Growth of AI
Developments of the research community For centuries, artificial intelligence has been an intellectual challenge; only recently a glint of practical feasibility has resulted in a wide variety of applications (Tanimoto 1987). During the early and mid-1960s, ambitious projects in automatic English/Russian translation not only failed to produce the promised systems but dampened the respect and enthusiasm for Al as a field. The field has, however, recovered now. Many scientists, engineers, researchers, and programmers are studying Al techniques and building AI systems. National and international organizations dedicated only to Al are being developed, and they are growing. In the US, the American Association for Artificial Intelligence conducts a conference every three or four years, where research results are reported, tutorials are offered, and an exhibition of equipment and books is held. TESLA
Industrialization of AI The recent, much deeper understanding of the problems and solutions in the major domains of AI has helped to provide a solid base for many AI systems and applications. Certain ventures in computer-assisted language translation have recently been launched, even though the procedures of natural language translation have not yet been understood completely (Tanimoto 1987). Machine vision is now practical in the fields of robotics, biomedical microscopy, and material analysis, even though many basic questions of vision have yet to be answered.
The role of the Al practitioner During the next decade, a majority of AI engineers
are likely to be designing expert systems. They are focusing their work in the field. of medicine, finance, and anthropology to develop suitable representations for the knowledge in each field. Knowledge must be put into a form from which useful inferences can be made automatically. In addition to developing knowledge representations, suitable displays and means of access must be designed for users. Natural language and cathode ray tube (CRT) interfaces must be designed with capabilities particular to each application. After an expert system has been designed, it must be maintained, new knowledge added, and unnecessary heuristics replaced from time to time because there is usually room for improvement in fields such as medical diagnosis, mathematical theorem proving, etc. New technologies must be incorporated into expert systems as and when they become available. In the coming years, the basic mechanisms of machine learning and problem-solving will be studied continuously. The field is sufficiently rich, and the many basic issues such as optimal search, probabilistic reasoning, and inductive inference will provide open problems for many years.
Artificial Intelligence and Industry
For a long, artificial intelligence has been subject to much criticism and has been the object of controversial discussions. The pros and cons of the name itself, the contents, and the results of the field have been argued (Glorioso & Osorio 1980). It has been maintained that Al has no theory to offer and nothing can come out of its shaky grounds. Consequently, the industry felt secure in not trying to take note of the possibilities of artificial intelligence until very recently.
It is quite clear that the situation is dramatically changing now. There are, in fact, many reasons for this in-depth evolution, and we shall review them in detail. Three events advanced the idea that AI was to be investigated and assessed by the industry itself. The first event was the accomplishments of AI itself, especially in its knowledge-engineering component. The second one was the hardware evolution which made it feasible to implement ideas relying on more powerful or cheaper technology, and which opened a wide potential market to new and unsophisticated users requiring new ergonomic, behavioral attitudes, to which traditional computers could not do justice. The third event was the need for industry. services and the society to invent solutions to increase productivity and satisfy ambitious social needs.
What is it that makes problem-solving techniques of interest to AI? The three main reasons are development in knowledge engineering, major technical breakthroughs, and the fulfillment of new needs. Al has influenced the industry mainly in the following areas: factory automation, programming industry, office automation, personal computing, education, and social changes.
AI and the World
Artificial intelligence researchers have been trying to develop a computer that thinks. They have to keep in mind the nature of the world when they introduce intelligent systems (Charniak & McDermott 1985). What would the existence of such machines? say about the nature of human beings and their relation to the world around them? This is one of the main questions asked to determine the nature of human values. Al has tremendous applications in various fields. One such area is robotics. We can use robots to weld various machine parts or to transport material from one place to another. But how one gets a robot arm to move from one place to another without damaging any other thing was one of the standard questions.
Another area of AI research is natural language comprehension. There is a program that translates a natural language query into a database query language an artificial language designed has to be used by a computer database. Some researchers are trying to create machines to perform medical diagnoses. Further, an Al program that has replaced people in its domain of expertise is the XCON program. Some Al programs perform mathematical calculations, predict the presence of mineral deposits, use TV cameras to see the world and identify objects, verify the designs of electronic components, and much more.
