Artificial Intelligence & Internet

by Bizhan Nasseh

1. Introduction

The imagination of creation of intelligence in the machine has been with the human for a long time, but the formation of this discipline is from the work of various groups in the United State and Europe in the years before and after World War II. The earlier research included cybernetics, behavior, and theories of computation. In the 1950s, in the scientific world the potential for computer reasoning began to be realized. In 1956, in order to direct and organize research efforts in this field, many active researchers came together for the "Summer Research Project" at Dartmouth College. This meeting had a major impact on the Artificial Intelligence(AI) research for the next few decades. The main goal of AI was and still is to replicate the functionality of the human mind. Over the years, AI covered the following range of topics:

     In 1960, NASA asked researchers at Stanford University to build a program that can perform chemical analysis of the soil on Mars. The knowledge of an expert chemist had to be encoded into a computer program. This was the starting point for the creation of an expert machine. The created program was called DENDRAL which continues to prosper and in some cases outperforms human experts. This system was followed with some other major rule-based systems such as MYCIN and XCON.

     The explosion of the Internet and its applications in communication, business, and education sooner or later will have a great influence on the lives of every citizen. The addition of the National Information Infrastructure (NII) is changing the ways in which of business, communication, and teaching and learning are planned and conducted. Artificial Intelligence can play a vital role in the utilization of resources on the Internet.

     There is not adequate attention and study about the contribution of AI in the Internet, but AI has potential not only to address many scientific problems, but also find many practical way to help general users in the utilization of the Internet. AI can make computers and the Internet easier to use by providing intelligent interface which can help user to find needed information and adjust the process to individual skills and pattern of usage. AI representations can help the development of a flexible infrastructure for the Internet. The Internet contains information and data from a wide variety of sources, including natural language, video, audio, mathematical equations, animation, images, and databases. AI technology can enhance the Internet usage-- speech and image processing can allow to extract and identify multimedia content, knowledge representation enable data translation services to convert information from one format to another, or expert technology can provide tutorial and problem solving. These AI tools will help in constructing advanced user interfaces and development of the complex systems needed for the easier utilization of the Internet applications. In the following sections, I will discuss AI contributions in detail with introducing intelligent partner for better utilization of the Internet applications.

2. Intelligent Partner

Current computer systems and Internet applications are difficult to operate and access even by experienced users. Current user interface forces users to adapt to the machine instead of adopting machines to the individual users' needs. The ease of Internet utilization is a major issue and need which must be addressed as soon as possible by higher educational institutions and business sectors. How we can expect all the general users to memorize commands and options for locating information and accessing resources. The same time we are witnessing daily changes in the procedures of operations and amount of resources on the Internet. Explosion of information and resources on the Internet will make the effective utilization of applications such as World Wide Web difficult for regular users.

     Intelligent partner is a great contribution of AI to the utilization of the Internet applications. It is capable of a goal-oriented behavior in the needed environment. The structure of the intelligent partner is a dynamic knowledge- base, knowledge presentation, computational structure, and database which generates and directs the partner's behavior in the environment. The main capability of the intelligent partner is interfacing with users, customizing activities, providing language processing, learning activities, and image understanding and synthesis. The last three capabilities are functions of an ultimate form of the intelligent partner. In the simple form, a design of an intelligent partner can combine a few capabilities such as intelligent interface, intelligent tutorial, and intelligent problem solving. One of the main problem is integration and interaction of all the developed components with each other in real-time performance. There are some major research and development activities in some of the higher educational institutions in the design of the some kind of intelligent system which can be incorporated with the Internet activities. These types of activities need investment, commitment, and support of the organization. In the following subsections I will present my idea about components of intelligent partner in the form intelligent interface and information finder, intelligent tutorial, and intelligent problem solver.

2.1. Intelligent Interface

Today, most of the interfaces of the Internet applications require the memorization of many commands and sequences of menus by users. The Internet interfaces need to be intelligent in order to adjust automatically to the user's skills and patterns of usage. An intelligent interface is very important element of encouragement for the general public in the utilization of Internet applications. User should have the choice to give requests in the form of speech, text, or query language and intelligent interface should be capable of designing a unique process for achieving the request. The intelligent interface should be able to customize access methods based on the users needs and instructions. Current interfaces can process and generate data far more easily than can present and interpret it. When a query returns huge amount of data, user need intelligent interface to present data in the best suited form such as chart, graphic, natural-language, text, video, and audio. The main technical concern is to adapt computer to summarize and synthesize all the accessed information in the best form of presentation which suits the user's needs. Some of the technology such as generation of grammatical text from certain knowledge-representation formats and synthesizing speech from unrestricted text is already here. The AI technology for automatic design of effective graphics is still in its research and development stages. Some other AI technology such as intelligent decision making for combination of appropriate modalities automatically, will be achieved in the near future. The heavy investment in communication and Internet applications, and necessity of utilization of Internet resources by general public will direct and provide opportunities for many new visions and ideas in the development of intelligent interfaces for users.

