tibA.I. Status

[jcarley]

Status of tib A.I.
Source Hosting: http://code.google.com/p/tibai/
Tibia Client Support: 8.5

Current Functions:
Pathfinder: In progress
- Local Search: Complete
- Illegal Tile Mapping/Avoidance: Complete
- Tibia Wide Search: In development
- Intelligent Heuristic: In development
- Multi-threaded searches: Comming Soon!
- Advanced Diagnostic Interface: Comming Soon!

Combat Brain: Comming Soon!
- Opponent Database: Comming Soon!
- Unguided Learning Algorithm: Comming Soon!
- Danger Assessment: Comming Soon!
- Intelligent Planning: Comming Soon!
- Terrain Assessment: Comming Soon!
- Combat Strategy: Comming Soon!
- Multiple Enemy Tactics: Comming Soon!

Social Brain: Comming Soon!
- IAC Concept Maps: Comming Soon!
- Multi-Target Conversations: Comming Soon!
- Acquaintence Memory: Comming Soon!
- Conversation Scripts: Comming Soon!
- Concept to Sentence Translation : Comming Soon!
- Humanization by Planned Errors: Comming Soon!
- Intelligent Topic Avoidance: Comming Soon!

A* Strategizer: Comming Soon!
- More info comming soon.

[ame]

Quite the list there!

Just make sure you focus on the "important" bits. Ask on the TibiaAPI if you need some functionality added.(If you think it'd be of any use for someone else;))

[DarkstaR]

I cant wait to see what happens with this bot
Sounds amazing and just by reading what you've said about it in other threads, I'm sure you'll accomplish it :)

[OsQu]

Looks GOOOOOOD! [pronounce with Jim Carry voice from Bruce Almighty]

If you need help with anything (TibiaAPI or any other stuff) don't hesitate to ask it! I'm fully supporting this project.

[CristoferMartins]

grats for the idea jcarley.

[jcarley]

Thanks. I'm still developing the particular strategy I'm going to take to get all of this accomplished. It's still a long ways from being completed but I think once the combat and pathfinding portions are completed I'm going to try and get some help to improve and extend that part. The fact that it will be functional as a cavebot will generate more interest in this project as well.

BTW I just got off a 40 hour solo road trip that afforded me a great deal of time to think some things out. I think I came up with a really sweet way to assemble super long strategy paths using objects that hold the quantified state of the AI including the relevant qualities of its controlled character and a modified A* algorithm. Also I figured out a really elegant, almost biological, way to implement a need system that could take advantage of unforeseen events. I planned out a way to formulate any subject in tibia into an IAC(Interactive Activation and Comptetition Network) concept map that could easily be expanded to allow for potential actions and the translation of those actions and intentions resulting in sentences for use in a conversation (enough to eliminate suspicion as a bot at least). And lastly I decided on a final goal for my pathfinder which led to a method for the logical arrangement of the code for the whole project.

It was an exciting couple of days. I'll post more on these subjects in seperate threads as time permits. I have to sleep now though.

[jcarley]

Just an update:

I've had to take a short break from this project as you may have noticed. My time constraints came to a critical threshold and something had to be removed. Tibia programming pulled the short straw. I definately plan to resume this project but I first have to take care of some things to free up some time. Thanks for your patience.

Before I stopped I reached a point where I could finally visualize the final product with all its connecting parts. That 'vision' made me realize that some portions of my code need to be re-evaluated. As a result I plan to go back and recode some portions of the existing project. I'll let you know when I have this figured out. Thanks.

[jcarley]

I'm starting up development on this project again. Updating tibiaAPI as we speak.

[Jesper_Kladden]

I love you JCarley ... <3

[jcarley]

so does your mom

[jcarley]

My sources have been updated to work with Tibia 8.41. All it does for now is walk. But it walks like a champ. Test it out. It can go to any location that doesn't require doors or stairs. Go to carlin and try the XYZ set. Try it with no map files and try it with full maps. It will also learn where it can't go such as other peoples houses.

[jcarley]

I freed up time to work on this the past 3 days.
I added a thread to facilitate datamining in the background. The data mining process draws conclusions from other information for use by the A.I. It currently generates connections between 16x16 clumps of tiles i call macro-tiles. This information is used to give a more accurate heuristic when path-searching. I believe this will reduce the search area by nearly 1000% on average.

Accurate mining of the macro-tile data is functioning properly
I'm now developing the heuristic function- (its tough because you never can be sure if the macro-tile data is up to date)
next is to create a two way path search - (this will prevent very long searches if the destination cannot be reached)

I'll also submit a code update soon

[muttley]

Is project still alive?

[jcarley]

Is project still alive?


it is, it is
its just been so much work, this project is very close to my heart and I'm working tirelessly on developing the AI but its been choked out a little by my work projects. I am definately still developing this project though

[jcarley]

My main developmental focus right now is on organizing data based around grammar. The goal is to teach my A.I. to learn english.

Theory: We neurologically store information in a way that mirrors the syntax of our language. Take the following sentence as an example:

"All trees have trunks and branches but only some trees have leaves."

