Civilization V DLL Source Code coming with Fall Patch

[DonQuiche]

As for moving people to Mars, human expansion is necessary for our survival of potentially cataclysmic events as well as to relieve the burden on our own home planet.

A for moving people to Mars, this would be a living hell.
So I vote for other people being deported there and me staying here. Anyway, even if you want to imagine terraformation and such, you're still limited to two planets and the last time I checked physics said it is not possible to go faster than light, so bye bye other systems (physics laws are hard es that never tolerate any exception - at best you can find one more gentle than the others and willing to beat them up for you, but so there seem to be none this time).

By the way, people tend to put a lot of faith into science, but nature is inflexible, so I do not think that everything you imagine is possible. By the way, medicine did greatly improved the children's health (50% of them were dying before being 10 in the renaissance) and survival to parturition for women, but the life expectancy once you reach 60 is about the same than it was a century ago, although we're dying from different things.

For the rest of your post, you're assuming that AI would unavoidably being architectured like humans. But nothing proves that intelligence needs emotions in order to establish priorities. And what you say about the limbic and prefrontal cortexes does not actually prove that emotions are required for humans (!= AI) to establish priorities, it may just be that the limbic cortex is required for both emotions and priorities.

Also, as I already stated, what nature does, we can do it better. So, I am certain that we can make machines that last longer than we do.

 

[PawelS]

Seems this discussion departed from the DLL subject significantly

 

[3335d]

DonQuiche, have you heard of the Alcubierre warp drive which may allow us to go faster than light should it be possible? Recent research shows that it may take less energy than previously thought, and NASA has taken it seriously enough to attempt to prototype it in the laboratory. Also, time dilation means that people on even sublight spacecraft will indeed survive the journey, provided the craft is fast enough.

What I mean by medical advances include entirely new things like replacing diseased or broken organs, curing cancer and heart disease with nanotechnology, and generally promoting a healthier lifestyle, which should extend our lifespan greatly. Genetic therapy means that we will someday be able to cure all genetic diseases which plague us. Better control of infectious diseases will reduce mortality rates in poor countries even more than it has already done. While the increasing ease of manipulating genetics means that eventually even high school students may be able to weaponize viruses such as the Spanish flu, high school students will also be able to quickly make defenses against these viruses (assuming that they have that level of intelligence at that age).

Making the move to magnetic levitation for transport will reduce our transportation energy needs by 90% and help to heal our planet, as will superconductors and the increased deployment of environmentally friendly energy sources. Despite our limitations as biological beings, it is still possible for humans to do all that robots can do and more.

As for emotions, tell me what "rational" way can quickly choose between a, b, c, x, y, and z, provided that none of them harm the individual choosing between them. Rationalizing the difference on a data set with millions upon millions of elements would be an almost impossible task, explaining why shopping is such a pain for people with unconnected limbic systems, and explaining also why robots without emotion would falter as well.

As for your comment about humans doing better than nature, it's possible that we can apply this doing better to our own body. Refer back to my paragraph about advances in medicine.

With all due respect, you seem to be a bit behind on some of the advances in scientific knowledge that have recently surfaced that may make the stuff of science fiction science fact, perhaps within our lifetimes

Seems this discussion departed from the DLL subject significantly

 

[crawf0rd]

Back to AI. Here's my dream-adaptive-ai:
-The game would keep track of what the human player builds
-When the player wins, it dumps into the AI file the amount of times the player built each unit and change it into a flavor for that certain victory type (the more he built a certain unit, the higher the flavor). If the player has bad strategies, so does the ai. It throws out the need for a wide amount of difficulties.
-Rinse and repeat for religious beliefs, buildings, improvements, etc.

Now I have zero clue the realism of something like this. I'm sure if something like this were possible, it would've been done. A man can dream, can't he?

 

[3335d]

Back to AI. Here's my dream-adaptive-ai:
-The game would keep track of what the human player builds
-When the player wins, it dumps into the AI file the amount of times the player built each unit and change it into a flavor for that certain victory type (the more he built a certain unit, the higher the flavor). If the player has bad strategies, so does the ai. It throws out the need for a wide amount of difficulties.
-Rinse and repeat for religious beliefs, buildings, improvements, etc.

