What is Artificial Intelligence?
Artificial intelligence (AI), also known as synthetic intelligence, is a branch of science and engineering which studies and designs intelligent machines, and in particular, intelligent computer programmes.
It is also often used to describe the property or characteristic of machines or computer systems that demonstrate intelligence. Intelligence in this case being traits such as learning, reasoning, planning, communication, knowledge and perception, as well as the physical ability to move and manipulate objects.
AI research first began after World War II, when several scientists began to independently work on different ‘intelligent’ machines. Alan Turing, an English mathematician, was probably the first and was probably also the first to decide that AI research benefited more from programming computers than from building machines. Since then, most AI researchers base their work on programming computers.
AI and Human IntelligenceThe ultimate goal of AI researchers is to create computer programmes which can solve problems as well as any human. Thus, a large part of AI research is directed towards simulating human intelligence. However, AI researchers are not bound by biological constraints and therefore they can also tackle problems that require far more computing than humans are capable of.
There is some doubt over whether human-level intelligence can ever be achieved, some researchers feel that it can be if a large enough number of programmes are written and a large enough knowledge base is accumulated. Most AI researchers feel that computer systems will always lack the new fundamental ideas required and therefore, it is difficult to predict if human intelligence can ever be achieved.
AI and ComputersWhile many researchers have attempted to invent non-computer machines which exhibit artificial intelligence, so far they have always had to simulate the machine on a computer first, due to the large costs involved in building such machines.
The computer simulations have always performed so fast that that the researchers have doubted the worth of actually building these machines, as they will probably never be superior to computers in speed and performance. Certainly, with the ongoing development of computers, they will only get faster and faster, with more and more sophisticated programmes meaning that they continue to be best agents for artificial intelligence.
Applications of AIAI borrows from several fields, including neuroscience, computer science, cognitive science, psychology, linguistics, operations research, control theory, philosophy, logic, optimisation, economics and probability. It involves operations such as logistics, robotics, speech recognition, control systems, facial recognition, scheduling and data mining, among others.
AI is applied in many areas:
- Game Playing – machines which can play well against people, mainly through brute force computation. For example, an AI master level chess machine would be able to beat a world champion by being able to consider and assess 200 million positions per second.
- Speech Recognition – although an application that is of great practical convenience, unfortunately, computer speech recognition has only limited use; for example using voice dictation to replace keyboard typing. However, many people who have tried this have returned to using a keyboard and mouse manually in preference.
- Understanding Natural Language – again, something that is only possible in limited areas, this involves the computer being provided with an understanding of the domain a particular text is about. It is not enough to just get a sequence of words into a computer or even to parse sentences.
- Computer Vision – trying to mimic the human ability to interpret the 3-dimensional information from the world around us. At present, there is some limited ability for computer vision to represent 3-dimensional information directly.
- Expert systems – this involves the knowledge from experts in a certain domain being embodied in a computer programme, for carrying out a specific task. For example, one of the first expert systems created was MYCIN which diagnosed bacterial infections of the blood and suggested treatments. While it performed its primary task better than medical students or even practicing doctors, it could only do this within set limitations – namely that its interactions were dependent on a single patient being considered and its framework only considered bacteria, symptoms and treatments but could not include doctors, patients, hospitals, recovery, death and events occurring in time.
- Heuristic Classification – an extension of expert systems, this involves inputing information in one of a fixed set of categories, from several sources of information. For example, modern credit card transactions use this system by considering information about the owner of the credit card, his record of payment, the item he is buying and the establishment is buying from – and then advising whether or not to accept the purchase.