This series of articles is focused generally around the topic of Artificial Intelligence (AI). We will start by looking at what AI is, and cover the various ways in which it can be implemented and applied using computers and modern technology in further articles.
Part 1 – An Introduction
Artificial intelligence is a very broad field, and far from being isolated to computing it encompasses many other disciplines such as philosophy, neuroscience and psychology. It is important to note though, that rather than just seeking to understand intelligence, AI practitioners seek also to build or create it. The uses and applications of AI are many and varied, and although many think of humanoid robots when we discuss AI, you may be surprised to know that we already encounter applied AI in our day-to-day lives.
AI is full of big questions – how does an entity (either biological or mechanical) think? How does it understand or solve a problem? Can a machine truly be intelligent? What is intelligence? The answer to these questions may not be easy, but there is an answer staring us in the mirror so we know the quest to find out is achievable.
Through this series of articles I will be exploring the many different approaches, sub-fields, applications and questions that we encounter when exploring this vast and exciting field of research.
Part 2 – What is AI?
Firstly I would like to say that the term Artificial Intelligence (AI) means different things to different people. In fact even the words we use to describe the topic are ambiguous. The term artificial can have subtly different meaning; consider what we mean when we refer to ‘artificial light’. This is real light, that has been created by a man-made source. It functions exactly as we would expect light to function, and from a physicists point of view it simply ‘is’ light. When we refer to ‘artificial grass’ however, we use the word artificial to mean something subtly different. Artificial grass is not grass. It is not a plant, it is not made of the same material as the plant, and it does not share all the properties of real grass. It does however perform the main functions of grass adequately, and may often fool people into believing that grass is present.
The term intelligence is also open to interpretation, and so we end up with some very different definitions of what AI actually is. The definitions we come up with however tend to fall into a one of two categories – they are either focused on the process used to achieve the goal, or on behavior. For example, Luger & Stubblefield define AI as ‘The branch of computer science that is concerned with the automation of intelligent behavior‘ Whereas Winston defines it as ‘The study of the computations that make it possible to perceive reason and act‘.
We also must consider how we measure success, and again there are a couple of common standards. We tend to either assess our system when compared to human performance, or against an ideal concept of intelligence often referred to in the field as ‘rationality’. A system is rational if it makes the correct decisions.
Broadly speaking we end up with four acceptable goals in producing AI – systems that think like humans, systems that act like humans, systems that think rationally, and systems that act rationally. In the next part of the series we will start to examine each in more detail.