Artificial Intelligence (ECE 470/ECE 569A)

Introduction to AI 1

Dr. Deepali Arora

Tuesday Jul 7, 2020

1Prepared using textbook online resources available at:

http://aima.cs.berkeley.edu
1 / 23
Outline

1 What is AI?

2 Rationality

3 Agents and Environments

4 PEAS (Performance measure, Environment, Actuators,

Sensors)

2 / 23
What is AI?

Artificial Intelligence
Methods for applying computers to problems that require “intelligence”
Study of the fundamental limits of “intelligent” behavior by computers.

What is Intelligence?

3 / 23
Acting Humanly: The Turing Test

Turing (1950) ”Computing machinery and intelligence”:

”Can machines think?” −→ ”Can machines behave
intelligently?”
Operational test for intelligent behavior: the Imitation Game

Problems:
Not reproducible
Not amenable to mathematical analysis
4 / 23
Acting Humanly: The Turing Test

Abilities Required for Turing Test

• Natural Language Processing (understanding,

generation)
• Automated Reasoning
• Learning
• Knowledge Representation and Storage
• Vision (for “total turing test”)
• Robotics (for “total turing test”)

Problem: Tends to focus on human-like errors, linguistic tricks, etc.

Does not product useful computer programs

4 / 23
Think Like Humans: Cognitive Science

• Goal: Develop precise theories of human thinking

• Cognitive Architecture (e.g., SOAR, ACT-R)
– Software Architecture for modeling human
performance
– Describe task, required knowledge, major subgoals
– Architecture follows human-like reasoning
– Makes testable predictions: Time delays during
problem solving, kinds of mistakes, eye
movements, verbal protocols, learning rates,
strategy shifts over time, etc.
• Problems:
– Identifiability: It may be impossible to identify the
detailed structure of human problem solving using
only externally-available data. “Optimal”
performance is an excellent predictor of human
performance in most routine tasks.
4 / 23
Thinking Rationally: The Logical Approach

• Ensure that all actions performed by computer are justifiable

(“rational”)
Facts and Rules in Theorem Prover
Formal Logic

• Rational = Conclusions are provable from inputs and prior

knowledge
• Problems:
– Representation of informal knowledge is difficulty
– Hard to define “provable” plausible reasoning
– Combinatorial explosion: Not enough time or space to
prove desired conclusions.

5 / 23
Acting Rationally

Rational behavior: doing the right thing

The right thing: that which is expected to maximize goal

achievement, given the available information

This course is about designing rational agents

Rational agents do the best they can given their resources

5 / 23
Rational Agents
Adjust amount of reasoning according to available resources and importance
of the result.
An agent is an entity that perceives and acts
Abstractly, an agent is a function from percept histories to
actions:

f : P∗ → A
For any given class of environments and tasks, we seek
the agent (or class of agents) with the best performance
Computational limitations make perfect rationality
unachievable
→ design best program for given machine resources

6 / 23
Agents and Environments

Agents include humans, robots, softbots, thermostats, etc.

The agent function maps from percept histories to actions:

f : P∗ → A
The agent program runs on the physical architecture to
produce f
7 / 23
Vacuum-Cleaner World

Percepts: location and contents, e.g., [A, Dirty ]

Actions: Left, Right, Suck, NoOp, Dump

8 / 23
Vacuum-Cleaner World
 agent: robot vacuum cleaner
 environment: floors of your apartment
 sensors:
 dirt sensor: detects when floor in front of robot is dirty
 bump sensor: detects when it has bumped into
something
 power sensor: measures amount of power in battery
 bag sensor: amount of space remaining in dirt bag
 effectors:
 motorized wheels
 suction motor
 plug into wall? empty dirt bag?
 percepts: “Floor is dirty”
 actions: “Forward, 0.5 ft/sec”

8 / 23
A Vacuum-Cleaner Agent

What is the right function?

Can it be implemented in a small agent program?
9 / 23
Rationality

 Performance Measure: Criteria for determining the quality of

an agent’s behavior
 Example: dirt collected in 8 hour shift

 Avoiding Omniscience
 An omniscient agent is one that can predict the future
perfectly. We don’t want this!

 Agent: Mapping from percept sequences to actions

10 / 23
Rationality

Fixed performance measure evaluates the environment

sequence
one point per square cleaned up in time T ?
one point per clean square per time step, minus one per
move?
penalize for > k dirty squares?

A rational agent chooses whichever action maximizes the

expected value of the performance measure given the
percept sequence to date

10 / 23
Rationality

Rational = omniscient
percepts may not supply all relevant information

Rational = clairvoyant
action outcomes may not be as expected

Hence, rational = successful

Rational =⇒ exploration, learning, autonomy

11 / 23
Task Environments

Task Environments are the”problems” to which rational

agents are the solutions.
Specified using PEAS (Performance measure,
Environment, Actuators, Sensors) description
Properties:
Fully observable/Partially observable
Single agent/Multiagent
Deterministic/Stochastic
Episodic/Sequential
Static/Dynamic
Discrete/Continuous
The hardest case is partially observable, multi-agent,
stochastic, sequential, dynamic, continuous

12 / 23
Agent Types

Four basic types in order of increasing generality:

1 Simple reflex agents

2 Model-based reflex agents

3 Goal-based agents

4 Utility-based agents

All these can be turned into learning agents

13 / 23
Simple reflex agents

14 / 23
Model-based reflex agents

15 / 23
Goal-based agents

16 / 23
Utility-based agents

17 / 23
Learning agents

17 / 23
PEAS

To design a rational agent, we must specify the task

environment

Consider, e.g., the task of designing an automated taxi:

Performance measure??
Environment??
Actuators??
Sensors??

19 / 23
PEAS

To design a rational agent, we must specify the task

environment

Consider, e.g., the task of designing an automated taxi:

Performance measure?? safety, profits, comfort, . . .
Environment?? streets, traffic, pedestrians, weather, . .
Actuators?? steering, accelerator, brake, speaker, . . .
Sensors?? video, accelerometers, GPS, . . .

20 / 23
Internet Shopping Agent
Examples of agent types
Agent Type P E A S

Medical Healthy patient, Patient, hospital, Display Keyboard entry

Diagnosis minimize costs, staff questions, tests, of symptoms,
lawsuits diagnoses, test results,
treatments, patient’s
referrals answers

Satellite image Correct image Downlink from Display Color pixel array
system categorization satellite categorization of
scene

Part-picking Percentage of Conveyor belt Jointed arm and Camera, joint

robot parts in correct with parts, bins hand angle sensors
bins

Interactive Maximize Set of students, Display Keyboard entry

English tutor student’s score testing agency exercises,
on test suggestions,
corrections
22 / 23
Kinds of Environments

• Fully-observable vs. Partially-observable

– Fully-observable = Markov
• Deterministic vs. Stochastic
– Strategic: deterministic except for the actions
of other agents
• Episodic vs. Sequential
• Static vs. Dynamic
• Discrete vs. Continuous
• Single agent vs. Multiagent

23 / 23
State Representation

Three ways to represent states and the transitions between

them:

23 / 23