     Finally, someday users should be able to make requests based on what they want to accomplish and leave the problem of determining how to achieve that goal to the intelligent interface. Higher educational institutions always have been pioneer in the AI research and development. There is a great opportunity for adding AI power in the form of intelligent agent or partner to the Internet utilization.

     The following example is a simple illustration of the mass and variety of information from diverse sources which emphasizes the necessity of an intelligent partner. AI learner wishes to find information about AI on the Web. The intelligent interface system(IIS) asks which area of AI and the learner answers with the research area of the AI. The IIS asks current research, past research, future research, or all of them. Learner selects the current research. IIS will arrange search query language and communicate with the browser. The results are in the form of texts (from a server in USA), graphs and images (from a server in Japan), audio (from a server in Australia), and video (from a server in Germany). How much time and efforts user must spend in order to make knowledge from these information? Does AI technology have the power to use knowledge representation, speech and language processing, and image understanding all together in order to provide a useful and related information in the form of knowledge for user? This is a major processing ability which can reduce the volume of retrieved data and information by means of combining, abstracting, and summarizing. This is a wish list and vision, but I am sure many proactive, aggressive, and quality research institutions already are planning and investing toward this type of research in the AI technology for the Internet.

2.2. Intelligent Tutorial System

The advancement in the communication and computer technology, and necessity of the utilization of the Internet applications provide a strong reason and foundation for the design of a wide range of intelligent tutorial systems. In the next few years we are going to witness the invasion of the Internet applications such as Web in the daily life (work, entertainment, education, and communication) of the general public. This invasion provides a great need for research outcomes in the area of users' characteristics, level of competency, and types of utilization. Intelligent tutorial system will be developed by using system technology. The differences between CAI and Intelligent tutorial system is CAI system have no understanding or knowledge of subject domain or teaching function. The intelligent system posseses a knowledge of the subject domain and teaching. The intelligent system should be able to infer why a student is having difficulties in the operation of Internet application and reformat the instruction materials in a such a manner as to overcome the student's problems. The knowledge-base of this type of a system has knowledge about various levels of users competency and the system should be able to infer which level a particular learner belongs to and presents the related training accordingly.

     In addition to the separate intelligent tutorial system which I explained above, there should be an intelligent tutorial-while doing system. These intelligent system will be activated in the case of mistake or problem by user. This system first acts as an intelligent problem solver, providing intelligent advice for continuation of operation, but at the same time it provides options for learning opportunities about the mistake and problem. If a user selects to learn more about mistake, the intelligent tutorial system provides some examples and needed knowledge. At the end of the operation the system will remind the user about needs to learn more about the part that user had problem. The ultimate form of capability of this type of intelligent tutorial system is to evaluate user's competency after each utilization and update its knowledge base for more customized and error free utilization. The system also updates database with the information about the learning needs of user in the better utilization of the Internet applications. The intelligent system will advise user for improvement of competency with design of proper learning activities.

2.3. Intelligent Problem Solver

As computers began to spread through engineering and accounting applications, people interested in lifelike machine began to research and to use them as symbol processing systems. They realized that they can program computers to do things that required thinking by humans (solve problems and prove theorems). Computers as expert systems became major contributors in the problem solving in variety of disciplines: as physicians to analyze certain kinds of disease (MYCIN system by Edward Shortliffe), as engineers (EL system analyzing the electrical diagrams, or programs which help design the integrated-circuits), or computer program can learn from example and precedents (INDUCE system by Ryzard Michalski). With all these capabilities computers can be clearly looked at as intelligent problem solvers.

     In the Internet world the diversity and volume of accessible on-line data is increasing dramatically. As a result, existing tools for searching, browsing, and problem solving are becoming less effective. The increasing use of non-text data such as images, audio, video, and graph has amplified the need of an intelligent problem solving system. In addition to problem solving, this intelligent system should have capability of problem definition. In the case of mistake or uncertainty, this system should make decision for continuation of query in the complex and diverse world of the Internet. The intelligent problem solver for the Internet application is different from other problem solving systems, but today's AI technology and AI application developers have the potential for development of this complex system. The intelligent problem solving constantly should update knowledge-base and database for helping user in the future activities.