This sentence is informative. A child could take this information and use it to extrapolate other conclusions or questions. This information is stored based on its relationship to other information. The different types of connections are effectively abstract concepts that allow us to develop more complicated thought. In the sentence above a relationship is formed between the tree concept and the trunk concept in establishing that trunks and branches are components of all organic trees. The tree concept also becomes a base concept and two child concepts are formed based on this information that of leaf-less trees and leaved trees. The tree concept then is largely enhanced. The A.I. mind knows it has 1 trunk and potentially multiple branches it also knows that there are at least two types of trees that can be distinguished based on if they have leaves or not. The written word 'tree' is connected as well.

By arranging all of the A.I.'s knowledge in this fashion I believe I can facilitate language. A thought may be seeded from the desire for experiece. The A.I. begins to try to satisfy that desire. It determines that the ideal hunting scenario is to be using its two handed sword to fight slimes rots and orcs in folda. It compares the difference between its current state, being on the top floor of the depot in Thais with its sword and shield, and its desired state, being in the hunting spot in folda with its two handed sword. Changing one thing at a time it determines the best way to create the scenario. If in this scenario it was asked "What are you doing" the following would result. The action 'doing' is applied grammatically to the concept 'you'. The concept 'you' directed at the AI is the same as the concept 'I' as originating from the AI. Therefore the collective concept I/you's doings/plans/goals is the subject of the sentence. 'What' establishes how to treat the sentence. It is interrogative in that it is requesting information on the subject. The appropriate concepts are lit up. First the concept relating to the AI's person. 'doing' in this sentance is a component concept of 'you'. The AI has a small understanding of itself which includes its complex current activity which happens to be following the steps of this plan it developed to hunt in folda. The word doing is 'lit-up' as well. These two words must connect along a compatible series of conncetions. The base 'do' of 'doing' is connected to the plans, actions, and activities concepts which are components of the entity concept which is a base of every person including the AI which is connected to 'you'. Its established then: The AI's activity is the subject therefore the source of the material for the response comes from the plan queue of the AI. The goal concept is what is returned which is 'going hunting'. If that subject becomes the focus of future questions then it can be broken down into smaller components.

The short-term memory will allow complex concepts to be formed into single concepts temporarily. The concept of going to folda can be reduced to a single 'neuron' with its appropriate connections that detail the plan and everything associated with that. This neuron can be simplified later by removing unnessary connections and stored in longterm memory or it can be deleted altogether.

Each concept is effectively a neuron. Each neuron has a life span that extends based on the lifespans of the neurons it is connected to and the number of times it has been accessed(if it has been repeatedly accessed its very likely that it will be accessed again in the future). It also has a number of incomming connections which connect to concepts that it is dependant upon and a number of outgoing connections that connect to concepts that depend upon it. If long term information about a tree is stored then the concept that forms the source of that information retains an extended lifespan as well. The concept that forms the person that shared the information would be extended as a result. Every time that neuron is accessed the lifespan is extended based on the number of outgoing connections it has(how many other concepts depend upon it). When a neuron is destroyed the connections it has to other neurons is destroyed as well. If a neurons last connection is destroyed it automatically is destroyed also.

In this fashion a semblance of intelligence actually begins to appear. This was very long but very critical to what I am doing.

[jeremic]

If you can model the idea of success, neural networks can achieve almost anything! Nice project you have going by the way.

[DarkstaR]

This is part of what made me want to study Computational neuroscience in college, tbh.

[jcarley]

So this project has taken a back seat to my work for the past few months due in part to the fact that I'm reworking my theories on concept formation in the human brain.

Its based around the idea that a single neuron could be the conceptual nexus for any concept no matter how complex. Synapses are effectively logical connections. My running theory is that every type of logical relationship that can be made is represented synaptically. Those neuronic concepts are connected with logical synapses to other concepts. The sum of these connections defines the new concept.

Its like a very fundamental type of inheritance. Concepts built on other more basic concepts. But inheritance is only one type of relationship. For example a lemon tree would be a base concept of the lemon tree in my back yard. But the lemon tree in my back yard also has a possessive relationship to its specific lemons which taste better than most lemons.

Needless to say that has become some complicated that I've had to take a second look at some fundamental theories in order to understand the rules that let these relationships form.

I think its based on perceptions. I think the most basic concepts are formed around what we take in through our senses and then language builds logic into that world. Our conciousness sets at the highest possible level at all times so as soon as a foggy mess of thousands of neurons gets organized so as to be represented by just a few then our consious thought just uses those rather than work with all the other stuff.

I think a shortcut is in language. Within the rules of grammer exist the ability to describe any possible type of thought or progression of thoughts. I think a closer study of both of these concepts will give me clues as to how to proceed from here.

I've got work to do on this still.
Consider yourselves updated.

[markdefer]

project dead?
seems like a very ambitious one :)

sorry for necro, really curious about this

[jcarley]

yeah pretty much, the research lead in other directions and I didnt have time to finish

[markdefer]

yeah pretty much, the research lead in other directions and I didnt have time to finish


really sad, if complete this project would "lead the new generation" of tibia botting for sure.