Transcript Revision Lectures - School of Computer Science

Introduction to AI
&
AI Principles (Semester 1)
REVISION LECTURES (Term 3)
John Barnden
Professor of Artificial Intelligence
School of Computer Science
University of Birmingham, UK
TODAY: Nature of Exam; Review of the term
TOMORROW: Question/answer session.
Nature
of the Examination
Format
 One and a half hours (for Intro to AI).
(Half of AI Principles exam.)
 Four questions, one containing choice.
 Suggestion: 10-15 minutes for initial read-through and thinking, then
up to about 15 minutes for answering each question, leaving about 15
minutes for final checking/refining.
 Some questions have several parts.
 Some questions broadly be similar in style to some questions in
formative Exercise Set Exercise, though simpler/briefer.
 One or two brief essay style questions/parts requiring you to recall
concepts, issues, examples, etc. from module material.
 Some questions/parts quite technical, others not.
Material, 1
My own lecture material.
Bullinaria slides pointed to from my list of weekly slides.
NB: This now includes his slides on Neural Networks (ask
me if you need any help understanding them).
Material from all Guest Lectures.
Chapters in the Weekly Reading Assignments on module
webpage.
Answers to the formative Exercise Set in Term 1.
Material, 2
Don't be spooked by previous examinations!!
My coverage of material is new.
Knowledge of textbook chapters other than those I've listed
isn't expected.
Knowledge of Bullinaria slides other than those I point to
from my list of weekly lecture slides isn’t expected.
Knowledge of fine technical details in guest lectures and
book chapters won't be expected. (Only expecting the main
concepts and overall grasp of main examples.)
But of course knowledge of all the above types could be
helpful and impressive.
REVIEW
of the material
(refinement of part of a Week 11
lecture)
Review, 1
Nature of AI: aims, applications, branches, issues.
Difference from CS in general.
“Intelligence” and its connection to “stupidity”.
Expert AI versus Everyday (“Common-Sense”) AI.
Why everyday AI is difficult.

Language processing, vision, planning, common-sense reasoning,
etc.
Review, 2
Why planning, common-sense reasoning, language
processing, etc. may need representation.
Why natural language is problematic for this … while also
having many strengths.
What we need to represent: entities (incl. situations,
feelings, …), properties, relationships, groups, propositional
structure, generalization/quantification, …
Types of reasoning we need to do.
Review, 3
 Taster of logic.

Captures entities, properties, relations, extreme forms of
quantification, basic forms of propositional structure. Can also
handle groups of entities.

Aims of logic: clarity and simplicity compared to NL; systematic,
sound reasoning; general applicability; common format for
comparison.
Intro to semantic networks (and frames).
Production systems.
Review: Guest Lectures
Chess, Computer games (NB: similarities, differences)
Learning, Neural networks
Evolutionary computing
Vision
Robotics, Agents
Philosophy
Review: General Themes in AI
Uncertainty, vagueness, conflict, missing info, diversity of
info.
Hence: satisficing, graceful degradation, heuristic
processing (i.e., using rules of thumb).
Context-sensitivity; relativity to agents’ purposes.
Task variability, learning, adaptation, repair (e.g., of plans).
Representation.
Reasoning.
Search.