Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Neruz]

The would certainly be interesting to see hapen in the game but how in the world could you have the game generate tactics for the predators like that?

Learning AI. Would eat the pants off a modern system, but systems X years from now might be able to pull it off.

Making Ai that isn't tricks and hacks mostly involves uber-computing and ultra-programming. Unless someone goes and makes a package where it's mostly done I don't think the time to develop a system like that is something we want to wait through.

A genetic algorithm might be able to develop the fighting styles, but it would have significant disadvantages. First and foremost, it wouldn't be realtime. So it would have to develop a non-situational fighting style based purely on information in the raws at worldgen, or we would have to sit through the algorithm every time we wanted to update a creature's fighting style. Could actually be done, but as you say, I sure as hell don't want Toady to spend his time on such a solution when he has much better things he could do. I think simplistic AI is fine for now. Maybe consider better AI (apart from sieges) for version 2 or some other far off date.

Well we're already seeing the introduction of genetic systems with Hair and Beards in the next version.


If you think about it; it makes sense. World Gen is exactly that, if i stop the game at year 1, then the world is literally a year old. So it's not in the least bit surprising that everything is living off preprogrammed instincts and hasn't had the chance to adapt to specific instances yet.

But if i let it gen to a thousand years, or more, then i'd expect to see some specialisation.

[Astus Ater]

If you think about it; it makes sense. World Gen is exactly that, if i stop the game at year 1, then the world is literally a year old.

This isn't completely true, rather you have a world with one year of recorded history. As evidence by creatures with no known year of birth. So there could already be preprogrammed instincts from the "dark ages" of your world.

[Neruz]

If you think about it; it makes sense. World Gen is exactly that, if i stop the game at year 1, then the world is literally a year old.

This isn't completely true, rather you have a world with one year of recorded history. As evidence by creatures with no known year of birth. So there could already be preprogrammed instincts from the "dark ages" of your world.

Which we already have. Read the rest of my post.


Good god man, you even managed to use the exact same words. Not reading other people's posts must be spreading, it's a disease i tell you!

[Akigagak]

CAN WE GET THIS THREAD LOCKED UNTIL TOADY IS READY TO ANSWER QUESTIONS, BECAUSE THIS THREAD IS GETTING RIDICULOUS LEVELS OF IDIOCY AT THE MOMENT.

[bluephoenix]

I finished up the lingering underground civ issues today -- how they are listed in the civ screen, an issue with how critters that become notable are linked to the underground civs, that sort of thing.

So UG civs done? 

[Footkerchief]

CAN WE GET THIS THREAD LOCKED UNTIL TOADY IS READY TO ANSWER QUESTIONS, BECAUSE THIS THREAD IS GETTING RIDICULOUS LEVELS OF IDIOCY AT THE MOMENT.

Calm down, or just don't read this thread if you can't stand to see it polluted.  The only current off-topic conversation is like a dozen posts long and that's nothing for this thread.  We're still having plenty of on-topic posts.

But yeah, Neruz, please resist the temptation to start slapfights.

So UG civs done? 

Sounds like it!

I want to know if this means that he will be done with the underground week, or if he just finished another part of the underground but still has a few more days before moving onto something else...

Like I just said:

This was the latest breakdown I could find of the underground to-dos: "I also organized the remaining underground notes and threw various things out for this release, so I'm down now to needed work in various categories. More or less it boils down to testing out the new life and populations more, testing out and fixing up the underground animal people civilizations, fixing map quibbles, deciding what to do with the rope stuff, and dealing with a few hanging issues with the stuff I put in the spoiler before."

So after he finishes underground civs, there's still the rope issue and possibly some other stuff.  But it's getting closer!

[Shoku]

If you're talking aboult Interactive Fiction, that's all scripted. It wouldn't be much use in this sort of situation. Now I have to give Zork another go. Thanks for killing my week. 

Scripted like sounds would have to be? Yeah.
I was being a bit broad though aimed at varying perceptions. Giving out different descriptions of the same stuff depending on whatever.

Sure I bet most of it follows a really predictable tree of go here do this but that's not universal.

The would certainly be interesting to see hapen in the game but how in the world could you have the game generate tactics for the predators like that?

Learning AI. Would eat the pants off a modern system, but systems X years from now might be able to pull it off.

Making Ai that isn't tricks and hacks mostly involves uber-computing and ultra-programming. Unless someone goes and makes a package where it's mostly done I don't think the time to develop a system like that is something we want to wait through.

