/Problem/
I have listened through the PodCasts and there is specific attention to the addition of super tokens to simultaneously solve issues of AI settlement vs. settlement relationships with an actual globally noticeable effect which assists in the development of story and meaning in the world. This mechanic is very interesting, and unless I misheard or misunderstood I believe it should be split in two. It should be a mechanic of local and regional story development and global story development, working symbiotically (it is put in a similar manner). In lieu of the latest podcast, this is not "a problem" but "what I don't like"
/The opportunity available/
I believe it is good mechanic and is probably much easier than my solution, however it will pan out better in the end. Super tokens as global effect triggers are ideologically separate from the issue presented by factions. I believe that the moment the AI settlements were given distinctions with color and architecture along with the notion of a persistent world really presented a different idea. We are used to generic mobs and even scripted NPC responses so we believe that each AI inhabitant is a target. Instead the AI, even if not apparently unique, has unique history, experience and interactions. This opportunity is unique in an MMO, even for such a small comparable size.
/Individuality/
This is even further emphasized by the notion of history with fewer AI settlements but the possibility of destroyed ones. This primarily depends upon the specific structure of the AI Director. I assume this is a variant of a multi-agent system, which is based upon nesting each settlement as an agent and then each individual as another agent within that. One of the fundamental ways this is abused in other games is by treating a village of AI as individuals who think in parallel where as we seek to give them human behavior which is distributed. I mean each person must share information, language and our physical body obfuscates the concurrent computation. These can be indistinguishable. This behaviors of one homogenize the group behavior of a settlement. Intuitively this should not be so, if we seek more human interactions from an AI.
/4 Issues to Address First/
I mean that it is necessary to have a more complex system than the common way a retreating mob appears to convey information of threat and danger to instill a call to arms. As it is a persistent world, first we can expect that with various degrees often emotion strength is inversely proportional to time. Second this again implies history and knowledge, or simply they should use knowledge token. It is of course not necessary for the AI to publicly advertise, however I imagine simply a in-discriminant sign since it cant be read. Third a coherent structure in AI society, each inhabitant will act as an agent together with the other inhabitants. Last persistence implies finitude, objects decay and disappear, just as AI inhabitants, but their knowledge and influence is lost when they die.
/The Synergy and Evolution/
The most immediate problem for programming this for anyone should immediately be the AI. If we truly seek reciprocal human-like interaction with the AI then we must initially mesh the original routines as AI have no previous history. This could then be removed and tested and if successful pushed down to 3. Otherwise it may be necessary to continue to debug or even used the method mentioned, a Monte Carlo method simulation of two small settlements, adjusting parameters. The game world should actually just be made a randomly generated but small grid for 2 or 3 settlements.
/Beyond that -- Semiotics and Intention/
After this it is really far off, but I will note that towards the end of the last talk, there is essential made about AI intent. This quickly lead into AI developing animosity for a player or AI using competing resources. This further leads to the comical vespene gas analogy, which is quite nice. Despite either the point I believe Eskill impresses, although should perhaps more is the distinction between deduction and inference based knowledge systems. Loves size and structure allows it stray from RPG and use RTS elements like Fog of War and FPS equivalent line of sight and radar/stealth. This leads to the issue mention on AI variation.
/AI variation/
It is mentioned that aesthetic changes to AI are planned, and I believe he mentioned he wanted to tie this into more internal AI customization but I believe this is actually along with my ideas. Much of what video games are about is graphics and much of the human brain is spent on visual processes. A well formed connection or distinction between form and demeanor gives a strong psychological. To reinforce story elements the current progression must have meaning, perhaps the leader of an AI settlement told colluded with against another, one which nearly took your settlement and is quite ugly. These factors matter no matter how little. Additionally as AI settlements spawn new individuals, it is simple to give names and relationships and even denote this graphically whether physically or by ornamentation, in the tribal age AI competed somewhat RTS style and could be discriminated with clothing, etc. On the other hand Ryzom was not a too well known MMORPG with races that had an initially unknown histories so as to assist in the impression of genuine interaction with an alien individual. I highlight Ryzom specifically because the common fantasy races and sci-fi universes are quite recognizable, this race variation is the typical method in MMOs but when I had a trial of Ryzom one of my first impressions was the combination of morphological,linguistic and cultural traits.
/Final Super Tokens and modeling/
Super tokens will be a great addition, beginning a global event which creates a reaction in both players and AI, their interaction beginning the next iteration. I realize that was rather long and difficult. However in keeping with stochastic nature of the procedural generation programming the 4 steps I spoke about amounts to adding additional parameters, determining the correct values to more human and complex is an issue of modelling the behavior with a Bayesian Network. When I mention passing messages and social structure that is quite literally an issue of Belief Propagation. After the distribution of each random variable is found to be within range of successful testing, personality traits and other AI distinctions can be generated from changing different distributions.
