved.

[Dr Mario]

=
I replied before that I don't mean voice commands. I mean written text. As in "you write English text with keyboard". That's something I can say is possible and I could probably do it myself in a game if I had the time and resources.

You can try, but decades of work on text parsing and translation software have shown that even strict text instructions are an absolute bear to get working properly. Assuming you can get the grammar working properly (good luck here, there's no universal agreement on how the language is meant to be constructed and pretty much everyone violates most of the widely agreed-upon rules more or less constantly - the kicker is that they don't do so in anything resembling ways that are coherent across the population, its very individual) you still run headon into the issue that language usage can vary radically across demographics. Just getting the dictionary meanings down is a bastard (remember, many words have multiple overlapping usages), dealing with localised denotations is a problem on top of that and getting the right connotations? that's months of work per group.

[gezegond]

You can try, but decades of work on text parsing and translation software have shown that even strict text instructions are an absolute bear to get working properly. Assuming you can get the grammar working properly (good luck here, there's no universal agreement on how the language is meant to be constructed and pretty much everyone violates most of the widely agreed-upon rules more or less constantly - the kicker is that they don't do so in anything resembling ways that are coherent across the population, its very individual) you still run headon into the issue that language usage can vary radically across demographics. Just getting the dictionary meanings down is a bastard (remember, many words have multiple overlapping usages), dealing with localised denotations is a problem on top of that and getting the right connotations? that's months of work per group.

You won't need all that in a game. You just make it to understand (in a sort of limited sense) the in-game logic. What you're talking about is near human like intelligence which obviously isn't easy.

But I promised I will stop discussing this. I won't reply anymore. (Or try not to)

[Gare]

Long posts make me moist.

[gezegond]

Long posts make me moist.

tl;dr I think Gen5 was better than Gen7. Raype tries to prove otherwise.

What started the discussion was this post. Which I used to show that Discs are teh awesomez and therefore Gen5 is TEH BEST GEN EVA.

[Dr Mario]

You won't need all that in a game. You just make it to understand (in a sort of limited sense) the in-game logic. What you're talking about is near human like intelligence which obviously isn't easy.

You kinda need something like that if you want actual language-based input.

If you're content to use the old text parser approach of noun the verb, then you can bypass all this. That just comes with its own ways of frustrating the player (and design limitations. The format really starts to struggle in any context where you might have multiples of the same object onscreen concurrently). It also makes its limitations very apparent, which is part of the reason that it fell out of use in favour of GUI-based control.

[gezegond]

You kinda need something like that if you want actual language-based input.

If you're content to use the old text parser approach of noun the verb, then you can bypass all this. That just comes with its own ways of frustrating the player (and design limitations. The format really starts to struggle in any context where you might have multiples of the same object onscreen concurrently). It also makes its limitations very apparent, which is part of the reason that it fell out of use in favour of GUI-based control.

My approach would be to use patterns to categorize and then sub categorize the text. (For instance, is it an order, a question, small talk, etc. and then sub categorize for instance the question, is it asking about why, where, how etc.)
Then find keyword(s) in the sentence. ( These are in-game words, such as names, places, objects, etc. which can be in a list)
Each character would need a knowledge database, which it would have to cross check the keywords with to produce relevant replies.
When character is created, the knowledge database is randomized. If it's story character the knowledge database is designed by developers.

This is something I thought up in minutes. Obviously there's gonna be more to it. But it IS possible. I just hate the "no one has done it before, therefore impossible" kind of thinking.