AI | Itgsgroupwork

The term “intelligence” is an abstract concept with definitions that varies depending on the context. According to Stuart Grey, people would describe intelligent to be things such as:

The ability to respond to environments
The ability to learn new skills or knowledge
The ability to use logic or reasoning to create conclusionsThe ability to learn from past experiences
The ability to evaluate and make judgments

In computer science however, the term “intelligence” is used as a bases for designing systems that behave similarly to humans.

What is Intelligence?

Artificial Intelligence

Artificial Intelligence, or IA, is systems that simulate human actions or decisions through a series of facts or rules. Although the method of achieving these actions are very different from that of humans, the goal of IA’s are just to achieve human-like actions.

Unlike humans, artificial intelligence often specializes in one task, such as playing chess or cleaning rooms, rather than containing a generalization of any situation like humans.

Issues and goals that arises from artificial intelligence includes the further ability to reason, planning, leaning, communication, perception and the ability to move and interact with object. Artificial intelligence is capable of accomplishing some of these tasks. However, they are only capable of accomplishing them to an extent, unlike humans.

Source (Everything else is from the regular book):

This list of intelligent traits is based on the topics covered by the major AI textbooks, including:

• Russell & Norvig 2003

• Luger & Stubblefield 2004

• Poole, Mackworth & Goebel 1998

• Nilsson 1998

Expert Systems:

Expert Systems are very similar to Artificial Intelligence. The difference between the two is that Expert System has a much stronger emphasis on answering a specific knowledge domain (one specific knowledge area) through the use to programmed logic and rules. Expert systems are capable of things such s medical diagnosis and image screening (such as x-rays), identifying agricultural pests and diseases, spell checking, finance processes and more.

An Expert System Shell is a set of programs, which allows the user to build an expert system through their own creation of logic, knowledge and rules. There are multiple components of an Expert System Shell:

User Interface: User interface presents questions and accepts input from the user. Inputs may be in various forms such as yes/no, multiple choice or text. The User Interface also presents answers once they are determined.

Knowledge base: Knowledge Based contains data and facts, which are used to form knowledge in a specific knowledge domain. The information in the Knowledge Base is intended to be a replication of a human. Because many experts in a particular field, such as doctors, are not experts with computers, this system will first have to be prepared by a knowledge engineer.

Interference Engine: The Inference Engine provides knowledge by matching the user’s input wit the data in the Knowledge Base in order to find appropriate answers. This is accomplished by using Inference Rules, which witch is how items of data relate to one another.

Expert Systems uses multiple Knowledge Domains to provide the user with information or knowledge:

Decision Trees: Decision Tree uses Boolean Logic. It is a series of questions that leak up to another until it finds the answer. Each question is answered from two choices, typically Yes/No questions. The strength of this system is that is quick to both use and create. Limitations include the lack of answers, as some questions are very hard to just answer “yes” or “no”.

Inference Rules: Inference Rules is similar to Decision Tree, however it does not follow “yes/no” logic. Instead, it uses “IF…THEN” logic. It is a process of chaining, which can gives answers according to the conditions.

Fuzzy Logic: As other types of Knowledge Domains uses concrete answers such as “yes/no”, Fuzzy logic does not. Instead of having to be as part of a set or not, Fuzzy Logic allows answers to be partially in one set. This is useful for questions such as “What is the temperature?” an answer can be between “Hot” and “Very Hot” instead of having to be “Hot” or “Very Hot”.

There are many limitations to Expert Systems due to its limited amount of answers that users can provide. Due to the limited choices of answers, users may receive false answers from using Expert Systems. Many modern IA researchers are no longer developing Expert Systems because of this issue.

Applications of AI: Artificial Examiners

This is a case study, which took place in the University of Buffalo, New York. Researchers have created software, which can read and grade exam papers. Although assessing multiple-choice answers are simple for a machine, Professor Sargar Srihari has created a system, which was tested on its ability to grade reading comprehension exams. The question for the exam was “How was Martha Washington’s role as First lady different from that of Eleanor Roosevelt?”

The system is programmed to look for key phrases and words, which are commonly found in papers given high grades by human examiners.

The system scanned the hand writing as input. It then process the hand written text into words it can understand. The system then will look for those key phrases before giving grades as output based upon the amount of the instanced of the key phrases.

The limitations of this includes the difficulty is translating the hand-written text s, as well as questions which involves evaluating, as key phrases would not be difficult to construct. This issue occurs with the difficulty of understanding context, as the system can only understand key phrases.

Pattern Reorganization

Pattern Reorganization systems can be used for, but not limited to, Speech Reorganization and Computer Vision application such as Handwriting Reorganization and Face Reorganization.

Patter Recognition systems are trained to use a set of Training Data. For Example, a system is trained to recognize handwritten text by showing samples of letters and train it to interpret them as “A” or “B”. The higher the training samples are, the more reliable the system it. The term Confident Interval refers to how certain their answer is.

Natural Language Processing

Natural Language Processing is the ability to understand or process selected human language(s). Inputting this process may take in forms such as voice commands. Outputs may include texts or repetition of what the user has said.

Some limitations of this processing are:

Understanding which meaning the user refers to as many words contains multiple meanings.
Complex rules concerning syntax and grammar.
Understanding idiom
Some words in some languages may lack direct translation. For example, in Mandarin, the word “he” and “she” are both pronounce as “ta”, even though it is spelled differently with text.