Now I have zero clue the realism of something like this. I'm sure if something like this were possible, it would've been done. A man can dream, can't he?

Not a bad idea. Such an AI may be able to approach the level of humans who play the same strategy again and again. CiV is such a limiting game, had it been more open-ended then humans would have a lot more room to strategize.

 

[DonQuiche]

Well, I suggest a moderator split the part about Mars and such in order to keep things clean.
I will avoid polluting this topic meanwhile.

 

[Jatta Pake]

I think the discussion regarding the need for human-like AI is fundamentally flawed. What is the point of focusing on making an AI more human-like? Do we really want an AI who rage quits when you take one of his cities? One that uses exploits to get an unfair advantage? One that trolls you? One that models a mouth breathing self proclaimed genius who gives you unwanted advice because he lives in mom's basement and does nothing but play Civ?

AI will NEVER replace human opponents because intrinsically we play against humans for the social aspect that is part of the "fun". Part of the glory of competition is besting an otherwise equal opponent. If that wasn't the case, humans would flock to sporting events to watch a motorcycle race a dude on foot.

The purpose of AI is to create "fun" for the human player. Modeling human emotions may result in "fun" for players but what actually happens in the background is completely irrelevant. I think AI is the wrong term. I would call them "Fun-anators" but I doubt that will catch on.

My version of AI (Fun-anator) would do the following:
-Be adaptative based on player feedback. The system would guage player emotion and react accordingly.
-What is a "fun" experience would not be defined as merely "happy". It would include all emotions. For example, someone could play a frustrating game of being kicked around as the underdog all game to finally eek out a win in the final ten turns. These are the games we remember forever.
-The AI would adapt to the individual. As the player gets better, the AI could scale up difficulty as well. Alternatively, others may want a tough AI at all times to punish them mercilessly until they master a certain level of play. The AI should know the player.
-The AI would demonstrate novel techniques to the player and the ability to surprise. The AI would also be tuned to generate strong emotions in the player but balanced so hatred was not generated at the game itself. It's fun to hate an AI, even ones that do jerk moves just to make you mad, but it's not fun to hate the entire game. On the flip side, an AI who remains loyal against his own interests generates feelings of friendship or even pity.

We don't need human like AI. Time would be better spent allowing humans to play each other. We do need AI that creates a fun experience and this is what the effort should be for AI development.

 

[ThisNameIsTooLo]

When the DLL source code comes out, will it be possible to change the way music is played? I'd like to have separate playlists for each era again.

 

[Biz_]

it's not that everyone wants the AI to be exactly like a human (although a lot of people DO want that)

it's that Civ is a game about resource management tradeoffs that is heavily reliant on a threat model and is very adversely affected by cheating

if the AI could be a threat without cheating, then people could play the actual game as it was designed without going through the headache of the multiplayer modes

the AI needs to know how to play the game as a prerequisite for being "fun," however you define it

 

[stfoskey12]

It throws out the need for a wide amount of difficulties.

But somehow I like the feeling of achievement when I move up to the next difficulty.

 

[The Little One]

YESYESYESYESYES!!!!!

Now if only I knew anything about .dll format

 

[PawelS]

YESYESYESYESYES!!!!!

Now if only I knew anything about .dll format

You don't need to know anything about the .dll format. But you need to know C++

 

[Moriboe]

I will be interested to see how the AI is coded, at least. A few things may be easy to change.

E.g., it has bothered me from the beginning that AIs don't go for victories that are easily within reach, just because they set a different goal for themselves 200 turns before. In Civ4 nearly all civs would go for a space victory if they had the chance and it made the end game more exciting. In Civ5 you can fall behind a full era and still win. Where's the satisfaction in that? Of course it would get frustrating if the top AI, usually swimming in money on immortal+, just buys all CSs every time, but that's another matter.

 

[Pazyryk]

I will be interested to see how the AI is coded, at least. A few things may be easy to change.