A genetic algorithm might be able to develop the fighting styles, but it would have significant disadvantages. First and foremost, it wouldn't be realtime. So it would have to develop a non-situational fighting style based purely on information in the raws at worldgen, or we would have to sit through the algorithm every time we wanted to update a creature's fighting style. Could actually be done, but as you say, I sure as hell don't want Toady to spend his time on such a solution when he has much better things he could do. I think simplistic AI is fine for now. Maybe consider better AI (apart from sieges) for version 2 or some other far off date.

Well we're already seeing the introduction of genetic systems with Hair and Beards in the next version.


If you think about it; it makes sense. World Gen is exactly that, if i stop the game at year 1, then the world is literally a year old. So it's not in the least bit surprising that everything is living off preprogrammed instincts and hasn't had the chance to adapt to specific instances yet.

But if i let it gen to a thousand years, or more, then i'd expect to see some specialisation.

No the-

When you're talking about programming something with genetics it means AI with weighted actions. Say we wanted to have the dwarves eat a little smarter. If could cost 100 points (unless they were starving in which case it became the cheapest option,) so they'd avoid it for cheaper activities unless everything else added up to more than 100. If interrupting a job to go eat costs 60 or so and the job takes a fair bit of time plus all that walking the total task could be bigger than 100 so they'd just eat first.

Blech, I don't think I did a very good job of that.

PS. Your avatar makes me think you're that one character in a show that makes no sense except to themselves and it's affecting how I read your posts.

If you think about it; it makes sense. World Gen is exactly that, if i stop the game at year 1, then the world is literally a year old.

This isn't completely true, rather you have a world with one year of recorded history. As evidence by creatures with no known year of birth. So there could already be preprogrammed instincts from the "dark ages" of your world.

Which we already have. Read the rest of my post.


Good god man, you even managed to use the exact same words. Not reading other people's posts must be spreading, it's a disease i tell you!

Though with "yes I said that" posts it's no wonder I'm viewing you like one of those characters...

[tfaal]

I've heard that the game will simulate fractures for rigid tissue layers, but I'm not really sure what that means in this context. Will a low contact area bludgeoning attack damage surrounding tissue, or is it handled by something more complex? Is the same system of fractures in place for human bone and rock man "flesh"?

[Footkerchief]

Will a low contact area bludgeoning attack damage surrounding tissue, or is it handled by something more complex? Is the same system of fractures in place for human bone and rock man "flesh"?

The fracture system is unified, yes.  A blunt attack with low contact area may damage softer tissues surrounding a rigid tissue, depending on the specific properties of the materials involved and the attack itself.  Here are some explanations from Toady on how it all works:

Spoiler: long quotes (click to show/hide)

How much does this (skin thickness) actually matter in terms of combat?  Will there be things that daggers can't cut through?

That's coming up very soon, so I can't say exactly how much.  The idea is that a thicker tissue would matter a lot, so a bulbous creature with a large layer of lardy rolls would have a great deal of protection.  It'll take some balancing, and I'm not to the point where I've tested it out.  In terms of daggers being able to cut in general, it currently uses the shear properties for this, and if the dagger is a harder material it will form some sort of cut (assuming the strike is forceful enough to get beyond the shear fracture point for, say, skin), but on a titan, that should be neglible unless you make a hobby out of it.

Does bludgeoning damage transfer?  (AKA is unlikely to do damage to skin or fat but will bruise muscle and break bone)

Yeah, though it's sort of cludged since there's a lot going on and I wasn't quite wrapping my head around everything.  Right now it'll check the impact elasticity (which I guess is actually impact yield strain or something) of the outer layer, and that'll let it bypass impact fracture after a certain point (while still allowing bruises and ruptures without the layer eg skin actually breaking) even if the force is high enough (as the layer is assumed to have made way for the moving object in a sense).  So if you make a michelin man style creature that's just a purely elastic material, a mace will be unable to harm it unless it has blood flow and can bruise and so on, while relatively inelastic materials like bone will fracture, even through skin.  Skin has a lesser shear fracture than bone on the other hand, so a knife will cut through skin more easily than bone.

This can all use a bit of work, but I'm hoping to get something that seems sensible during play out of it this time around.

So we'll also have armor acting like a pseudo tissue layer for purposes of transmitting bludgeoning damage and so forth?

If an iron sword strikes some steel plate armour and fails to penetrate it, and the damage is converted to bludgeoning (AKA the force of the blow is spread out over the plate), will bludgeoning damage be inflicted to the creature beneath the armour?

Additionally, if we're talking about a steel-skinned creature (Giant Magma Ants?) instead of a steel-armoured creature, is the force transfer handled differently?

