Introduction to Artificial Intelligence
Artificial Intelligence (AI) is a branch of computer science concerned with the development of computers to perform tasks that require "intelligence". Tasks requiring "intelligence" are generally considered to include problem solving, learning through experience and automatically adapting to new scenarios.
AI is a broad topic with research divided into subfields with different approaches and aims. Strong AI is concerned with getting computers to think like humans and have human traits such as consciousness. Strong AI is an important topic for computer science, philosophy and science fiction. Weak AI (also known as "applied" or "narrow" AI) is concerned with getting computers to act rationally to match or exceed human performances at specific tasks.
The use of AI techniques has been applied to many real-world problems. Applications of AI include:
- Natural language processing.
- Hand writing recognition.
- Robotics (e.g. for use in industry).
- Data mining - discovering patterns in large data sets (e.g. for fraud detection).
- Computer games - making the behaviour of computer controlled characters more realistic.
Widely publicised achievements in AI include:
- The Google driverless car.
- Deep Blue - a chess-playing computer capable of beating the best human players.
- Watson - a system capable of understanding and answering quiz questions asked in natural language. Watson successfully competed against other human contestants on the American quiz show Jeopardy.
- Siri - an "intelligent personal assistant" available for the Apple iPhone. Siri uses a natural language interface for answering questions.
Research into AI has lead to the development of a number of general methods for solving difficult problems.
- Function optimisation - a number of techniques have been developed that, starting with one or more initial attempts at a solution, attempt to move closer to an ideal solution using a process of iterative improvement. Techniques include naive random searching, hill climbing, simulated annealing and genetic algorithms. Genetic algorithms are inspired by the natural process of evolution and are part of the wider AI subfield of evolutionary computation.
- Rule-based expert systems - general reasoning algorithms have been developed that work with a collection of facts and rules for a specific domain (e.g. medical diagnosis) to mimic the actions of a human expert. There are different approaches to modelling the rules (e.g. first-order logic and fuzzy logic).
- Neural networks - development of artificial neural networks (ANN) has been influenced by biological neural networks. Artificial neural networks are used to find patterns in data or model complex relationships.
- Probalistic methods - a number of techniques have been developed that allow software to make intelligent decisions even when the knowledge of the problem is incomplete or uncertain. Techniques include Bayesian networks and Markov models.
- Statistical classification - a number of techniques have been developed that use pattern matching to classify an item as belonging to a particular group. Techniques include decision tree learning and the k-nearest neighbours algorithm.