The basic work of an Al researcher is to write a program that considers pictures from slightly different viewpoints and finds points in the different views which correspond to the same place in the scene (Charniak & McDermott 1985). If we are talking about robots in an application of welding a car body, there is no evidence for the robot to see what it is doing. However, while designing a robot, we have to ensure that it recognizes how far away something is. Stereoscopic vision is useful here because objects will appear at slightly different places for the two eyes, and the discrepancy is a good clue to how far away something is. This is considered one of the major technical issues challenging Al researchers in the world.
Current Trends in Applied AI
Several techniques are currently popular within the area of applied AI. Each technique employs a different method of intelligent behavior and is applied to problems that increase its power to reach solutions (McDonald et al. 1997). The key techniques presently in use are expert systems, fuzzy systems, neural networks, genetic algorithms, and swarm intelligence.
An expert system differs from other Al systems in that it attempts to explicitly embody expertise and knowledge within the software. This has been applied in many engineering applications (Kosko 1997). It is recognized that the analysis of data can be subjective. To allow for this situation, fuzzy systems using fuzzy mathematics have been employed. Such systems allow an element of fuzziness or vagueness to be associated with data. As a simple example, a load demand can be manipulated by a fuzzy system by converting the actual load into one of the following quantities: very high load demand, high load demand, medium load demand, less load demand, and very less load demand. A useful application of such fuzzy systems is in decision-making under uncertain or hypothetical situations.
In terms of practical applications, due to their learning capabilities, neural networks are often employed where a problem requires estimation, prediction, or classification. Neural networks do not store knowledge or expertise explicitly but implicitly characterize behavior through a learning process.
A common class of problems in AI applications is that in which the solution is found through some type of search strategy. Genetic algorithms are favored for search problems that require the identification of a globally optimal solution. They are based on the natural evolution process and operate by combining the best solutions at each stage to obtain the next generation of solutions. This process continues as an iterative method until the optimal solution is found (Mitchell 1998). The popularity of the genetic algorithm approach can be attributed to the fact that it finds optimal solutions without the search becoming trapped in local minima or maxima. Genetic algorithms can also be applied to problems with a large number of variables. Similarly, swarm intelligence utilizes the behavior of non-human living for problem-solving.
- Greater Cloud and AI collaboration
- AI solutions for IT
- AIOps become more popular
- AI will help in structuring data
- Artificial intelligence talent will remain tight
- Large scale adoption of AI in the IT industry
- AI Ethics is the focus
- Augmented Processes become increasingly popular
- Artificial Intelligence will become more explainable
- Voice and Language Driven intelligence
MODELLING, SIMULATION, AND AI
Artificial intelligence attempts to understand intelligent entities. Therefore, studying AI helps us to know more about human intelligence and ourselves. Al has produced many products in the last two decades, and computers with human-level intelligence will have a huge impact on our everyday lives and the future civilization (Geoffrey Gordon 2001).
Al is truly a universal field. Scientists from varied fields such as medicine and computers can contribute to it, and in turn be benefited from it. The term ‘artificial intelligence in its broadest sense covers the number of technologies that include but is not limited to, expert systems, fuzzy logic, neural networks, genetic algorithms, and swarm intelligence. Interestingly, most of these technologies have their origins in biological or behavioral phenomena related to humans or animals, and many of these technologies are simple analogies of human and animal systems.
Al may be defined as the branch of engineering science that is concerned with the automation of intelligent behavior. On the other hand, the definitions of Al programs are many; for instance, programs that perform tasks that, if performed by a human being, would be considered intelligent. These programs run on digital computers, just as do ordinary programs. Admittedly, they can be applied to a wide range of tasks for which conventional programs are ill-suited, such as complex optimization, robotic control, strategic game playing, automatic theorem proving, forecasting, decision making, and reproducing the expertise of human experts.
We can make two interesting observations here. First, the difference between the mind and the physical world has become so clear that the process of thinking can be discussed in isolation from any specific sensory input or worldly subject matter. Second, the connection between the mind and the physical world is so. tenuous that it requires the intervention of God to allow reliable knowledge of the physical world. This view of the duality between the mind and the physical world underlines all of Descartes’ thoughts, including his development of analytic geometry. How else could he have unified such a seemingly worldly branch of mathematics as geometry with such an abstract mathematical framework as algebra?
