Agent - Amazon S3

Download Report

Transcript Agent - Amazon S3

Important This Week
•
Important this week:
• Register for the class on edX
• Do the math self-diagnostic if you have any doubts whatever
• Do P0: Python tutorial (due on Wednesday 9/3 at 5pm)
• Do HW1: Agents and uninformed search (due on Monday 9/8)
• Register for the class on Piazza --- our main resource for discussion and communication
•
• Get (optional) account forms in front after class
Also important:
• Sections start on 9/8. Please attend your own section; you can attend another only if there is space.
• If you are wait-listed, you might or might not get in depending on how many students drop. Contact
Michael-David Sasson ([email protected]) with questions.
• Office hours today at 3.30 in 730 Sutardja Dai Hall (access via west stairs)
CS 188: Artificial Intelligence
Introduction, contd.
Instructor: Stuart Russell
Future
 We are doing AI…
 To create intelligent systems
 The more intelligent, the better
 To gain a better understanding of human intelligence
 To magnify those benefits that flow from it
Future, contd.
 Progress is accelerating, partly due to an industry arms race
 Once performance reaches a minimum level, every 1%
improvement is worth billions





Speech
Text understanding
Object recognition
Automated vehicles
Domestic robots
What if we do succeed?
 “The first ultraintelligent machine is the last invention that
man need ever make.” I. J. Good, 1965
 Might help us avoid war and ecological catastrophes, achieve
immortality and expand throughout the universe
 Success would be the biggest event in human history …
 and perhaps the last
Reasons not to worry
 “AI will never reach human levels of intelligence”
 “OK, maybe it will, but I’ll be dead before it does”
 “Machines will never be conscious”
 Consciousness isn’t the problem, it’s competence!
 “We design these things, right?”
 Yes, and the genie grants three wishes
 For almost any goal, a superintelligent system will…
 Acquire as many resources as possible and improve its own algorithms
 Protect itself against any attempt to switch it off or change the goal
Precedent: Nuclear Physics
 Rutherford (1933): anyone who looks for a source of power
in the transformation of the atom is talking moonshine.
 Sept 12, 1933: The stoplight changed to green. Szilárd
stepped off the curb. As he crossed the street time cracked
open before him and he saw a way to the future, death into
the world and all our woes, the shape of things to come.
 Szilard (1934): patent on nuclear chain reaction; kept secret
So, if that matters…..
 Along what paths will AI evolve?
 What is the (plausibly reachable) best case? Worst case?
 Can we affect the future of AI?
 Can we reap the benefits of superintelligent machines and avoid the risks?
 “The essential task of our age.”
Nick Bostrom, Professor of Philosophy, Oxford University.
CS 188: Artificial Intelligence
Agents and environments
Instructor: Stuart Russell
Outline





