Speech Recognition
Translators - translates from one language to another
Phone booking Systems - for payment of bills, ticketing information etc
Voice Commands - for secure entry and for controlling computer systems with spoken commands
Affected by background noise, slang, different accents, if someone has the cold or a sore throat or speaks to quickly, ambigous sentences.
Must be trained to recognise your voice
Handwriting Recognition
built into palmtops and smart phones
can recognise your handwriting and turn it into typed text
Must be trained to recognise your handwriting
Expert Systems
A system that recommends advice based on the answers you give to questions it asks you on its limited domain
+++ 24/7, combined knowledge of several experts, can explain and justify its decisions, consistent results
--- narrow domain, loss of jobs, no common sense
Examples - NHS 24/7 for medical advice, Whale Watcher - identify which whale you have seen, Legal Advice ....
Artificial Neural Systems
A electronic model of the human brain made up of thousands of simple interconnected processors
must be trained to do its job.
Learning to read postcodes to sort mail
Included in vision/speech/handwriting recognition
Stock Market Trend Predictions
Weather Prediction
++ more accurate and reliable, capable of learning and improving
-- cesigned for one specific task, cant explain its reasoning
Vision Recognition
Attempt to recognise and understand images
Uses include - satellite photo interpretation for military and weather
industry to check items on assembly line, airport security to recognise criminals etc
difficulties incliude - image could be blurred, objects could be in shadow, too many objects, object could be missing/unclear/distorted or you may only see part of an image from an unfamiliar angle
Intelligent Robots
robots which dont carry out simple programming, they can think and plan for themselves, can learn from mistakes and can adapt to new situations
uses include - space exploration, under water explanation, bomb disposal
Monday, 28 March 2011
Wednesday, 23 February 2011
The Turing Test
How does the Turing Test work?
Test a machines intelliengce - to see if a machine can communicate like a human by fooling another one
A human (machine A) sits at a computer terminal and asks any questions to two other computer terminals which are in seperate rooms.
One of these terminals is controlled by another human.
The other terminal is a computer being tested for intelligence.
If the human on machine A cannot distinguish between the two the test is passed.
Flaws with the test?
only tests one aspect of intelligence
Test a machines intelliengce - to see if a machine can communicate like a human by fooling another one
A human (machine A) sits at a computer terminal and asks any questions to two other computer terminals which are in seperate rooms.
One of these terminals is controlled by another human.
The other terminal is a computer being tested for intelligence.
If the human on machine A cannot distinguish between the two the test is passed.
Flaws with the test?
only tests one aspect of intelligence
Monday, 21 February 2011
Artificial Intelligence
Definition of Intelligence
"involves knowing and choosing"
Aspects of Intelligence
- Ability to Learn
- Ability to recall/remember information
- Problem Solving
- Adapating to new situations
- Reasoning and the ability to explain
Definition of AI
The Science of making machines do things that if done by humans it would require intelligence to do so
Why is it hard to define AI
AI is rapidly changing and advancing field
It has an unlimited scope
There are several groups working on AI, biologists, computer scientists, psychologists - they all see AI from their own perspective of the subject
"involves knowing and choosing"
Aspects of Intelligence
- Ability to Learn
- Ability to recall/remember information
- Problem Solving
- Adapating to new situations
- Reasoning and the ability to explain
Definition of AI
The Science of making machines do things that if done by humans it would require intelligence to do so
Why is it hard to define AI
AI is rapidly changing and advancing field
It has an unlimited scope
There are several groups working on AI, biologists, computer scientists, psychologists - they all see AI from their own perspective of the subject
Monday, 10 January 2011
Knowledge Bases and Semantic Nets
leadSinger(PersonX):-
is_a(PersonX, pupil),
likes(PersonX, karaoke),
voice(PersonX, amazing).
Rule with three SUB GOALS
For the overall goal to be true -> all subgoals must be true
The name of the diagram which allows us to visualise the relationship between the facts in a knowledge base is called ..... SEMANTIC NET
is_a(PersonX, pupil),
likes(PersonX, karaoke),
voice(PersonX, amazing).
Rule with three SUB GOALS
For the overall goal to be true -> all subgoals must be true
The name of the diagram which allows us to visualise the relationship between the facts in a knowledge base is called ..... SEMANTIC NET
Friday, 7 January 2011
Prolog
Programmable Logic
Allows to define FACTS and RULES
These collected together are called a knowledge base
To represent the fact
"McAlpine loves Binary" - loves(mcalpine, binary). (intentionally lower case)
This only works one way!!!
To represent the fact
"I have a spearmint silver car, 03 plate, 3 door, petrol"
car(spearmintSilver, 03, 3, petrol).
male(homer).
parent(homer, bart).
to make a rule for dad ????
dad(DadX, Child):-
male(DadX), 'the person dadX must be male to be a DAD!
parent(DadX, Child).
Allows to define FACTS and RULES
These collected together are called a knowledge base
To represent the fact
"McAlpine loves Binary" - loves(mcalpine, binary). (intentionally lower case)
This only works one way!!!
To represent the fact
"I have a spearmint silver car, 03 plate, 3 door, petrol"
car(spearmintSilver, 03, 3, petrol).
male(homer).
parent(homer, bart).
to make a rule for dad ????
dad(DadX, Child):-
male(DadX), 'the person dadX must be male to be a DAD!
parent(DadX, Child).
Subscribe to:
Posts (Atom)