Tweaking these will obviously take time, and one must be aware of why behavior (perhaps waving) are less common when used as a general model for many AI ( the probability of the state transfer to waving from the majority of states is very low perhaps ) . In the end the behavior of the AI may combine to make the settlement behave much differently. Alternative approaches for problem may be used (genetic algorithms for instance). Or it may be that the optimization may be achieved as a problem in convex optimization. In the end the next development stage should avoid issues like cows, dogs or other animals or tweaks to graphics. The main focus should be to start implementing historical AI, even if only one or two in a settlement at first. Tracking the lifetime of each historical AI and paying attention to bugs or complaints is crucial to develop AI that not only see will seek super tokens but do so with a consideration of their settlement and several others and even attempt to affect strategy and diplomacy.
http://en.wikipedia.org/wiki/Belief_propagation
http://en.wikipedia.org/wiki/Bayesian_network
http://en.wikipedia.org/wiki/Convex_optimization
/Special/
The use of a laser pointer is mentioned in the last talk. It would be an interesting approach to bridge communication gap by a combination of social prowess (gaining trust) and learning gesticulations through inference. This could be done through mouse gesture technology or since most computers have webcams they it may possible, as it certainly is with a Kinect. The result is particular shared relationship with the specific AI, one you trained and now can talk to. Seems like too much work now, but was really just funny to me.
Very interesting. I just
Very interesting. I just emailed Eskil regarding this post.
We're usually on the Love Teamspeak in the late evenings, you might want to stop by sometime if you have the chance.
In my opinion "Super tokens"
In my opinion "Super tokens" represents what has been a massive hole in LOVE. Its a hole that i think is hard to see but one that everyone will notice once I have filled it. Let me try to explain:
I believe that you can divide a story in to different levels of detail, and that a good story operates on many levels. Any story that is believable operates on some sort of logic, that connects events using cause and effect and strings them together in to a story line. Love has a massive amount of cause and effect, but it still doesn't make it a god story.
Lets take Starwars as an example. Starwars can be summed up by saying: "Rebels prevent the evil empire tries to rule the galaxy". There you go! that's 3 films in one sentence. You can easily see the mechanics at work here. Rebels apparently counter evil empires. But it doesn't make much of a story, in fact you wouldn't sit and watch 6 hours of film to figure that out. No the real action happens on a much smaller level. Another mechanic in the starwars universe is that droids in control rooms can turn off trash compactors. Starwas if full of these small mechanics, like that sand people ride in single file to hide their number or that the Millennium falken can do the Kessel Run in less than twelve parsecs. The thing is that while they give the story a lot of texture, they are not driving the story. Even if you have no understanding of what a parsec is, you can still enjoy the films.
What really makes starwars work it whats in between: the plot. it answers the question: How does the "Rebels prevent the evil empire tries to rule the galaxy"? in fairly simple terms that everyone who doesn't care about the Kessel run can follow. Guess what, that planet sized space station that blows up planets is bad, we better do something about it! Oh, it is protected by a force field? we better turn that off! These are mechanics too, and even if the characters are spread out around the galaxy working on repairing hyper drives or training to become jedis, the can all keep track of this story.
Love has some of the best micro cause and effect system of any game ever made. Pretty much everything in the world can be broken down understood and exploited. It also has a very good long term story where the entire world changes over time, and you can see settlements evolve sometimes for far longer then 6 hours. What is missing is what in between, that's the plot. The goal is to have the super tokens fill that void. There has been a lot of stuff there trying to be the plot in he past, like being attacked, or attacking the other tribes, but without a cause and effect they are separate events and doesn't get stringed together in to a plot.
So have the events been separate? No not at all, they have all been driven by a careful system of cause and effect, but the problem has been that it has not been a cause and effect system you have been able to follow. And this is real the big problem in Love; how do you keep all players on the same page, when they can be on different sides of the planet and do what ever they want? That is really what it comes down to: information and focus. Yes the Rebel alliance could all spread out and fight individual tie fighters spread around the galaxy, but no one can miss that big ass death star and it seems more important so that's where we are all going to meet up.
People who play begin to play Love tries to figure out what the plot is and they think its the configure screens or something, when that's really the hyper drive of the Millennium falken they are trying to figure out, and not at all the plot. What i hope to do with the super tokens is to give them that simple goal and plot to drive them forward and then they can learn all the details later.
This is why the super tokens have some very clear properties:
-They have Clear owner (So you know who wants it back if they lose it)
-They are easy to spot, and there is only one of them. (Focuses players)
-They are very valuable yet very dangerous. (Makes you care about their state)
-If turned off, they influence the entire world, so that no one misses it.
-Some times the game will compel players to steam and turn them off.
This is my goal, to create this over arching, well paced plot, that will pull all players in and give them a common cause, while also letting them improvise using the micro mechanics. Will it work? I don't know , I'm betting its all going to come down to clarity.
E
Compared to Eskils' and
Compared to Eskils' and Metahominids' posts my scrawling seem rather vagrant, but this seems like an appropriate place for my query.
So first, a question. Is the implementation of super tokens complete?
And secondly, in case anyone is interested and not already aware (most of you probably are), Dwarf Fortress is a neat indie game which sets the standard for procedurally generated story. Not to knock Love, I love Love and Love does Love very well. But Dwarf Fortress is in a class of its' own in terms of computer generated storyline and if that's something you're interested in its' worth a look. If, that is, you can get past the ASCII graphics, undecipherable interface, and monstrous learning curve.