Foundations

Getting a Handle on AI
- Some handwaving ideas [Coppin][RN][Luger]
  - Human(-like) results or not (superset? intersection?)
  - Human(-like) techniques or not (yes => human(-ilke) results) (no can still appear human(-like), or get different results)
  - Human(-like) hardware or not
  - The Turing test ... making a computer appear human ... ignore the interface.
- Dispelling the artificial myth ... computational intelligence ... using computers, any techniques, any results.
- Getting a grip on "intelligence"
  - Ambiguous sentences
    - Problem representation
    - Choices about words heard
    - Choices about word meanings (interpretation, semantics)
    - Choices about sentence meanings (interpretation, semantics)
    - Knowledge helps resolve choices
  - Someone who lives in Dreadbury Mansion killed Aunt Agatha. Agatha, the butler, and Charles live in Dreadbury Mansion, and are the only people who live therein. A killer always hates his victim, and is never richer than his victim. Charles hates no one that Aunt Agatha hates. Agatha hates everyone except the butler. The butler hates everyone not richer than Aunt Agatha. The butler hates everyone Aunt Agatha hates. No one hates everyone. Agatha is not the butler. Who killed Aunt Agatha?
  - Class examples of intelligence
  - Capturing Fanta
- Solving "hard" problems as intelligence
  - Hard problems
    - Finding the Jamba Juice shop
    - O(E) and O(N^k) problems
  - Solving easy O(N^k) problems is easy.
  - Solving hard O(E) problems requires intelligence.
    - Random leaps of faith = O(E)
    - Exhaustive search = O(E)
    - With intelligence = O(N^k)
  - Intelligence is the ability to solve O(E) problems with O(N^k) resources,
    through the use of guided search (making hard problems look easy)
  - "the ability to solve problems is generally taken as a prime indicator that a system has intelligence" [NS p.X]
  - (Some) humans can make good decisions, i.e., display intelligence. The Turing test is thus sound, but not complete.
- Solving hard problems - what (computing) techniques are necessary?
  - A problem has three components: Initial states, solution detector, state generator
  - Representing the problem requires computer science, but no intelligence per se.
  - Solving the problem by guiding the search is the intelligence
  - Knowledge is "required" to guide the search
    - Knowledge in terms of the symbols (syntactic)
    - Knowledge as a model of the environment (semantic)
    - Knowledge is not intelligence, but is "required" for intelligence.
An Intelligent Agent
- Senses a problem (often requires intelligence)
- Generates a solution (always requires intelligence)
  - Problem representation
  - Search for a solution
  - Knowledge for search guidance
- Actuates the solution (occasionally requires intelligence)
- The components of an intelligent agent individually may use AI, e.g., NLP sensors, AR over knowledge, robotic actuators.
- The environment
  - An intelligent agent exists in an environment that contains problems that the agent solves.
  - The complexity of the environment may stimulate or require complex (intelligent) activity.
  - Provides the meaning for all problem solving activity
  - Is a source of knowledge that may be extracted by the sensors
Emergent intelligence: Societies of agents
- Colonies exhibit intelligence (brains, ants, bakers, stock market) with limited individual agent intelligence
- Autonomous agents with specific capabilities
- Situated agents are aware of their part of the environment
- Communicative agents interact productively
- Structured societies are coordinated
- Societies are emergent - the whole is more intelligent than the sum of the parts.
- Need CS tools for these aspects

The Physical Symbol System Hypothesis - [CACM 19(3), pp.113-126]

Structure and composition
- Symbols
- Symbol structures
- Wider world of objects (physical and conceptual), including processes that manipulate symbol structures
Symbol structures designate processes and objects.
- The system can affect or react to the designated process or object.
- Processes and objects are different, but are designated by the same things (symbol structures).
- The thing that a symbol structure designates is determined by a separate mechanism. For symbol structures, the mechanism considers the arrangement of its constituent symbol structures.
- Designation provides the semantics of the symbol structures.
Interpretation
- A symbol structure designates a process by being the code that represents the process
- A symbol structure that designates a process can be executed
Completeness and closure
- A symbol may designate any object. Symbol structures designate objects according to their arrangement.
- Symbol structures are Turing complete
- There are processes for creating and manipulating symbol structures arbitarily.
- Symbols and symbol structures are stable.
- The system can hold infinite symbols and symbol structures (in principle).
The hypothesis: A physical symbol system has the necessary and sufficient means for general intelligent action.
- Any system that is intelligent turns out to be a physical symbol system.
- A physical symbol system of sufficient size can be organized to be intelligent.

Foundations [RN Ch.1.2]

Philosophy: algorithms, materialism, induction, backward reasoning
Mathematics: logic, computability, intractability, probability
Economics: utility, satisficing
Neuroscience: neurons, minds
Psychology: cognitive psychology, agents
Computer engineering: hardware performance, memory
Linguistics: Chomsky's theory, knowledge representation

AI areas [Lug Ch.1.2]

Natural Language Processing
Vision
Automated Reasoning
Game Playing
Planning
Knowledge Representation
Machine Learning
Robotics

Exam Style Questions

Give a definition of (artificial) intelligence in terms of computational complexity.
What are the three components of a classic definition of a problem?
Describe the operation of the Turing test as a criteria for success in artificial intelligence.
Draw a labelled diagram showing the components of an intelligent agent.
What facet of Philosophy suggests that the mind operates according to physical laws? What implications does this have for AI?
Describe how the environment affects the activities of an intelligent agent.
Define "emergent intelligence".
In the context of a society of agents, define "situated agents", "communicative agents", "structured society".
The Physical Symbol System hypothesis states that "A Physical Symbol System has the necessary and sufficient means for general intelligent action." What is the structure of a Physical Symbol System?
In the context of physical symbol systems, what are meant by "designation" and "interpretation"?
Describe the five requirements of completeness and closure for a physical symbol system.
Name three important contributions to AI from Mathematics.
Describe how the rapid improvements in computer hardware have advanced the capabilities of AI systems.
Name and briefly (maximum 20 words each) describe N areas of AI.