Artificial Intelligence (AI)
Introduction
There are various definitions of AI. John McCarthy came up with the term Aritificial Intelligence in 1955 and defined it as, "the science and engineering of making intelligent machines". The BCS defines it as, "the study of relationships between all aspects of intelligence of all forms and modelling them by computer systems".
Artificial Intelligence is important because it aims to make machines and systems generally better at what they do. Alan Turing (of Bletchley Park and Enigma fame) was very interested in AI and invented the Turing test, to evaluate whether a machine could convincingly mimic the intelligent behaviour of people. You can read about the Turing test here. Applications of AI today include expert systems, vision and image processing and natural language processing and machine learning.
Expert systems
We have seen in the previous section one area of the application of AI - expert systems. You can read more about expert systems here.
Pattern recognition
This area of AI seeks to digitise an image or sound and then analyse it and compare it to a database of patterns to try and find a match. There are lots of different applications of this technology.
-
- The police try to identify criminals from a digital image of a fingerprint or facial image.
- Individuals in pictures can be automatically tagged with their name on social media web sites.
- Face and retina scans can be used at passport controls to identify individuals.
- Casinos try to spot people they have banned from entering their casinos by using face recognition software.
- You can identify a piece of music on TV or over the radio by letting an app on your phone listen to a part of it.
- Terrorists can be identified from tapped phone calls when their voice is matched to a known example in a database.
- Quality control systems in factories can tell well-made products from sub-standard ones by taking pictures and comparing them to known standard examples.
- Objects can be identified from an image of them e.g. a stolen work of art.
- Machines can be positioned correctly using a camera e.g. so a weld can be done correctly.
Natural language processing and machine learning
This area of AI looks at the interaction between computers and people using their voice and working towards getting computers to 'understand' what a person wants. The programs that aim to do this often use something called 'machine learning'. This means that instead of trying to work out what the maths rules (the algorithms) are that describe something, a large block of data is used to try and build an algorithm that describes something and then that algorithm is used to make predictions and decisions. Based on the results and outputs of the predictions, the algorithm is revisited, fed the new data, improved and then are used to make more predictions and decisions. This results in more output data, which can be fed back into the 'improving the algorithm' cycle in an iterative cycle.
There are many applications of natural language processing including speech recognition and text-to-speech applications, translating between one language and another, providing answers and accurate responses to human questions, providing summaries of text documents, extracting and interpolating information from documents and so on.