3. Conclusion

We are witnessing an explosion in the amount of information that is available from various resources on the Internet. There is a major challenge to AI researchers for empirical methods that can be used in the development of computer intelligent partner (interface, tutorial, and problem solver) for the better utilization of the Internet resources by general public. The following areas are main contributors to the design and construction of an intelligent partner system for the Internet. Some of these areas needs additional research and study in order to be customized for Internet applications. Knowledge Representation:

     A representation is a set of syntactic and semantic conventions that make it possible to describe things (Patrick Winston, 1989). The syntax of a representation specifies the symbols that may be used and the ways those symbols may be arranged. The semantics of a representation specifies how meaning is embodied in the symbols and the symbol arrangements allowed by the syntax. All representation must provide some way to denote objects and to describe the relations that hold among them. Most information currently stored on the Internet uses one of two knowledge representation methods, database or natural language text. These two methods of representation of knowledge are not adequate, new knowledge representation methods such as narrow casting and semantic translation, combine diverse information, and accurate location of relevant information. The last one is the most needed one in the representation of knowledge from distributed sources. The integration of description languages with object- oriented and relational databases could provide value-added services. The development of standard language for encoding knowledge of scientific fields and World Wide Web hypertext libraries could facilitate support of semantically rich queries.

Machine Learning

Machine learning is of growing importance because of the rapidly increasing quantities of diverse data on the Internet and necessity for software that can automatically adapt to new or changing users' needs and runtime environments. The main technical concern is developing methods to automatically form general hypotheses from specific training examples. The development of machine learning has two interrelated problems which are software that improves automatically through experience and extraction of rules from a large volume of specific data. Consider the problem each user faces in locating information in the flood of data that will be available on the Internet. Machine learning techniques can lead to learn the interests of each user by observing what they search and read, then use this knowledge to automatically search numerous sources to recommend the most interesting or related ones. Methods such as decision tree learning, and neural network learning have been applied to data-mining problems such as assigning credit ratings based on bank records, and predicting medical treatment outcomes based on medical symptoms. New approaches have recently been developed, such as inductive logic programming which enable learning more expressive hypotheses than earlier learning methods. Machine learning methods for learning multimedia data will be increasingly important in the utilization of the Internet applications which will be needed for combining information from multiple databases. An additional Internet related topic for research in machine learning is social learning methods. It means small amount of information from multiple users are combined to provide individually customized advice of utilization for each of the users. For example after observing a small number of articles that a few users read, a social learning system can suggest additional articles by correlating these users' interests with each other, then recommending articles read by other similar users.

Language Processing

The goal of language processing is to create environment which communicate between machine and human in natural language. To make this communication closer to the human with human communication is a major challenge for researchers, scientists, and developers. Some of the major ambiguities are: language is a small part of presenting a request (other parts are intonations, gestures, and choice of words), people use different language structures, and sometimes a few short words have extensive meaning which depends on the audience to understand. Natural language processing has accomplished many major tasks such as techniques for parsing, semantic interpretation, generating report which is customized to context and task, word pronunciation, word sequencing, speech generation system, machine translation system, and content-based retrieval system. With all the investment and advancement in the language processing, still effective utilization of this AI technology in the Internet applications' for general public brings a new challenge for researchers and developers for year 2000 and beyond.

Image Understanding and Synthesis

Image understanding is to generate information from document page, picture, photograph, animation, or video. Image synthesis is to generate an image from available information in the form of charts, maps, views, and animation. Image understanding and synthesis are very important to the development of the intelligent partner, but accomplishment and integration of this AI technique in the intelligent partner is a major challenge for experts and developers. Image understanding and synthesis would revolutionize knowledge discovery and acquisition, even limited success in implementation of this technology in the intelligent partner will create extremely useful services for general users. The computer-vision technology has some achievement such as facial recognition, object recognition and reconstruction, hand movement and guesture recognition, and document analysis. The low-level image understanding and synthesis is relatively available, but there are many research challenges in the high-level processing, presentation, and reasoning about visual and geometric analysis.

     Finally, with National Information Infrastructure very soon all the schools, homes, and businesses will be able to access and operate variety of information from different resources. The general public does not have the opportunity for specialized training in order to be able to operate all the necessary activities on the Internet. The field of the AI can contributes effectively to the utilization of the Internet applications by bringing machine and human closer to each other. The intelligent partner has potential to create human-computer interface, tutorial, and help system which has ability to customize activities and to provide goal oriented environment for each individual users. AI can help general public not to be dependent on the human experts in order to learn and operate on the Internet. Many higher educational institutions already recognized the current and future role of the Internet in the human's life and work. Their today's research and effort is the key for tomorrow's success in the bringing human and machine closer to each other beyond of our imagination. Yes, AI can have a key role in the Internet utilization.

RETURN TO THE LIST

Copyright © Bizhan Nasseh 1996

[Home] [Personal] [Education] [Experience] [Research] [Hobbies] [Visions]