Agents and environments
Rationality
PEAS (Performance measure, Environment, Actuators, Sensors)
Environment types
Agent types
Agents and environments
Agent
Sensors
Environment
Percepts
?
Actuators
Actions
 An agent perceives its environment through sensors and acts upon
it through actuators (or effectors, depending on whom you ask)
Agents and environments
Agent
Sensors
Environment
Percepts
?
Actuators
Actions
 Are humans agents?
 Yes!
 Sensors = vision, audio, touch, smell, taste, proprioception
 Actuators = muscles, secretions, changing brain state
Agents and environments
Agent
Sensors
Environment
Percepts
?
Actuators
 Are pocket calculators agents?
 Yes!
 Sensors = key state sensors
 Actuators = digit display
Actions
Agents and environments
Agent
Sensors
Environment
Percepts
?
Actuators
Actions
 AI is more interested in agents with substantial computation
resources and environments requiring nontrivial decision making
Agent functions and agent programs
 The agent function maps from percept histories to actions:
 f : P*  A
 I.e., the agent’s response to any sequence of percepts
 The agent program l runs on some machine M to implement f :
 f = Agent(l,M)
 Real machines have limited speed and memory
 The program may take time to choose actions, may be interrupted by new
percepts (or ignore them), etc.
 Can every agent function be implemented by some agent program?
 No! Consider agent for halting problems, NP-hard problems, chess with a slow PC
Example: Vacuum world
A
B
 Percepts: [location,status], e.g., [A,Dirty]
 Actions: Left, Right, Suck, NoOp
A
B
Vacuum cleaner agent
Agent function
Percept sequence
Action
[A,Clean]
Right
[A,Dirty]
Suck
[B,Clean]
Left
[B,Dirty]
Suck
[A,Clean],[B,Clean]
Left
[A,Clean],[B,Dirty]
Suck
etc
etc
Agent program
function Reflex-Vacuum-Agent([location,status])
returns an action
if status = Dirty then return Suck
else if location = A then return Right
else if location = B then return Left
What is the right agent function?
Can it be implemented by a small agent program?
(Can we ask, “What is the right agent program?”)
A
B
Rationality
 Fixed performance measure evaluates the environment sequence
 one point per square cleaned up?
 NO! Rewards an agent who dumps dirt and cleans it up
 one point per clean square per time step, for t = 1,…,T
 A rational agent chooses whichever action maximizes the expected
value of the performance measure
 given the percept sequence to date and prior knowledge of environment
Does Reflex-Vacuum-Agent implement a rational agent function?
Yes, if movement is free, or new dirt arrives frequently
Rationality, contd.
 Are rational agents omniscient?
 No – they are limited by the available percepts
 Are rational agents clairvoyant?
 No – they may lack knowledge of the environment dynamics
 Do rational agents explore and learn?
 Yes – in unknown environments these are essential
 So rational agents are not necessarily successful, but they are
autonomous (i.e., transcend initial program)
 [dung beetle video]
The task environment - PEAS
 Performance measure
 -1 per step; + 10 food; +500 win; -500 die;
+200 hit scared ghost
 Environment
 Pacman dynamics (incl ghost behavior)
 Actuators
 Left Right Up Down
 Sensors
 Entire state is visible
Note: formal evaluation of an agent
requires defining a distribution over
Instances of the environment class
PEAS: Automated taxi
 Performance measure
 Environment
 Actuators
 Sensors
PEAS: Automated taxi
 Performance measure
 Income, happy customer, vehicle costs,
fines, insurance premiums
 Environment
 US streets, other drivers, customers
 Actuators
 Steering, brake, gas, display/speaker
 Sensors
 Camera, radar, accelerometer, engine
sensors, microphone
PEAS: Medical diagnosis system
 Performance measure
 Patient health, cost, reputation
 Environment
 Patients, medical staff, insurers, courts
 Actuators
 Screen display, email
 Sensors
 Keyboard/mouse
Environment types
Pacman
Fully or partially observable
Single-agent or multiagent
Deterministic or stochastic
Static or dynamic
Discrete or continuous
Known or unknown
Backgammon
Diagnosis
Taxi
Agent design
 The environment type largely determines the agent design





Partially observable => agent requires memory (internal state)
Stochastic => agent may have to prepare for contingencies
Multi-agent => agent may need to behave randomly
Static => agent has time to compute a rational decision
Continuous time => continuously operating controller
Agent types
 In order of increasing generality and complexity




Simple reflex agents
Reflex agents with state
Goal-based agents
Utility-based agents
Simple reflex agents
Agent
Sensors
What the world
is like now
Environment
Condition-action rules
What action I
should do now
Actuators
Pacman agent in Python
class GoWestAgent(Agent):
def getAction(self, percept):
if Directions.WEST in percept.getLegalPacmanActions():
return Directions.WEST
else:
return Directions.STOP
Pacman agent contd.
 Can we (in principle) extend this reflex agent to behave well in all
standard Pacman environments?
Handling complexity
 Writing behavioral rules or environment models more difficult for
more complex environments
 E.g., rules of chess (32 pieces, 64 squares, ~100 moves)
 ~100 000 000 000 000 000 000 000 000 000 000 000 000 pages
as a state-to-state transition matrix (cf HMMs, automata)
R.B.KB.RPPP..PPP..N..N…..PP….q.pp..Q..n..n..ppp..pppr.b.kb.r
 ~100 000 pages in propositional logic (cf circuits, graphical models)
WhiteKingOnC4@Move12  …
 1 page in first-order logic
x,y,t,color,piece On(color,piece,x,y,t)  …
Reflex agents with state
Sensors
State
How the world evolves
What my actions do
Condition-action rules
Agent
What action I
should do now
Actuators
Environment
What the world
is like now
Goal-based agents
Sensors
State
What the world
is like now
What my actions do
What it will be like
if I do action A
Goals
What action I
should do now
Agent
Actuators
Environment
How the world evolves
Utility-based agents
Sensors
State
What my actions do
What it will be like
if I do action A
Utility
How happy I will be
in such a state
What action I
should do now
Agent
Actuators
Environment
How the world evolves
What the world
is like now
Summary
 An agent interacts with an environment through sensors and actuators
 The agent function, implemented by an agent program running on a
machine, describes what the agent does in all circumstances
 PEAS descriptions define task environments; precise PEAS
specifications are essential
 More difficult environments require more complex agent designs and
more sophisticated representations