E.g., it has bothered me from the beginning that AIs don't go for victories that are easily within reach, just because they set a different goal for themselves 200 turns before. In Civ4 nearly all civs would go for a space victory if they had the chance and it made the end game more exciting. In Civ5 you can fall behind a full era and still win. Where's the satisfaction in that?

Yes, this seems like a good area to look at, in terms of AI gain / coding effort. A function could check things like civ gold, current CS situation, tech progress, city production, ...and then reset the targeted victory type.

You might need to add some functions to help the AI figure out how best to buy-off CSs (for diplo), or how to decide where to build space ship parts (for science). For the later, I have a strong suspicion that there is no AI at all. Has anyone ever seen the AI complete the space ship? I suspect that space ship parts are determined like other city builds: by a random number generator (i.e., no AI). In this case it would only be by luck if the AI can pull it off in any reasonable way.

Of course it would get frustrating if the top AI, usually swimming in money on immortal+, just buys all CSs every time, but that's another matter.

The frustrating thing is that the AI has to have such obnoxious handicap bonuses to be in the game at all.

 

[Putmalk]

Good thing I know c++ (not so much about classes yet, but by the end of semester I'll know that).

Cannot wait for the DLL.

By the way, I thoroughly expect this to come out on or near my November 10th birthday....

 

[smallfish]

You might need to add some functions to help the AI figure out how best to buy-off CSs (for diplo), or how to decide where to build space ship parts (for science). For the later, I have a strong suspicion that there is no AI at all. Has anyone ever seen the AI complete the space ship? I suspect that space ship parts are determined like other city builds: by a random number generator (i.e., no AI). In this case it would only be by luck if the AI can pull it off in any reasonable way.

Not sure if this would help, but here's an old shot of a GnK Immortal game where Sejong won SV in T313 instead of just building UN with his many CS allies:

Spoiler :

Would that be a reasonable attempt at victory?

 

[Pazyryk]

Would that be a reasonable attempt at victory?

Turn 313 on Immortal does not sound reasonable, unless you were doing something weird in that game like dominating the map but not finishing off AI players. I'm assuming many AI players had finished the tech tree long before turn 300 (that's my experience at Immortal, standard map size and speed).

They should finish the spaceship about 10 turns or less after they get the last tech for it. Why not? That's what a human player does and the logic is not at all difficult: 1) target appropriate techs in a reasonable order, 2) build SS parts in the highest production cities, cancelling any other builds already in progress. Yes, the AI can win this victory. But as I said I'm not convinced there is any specific AI for this. They will eventually get there just by picking techs randomly and randomly building SS parts in whatever city as building queues empty.

 

[GreekIrish]

Does this mean we'll be able to have total conversion like FFH? Or are those already possible?

Back to AI. Here's my dream-adaptive-ai:
-The game would keep track of what the human player builds
-When the player wins, it dumps into the AI file the amount of times the player built each unit and change it into a flavor for that certain victory type (the more he built a certain unit, the higher the flavor). If the player has bad strategies, so does the ai. It throws out the need for a wide amount of difficulties.
-Rinse and repeat for religious beliefs, buildings, improvements, etc.

Now I have zero clue the realism of something like this. I'm sure if something like this were possible, it would've been done. A man can dream, can't he?

That would be some very advanced AI if it could determine success based on build order and player actions. Use real scenarios as a guide and logically determine what's best. Survey thousands of players and you might have yourself some smarter AI.

 

[Pazyryk]

Does this mean we'll be able to have total conversion like FFH? Or are those already possible?

Total conversion is certainly possible right now. Something like FFH, with perhaps 70% of its mechanics, is possible without dll. The dll gets you the possibility for the remaining 30% (or so) of the special mechanics.

Problem is, however, that it takes 1000s of hours to make something on the level of FFH. The dll won't change that.

 

[deanej]

And it also won't change the fact that civ5 modding is not optimal to total conversion mods either. In civ4, you just created files with your stuff and didn't have any of the base civ stuff in them. In civ5, you have to explicitly delete everything not part of your mod, each item at a time.