Artificial Intelligence is a software component of a machine that learns.
Parts of AI are known, included in code, parts are unknown - to learn from available data & experiments.
Algorithm is in short a plan for a Solution, for example for 'cooking a cake' or for choosing Martial Arts style against a certain opponent.
Abstraction is in short generalization, but still on topic - with skipping part of the details for now.
As for the design:
- We wish for example to create or enrich a style to suit the needs against other style or against a certain opponent,
- We have initial style to start from or nothing at all,
- We have information about many or all of known & documented martial arts,
- We can have statistical data with styles performace against target style - including win/lose informations, amount of certain methods used (strikes, grapples, combinations, techniques, etc ...), amount of points scored with which methods as well, amount of succesful defenses & it's additional data as well,
- We can have models of combats (3D animation models) with scores & additional data as well,
- We can acquire more data in form of statistics or models, we can use even multiple models for single combat for better understanding as well (each of models can show something else, even if it's similar),
- We know the costs (measured in time to prepare mind/body & learn, for example average) of each of methods as well as our timeframe to learn the new style; we also known the 'building steps', methods that form more expensive methods as well,
- We are aware that certain longer combinations are tiring & sometimes dangerous as well,
- We analyze statistics & models to choose succesful combinations, add to proposed Solution (new style) & experiment as well. We can even generate animations (simulations) as part of experiments & learning as well.
Artificial Intelligence can help to handle large amounts of input data, categorize & make decisions based on that. We can use AI to analyze films (made simultaneously from multiple cameras at different angles) to create 3D VR model for further processing.
With proper model we can analyze fights in details, how each of moves can 'extend' into another moves, how opponent breathes, how others react to his strikes; by body reactions we can check how good his atemi & ki are.
The same we can use to analyze ourselves, to check for possible counters & defenses against these.
Then we 'know enemy' & 'know ourselves' - we can assess threats to protect against and adjust our style accordingly with counter-moves & their prerequisites.
AI can conduct experiment with models as well, simulating fights for more of statistical data.
Such data & additional software can help to plan trainings for a given person or group (against a certain opponent or against a certain style), both kihon, kata & kumite.