Yeah, that's basically how it works.  The armor is all checked in order from outer to inner, then it gets at the tissue layers, altering the character of the attack as it goes.  There's a little left to do there, so it's not completely clear to me which way it'll go as far as how much force is transferred (or if it matters if it's skin or armor -- it doesn't currently, though yeah, there is a difference in real life).  It has to be enough that skin doesn't stop bones from being broken, but not so much that armor is useless against impacts, depending on the mats and nature of the armor/clothing.

Obviously, from what you've said, it will be difficult to cut through, say, steel with iron. But HOW difficult?

I've only whacked the poor dwarf with the hammer.  Haven't tested armor yet, and all of this is very number dependent.  Ideally, there would be a point to upgrading equipment without making the prior equipment worthless.  A softer material can break a harder one with impact force currently if it's swung hard enough, but it would suffer itself in the process (though not so much if it is against something that just isn't as massive as it is).  There's a sort of mid-range spot I'm missing right now with wedge effects (like a dull axe), but I'll get at that at some point -- I have a contact area variable in the attack definitions that should make the pressure increase, but there are many things to balance.  There's a sweet spot we'll find eventually as long as everything is placed correctly.  Skill is probably still going to be the most important thing, unless you've got really really tragic atts or something (you can't be strength 0 unless you wait for your atts to rust for a long, long time, but if you theoretically did get down there, you wouldn't be able to cause any damage at all).

How do the strength values for metals interact with their effectiveness as weapons and armor?

Currently shear yield/fracture is used to determine edge effectiveness and the effectiveness of a material against a cut.  Impact yield/fracture is used to determine resistance against impacts.  Solid density will increase weight and therefore the impact of attacks.

[sproingie]

When you're talking about programming something with genetics it means AI with weighted actions. Say we wanted to have the dwarves eat a little smarter. If could cost 100 points (unless they were starving in which case it became the cheapest option,) so they'd avoid it for cheaper activities unless everything else added up to more than 100. If interrupting a job to go eat costs 60 or so and the job takes a fair bit of time plus all that walking the total task could be bigger than 100 so they'd just eat first.

That's not what genetic AI is about.   Genetic algorithms are about taking a program, running it through a fitness function, then taking the most fit programs and combining them into a new program.  Generally you don't go actually rewriting code, the program is just parameters to more fundamental behaviors, much the same way as DNA codes for proteins, not macro structures.

For most invaders, the fitness could be as simple as whether they live or die.  That approach probably wouldn't work too well for dwarves, and would realistically take too long for everything else anyway.

[Emily Murkpaddled]

Rereading those quotes detailing just how beautiful the new damage mechanics are is making me quite excited! It's actually a perfect one-two, though; combat is being made much more dangerous, while we also receive the tools to more properly train and manage the soldiers who will be involved. The idea that that choice in weapons and materials will now matter beyond simply suiting my tastes is very .

[Shoku]

When you're talking about programming something with genetics it means AI with weighted actions. Say we wanted to have the dwarves eat a little smarter. If could cost 100 points (unless they were starving in which case it became the cheapest option,) so they'd avoid it for cheaper activities unless everything else added up to more than 100. If interrupting a job to go eat costs 60 or so and the job takes a fair bit of time plus all that walking the total task could be bigger than 100 so they'd just eat first.

That's not what genetic AI is about.   Genetic algorithms are about taking a program, running it through a fitness function, then taking the most fit programs and combining them into a new program.  Generally you don't go actually rewriting code, the program is just parameters to more fundamental behaviors, much the same way as DNA codes for proteins, not macro structures.

For most invaders, the fitness could be as simple as whether they live or die.  That approach probably wouldn't work too well for dwarves, and would realistically take too long for everything else anyway.

And you think that someone who had thought having heritable hair colors was a genetic algorithm will understand what you just said?

[G-Flex]

Probably not, but that doesn't mean that giving a bad explanation helps either, honestly.

[Shoku]

I noted in the post how far short I fell. You could think of it as an invitation for someone to give a better example that does what I was aiming for.

Anyway a genetic algorithm wouldn't be a good way to get behaviors like running animals off of cliffs because the prey would have to already have the "run off cliff" behavior when running as a group, which is to say this would be a pathing issue involving tiles you can't normally path through. I think we've heard enough about that recently that I shouldn't need to go into details about it.

[gerkinzola]

Heh, I've just been reading those quotes and it got me thinking... would wooden armor be actually viable now against bludgeoning attacks as it can absorb and dissipate a force more effectively than say iron?