What is AI?
- Thinking Humanly
- Thinking Rationally
- Acting Humanly
- Acting Rationally
What does it mean to act rationally?
Do the right thing, often in AI means maximizing reward or minimizing error
What is an agent?
An agent is anything that is perceiving it’s environment and using sensors to gather data and actuators to act on the data
Fully Observable vs Partially Observable
if agent’s sensors are fully aware of the environment then it’s fully observable otherwise it is partially observable
Example of fully observable environment?
Crossword puzzle
Example of partially-observable environment?
Poker
Single agent vs multi-agent
We say an environment is multi-agent if we have multiple agents acting upon it
Example of single agent environment?
Crossword puzzle
Example of multi-agent environment?
Chess
Deterministic vs Schoastic
- We say an environment is deterministic if next state is only dependent on current state and agent action.
- It is scholastic is the next action is dependent on other agent actions or environment changing
Example of deterministic environment?
Crossword puzzle
Example of scholastic environment
Poker
Episodic vs Sequential
- Episodic environment is an environment where the agent’s experience can be broken down into it’s own episode.
- Sequential is different because it needs lookahead to perform correct actions.
Example of episodic environment?
Part picking robot
Example of sequential environment?
Crossword puzzle
Dynamic vs Static
- Dynamic environment is an environment that is constantly changing while agent is deliberating
- Static environment is environment that does not change.
Example of dynamic environment
Self-less driving cars
Example of static environment
Poker
Discrete vs Continuous
- Discrete environments can be described using states
- Continuous environments can be described using time
Example of discrete environment?
Chess
Example of continuous environment?
Self-less driving cars
What is PEAS?
P - Performance
E - Environment
A - Actuators
S - Sensors
What are agent types?
- Simple reflex agent
- Reflex agents with state
- Goal based agent
- Utility-based agent
- Learning agents
Can all agent types other then learning agents be transformed into learning agents?
True
