lecture 1 Flashcards
agent
perceiving its environment (throgh sensors) and acting upon the envirnment (through actuators)
human agent
environment: the world in which we are embodied.
sensors: eyes, ears, nose.
actuators : hands, legs, voice.
robotic agent
environment: the (part of the) world in which it is embodied.
sensors: (infrared) cameras, mic, temp sensors.
actuators: various motors and output devices.
sowftware agent:
environment: software that it interacts with, including IO.
sensors: keystrokes, networl packages, file contents.
actuators: writing to files, screen display.
Genral AI
building machines that can perform any intelectual task as well as or better the humans
Narrow AI
building machines that can perform a specific intellectual task very well.
Physical symbol system hypothesis
A physical system has the necessary and sufficient means for general intelligent action.
- A symbol is a meaningful physical pattern that can be manipulated.
- A symbol systen creates, copies, modifies and destroys symbols.
The chinese room arguement
Assuming that a computer can fool us into thinking its human, does that mean that it is intelligent.
“The computer doesn’t think, it just follows rules.
optimal solution
best solution according some measure of solution quality
satisficing solution
one that is good enogh, according to some description of which solutions are adequare
approx. optimal solution
one whose measure of quality is close to the best theoretically possible
probable solution
one that is likely to be a solution.
rational agent
Goal is to find optimal solution
bounded rationality
Goal is to find a solution that is as best as possible given limitations (time, memory, energy)
Abilities
the set of possible actions it can perform
goals
what it wants, its desires, its value
prior knowledge
what it comes into being knowing, what it doesnt get from experience
history of stimuli
current stimuli - what it recieves from the environment now (observation, percepts)
past experiences - what it has received in the past.
body and controller
an agent interacts with the environment through its body.
the body id made up of: *sensors that interpret stimuli
*actuators that carry out actions.
- the controller receives percepts from the body and the controller sends commands to the body.
controller
the controller is the brain of the agent.
agents are situated in time, they receive sensory data in time, and do actions in time.
controleers have (limitied) memoryy and (limited) computational capabiltiies.
the controller specifies the command at every time.
the comman at any time depend on the current and previous percepts (its history).
Belief state
an agent doe’snt have access to its entire history., it only has access to what it has remembered.
the memory or belief state of an agent at times encodes all of the agent’s history that it has access to.
the belief state of an agent encapsulates the info avbout its past that it can use for current anf future actions.
at every time a controller has to decide on: what should it do?
what should it remember?
flat
one level of abstraction
modular
interacting modules that can be understood separately
hierarchical
moduls that are (recursively) decomposd into modules
- flat representations are adequate for simple systems.
- complex bio systems, computer systems, organizations are all hierarchocal
kind of agents
- dead reckoning agent: doesnt prceive the world
- reactive agent: doesnt have a belief state.
- model free agent: agent maintains belief state that is used to maximize utility/reward.
- model based agent: agent uses model of the world as belief state to plan optimal actions.