Although in the eighteenth, nineteenth, and early twentieth centuries, the formalization of science and mathematics created the intellectual prerequisite for the study of artificial intelligence, it was not until the twentieth century and the introduction of the digital computer that AI became a viable scientific discipline. By the end of the 1940s, electronic digital computers had demonstrated their potential to provide the memory and processing power required by intelligent programs. It was now possible to implement formal reasoning systems on a computer and empirically test their sufficiency for exhibiting intelligence. An essential component of the science of artificial intelligence is this commitment to digital computers as the vehicle of choice for creating and testing theories of intelligence.
Digital computers are not merely a vehicle for testing theories of intelligence Their architecture also suggests a specific paradigm for such theories; intelligence is a form of information processing. The notion of search as a methodology, for example, owes more to the sequential nature of computer operation than it does to any biological model of intelligence. Most Al programs represent knowledge in some formal language that is then manipulated by algorithms. problem-solving
The evolutionary paths of artificial intelligence and simulation have only recently begun to converge. This convergence began as an outgrowth of research in cognitive psychology and mathematical logic (Haykin 2003). Until recently, the focus has been on explaining the working of the mind, and on the construction of general pose problem-solving algorithms. In contrast, simulation has developed from the purpose pro the need to study and understand complex time-varying behaviors exhibited by real physical systems. It is not surprising, therefore, that the world views of the two fields differ considerably.
The world view of artificial intelligence favors abstraction, generality, and elegance. The world view of simulation favors practical utility, precision, and reliability. The cultural forces separating the two communities have begun to subside, however, as a consequence of falling computing costs, increasing demands by sophisticated users for improved software performance, increasing access of inexperienced end-users to sophisticated computers, and increasing connectivity between databases and machines. After reviewing the historical development of both these fields, the differences between them seem to be more apparent than real.
INTELLIGENT SYSTEMS
A widely recognized goal of artificial intelligence is the creation of artificial scenarios that can emulate humans in their ability to reason symbolically, as discussed in typical Al domains such as planning, and natural language understanding. diagnosis, and tutoring. Intelligent systems can be constructed from explicit, declarative knowledge bases, which in turn are operated by general, formal reasoning mechanisms (Bielawski & Lewand 1991). This fundamental hypothesis of Al means that knowledge representation and reasoning-the study of formal ways of extracting information from symbolically represented knowledge-is are very important to the field of study (Brachman 1988).
Brachman’s description of Al relates to our general definition of intelligent. systems in two ways. It concentrates on knowledge representation and reasoning as the theme of any system that reflects intelligence and suggests that it is possible for a system to behave in this manner only if it contains formal mechanisms for representing knowledge and employing inference techniques that model conventional, computationally based systems.
Central to this type of intelligent activity is the ability to access and synthesize discrete pieces of information in creating a new understanding of any problem and its possible resolution. It has also been found that the uncertainty or fuzziness in certain systems, such as those involving temperature measurements, is too high, and this has led to fuzzy intelligent systems (Rajasekaran & Pai 2003).
Similarly, neural network intelligent systems have been developed based on the functioning of the human brain. A neuron is the fundamental unit of the brain and its behavior is similar to that of static neurons (Zurada 1994). Attempts to model the biological neuron have led to the development of the field of artificial neural networks (ANNS). The neural network intelligent system has been developed in order to facilitate predicting features in advance based on the previous details available. For example, if we are provided with the pattern of how, in the past, various data have affected the temperature during a day, we can predict the temperature for any particular future day. Similarly, we can use genetic algorithms and other evolutionary techniques, which have been developed mainly by emulating the nature and behavior of biological chromosomes.
Along similar lines, Marco Dorigo and his colleagues first proposed ant algorithms in 1991 as a multi-agent approach to solving difficult combinatorial optimization problems. There is presently a lot of ongoing activity in the scientific community to extend ant-based algorithms to many different discrete optimization problems.
