vimwiki/tech/AI.wiki

= AI =

AI or artifical intelligence is the art of making intelligent machines.

== Goals ==

* Expert systems
  * exhibit itlligent behaviours in one field, (IE chess AIs beat humans)
* Human intelligence
  * behave, think, and learn like humans

== Types ==

* AI
  * Machines which mimic human behaviour
  * not even close as of currently
* [[machine_learning]]
  * Allow machines to find statistical relationships in data to mimic patterns
* Deep learning
  * Subset of Machine Learning, makes multilayer neural networks feasible

== Agents ==

An agent is anything that perceives its enviroment thorugh some sensors and
acts upon the enviroment throught some actuators.

Some types of agents

* Simple reflex
  * Make decision based on current state of percepts, ignores previous state
  * Only work in fully observable enviroment
* Model-based reflex agent
  * Create a model of what is happening in the world, and based upon this
    model, make decisions
  * Work in partially obervable enviroment
* Goal based agents
  * Knowldege of enviroment is not always best info to know what to do
  * Agent knows its goal and desirable situations
  * Expansion of model based agent by having a goal to work to as well as a
    model of the world.
* utility agesnts
  * A goal based agent, but have a utility score that meansures how close the
    given state is to the end goal
  * Act based on goal as well as best way to get to said goal.
* learning agents
  * Agent taht learns from previous experiences. It starts to act with basic
    knowledge and then able to act and adapt automatically through learning

Example, taxi driver,

* Precepts
  * Cameras, GPS, microphone, etc
* Action
  * Steet, accelerate, break, etc
* Goal
  * safe, fast legal comfortable trip, maxmize profit
* Enviroment
  * Roads, traffic, pedestrians, customers

Any part of the enviroment that is fully observable is called a fully obersable
enviroment. Else it is partially observalble. Most aspects of the enviroment
are only partially observable.

A stocahstic enviroment is random in nature and cannot be determined completely
by an agent.

In a deterministic enviroment that is fully obersvable, agents do not need to
account for random chance.

If the enviroment can change while the agent is deliberating/computing what to
do, then the enviroment is dynamic. Otherwise, it is static. Static enviroments
are much easier to deal with, because an agent must not constantly deliberate
on when to change course suddenly.

If the number of possible states that can be percieved is bounded, then the
enviroment is called a discrete enviroment, otherwise it is a continous
enviroment. A chess game is discrete, because there is a finite number of moves
that cna be played.

If there is only one agent in the enviroment, then it is called a single agent
enviroment. HOwever if mulitple agents are operating in the enviroment, then it
is a multi agent enviroment. The design problems are much harder in a
multiagent enviroment.