The goal of programmers has always been to create a single near-perfect programming language that is in no way platform specific and can be programmed simply through interaction with the user to be able to write its own applications for performing services for the user. An example of this sort of system would be one in which the user would simply tell the computer that he wishes to write a letter, and the computer would write him a word processor that is designed. This sort of system however, would require advanced artificial intelligence – it would not only have to interact with the user, but would have to be able to perform the correct operation whether the user tells the computer to “help me write a letter” or to “let me word process”. The computer would have to have some form of intelligence in order to accept any situation, and be able to do what it thinks is the right course of action. There will of course be times at which the computer simply miss-hears the user, and so it must also be able to have limits of common sense that tell it for example, whether the user really wants to have a video conference with his grand mother whilst sitting in the bath.

These rules are fundamental in artificial intelligence and are what will actually make a computer intelligent, and not just be able to reply to a specific list of instructions and variants on those instructions. Another example of a rule like this would be one that enables the computer to learn. Without learning, it would be impossible for the computer to adapt to any situation. A form of learning is built into most speech recognition technology which enables the software to adapt itself to the user. In the past it was necessary for the human to learn how to use the computer, in the future; it will be the computers responsibility to adapt.

For this future to be realised, it is necessary for a lot of work to go into the development of AI. Many developments have taken place already, with regards to modules of things that we may relate to artificial intelligence – such as speech recognition, and of the guessing of functions by selecting key words from a passage. Much work has gone into the grammar and speech checking systems built into word processors, which can actually work out the context of a paragraph and give the user ideas on how to correct bad language. Common sense can also be built into the system through a series of rules, but for the computing to actually seem intelligent, it is necessary for it to be able to learn.

So far, it is learning that has proved most difficult. It is incredibly difficult to teach a computer to learn, since it is learning that seems to separate us from machines. Where a computer and baby might both start off their lives with a series of basic rules – in the case of the baby, primal feelings that tell it how to react to simple stimulus such as pain and warmth – it is the things that the baby can learn to do that separate it from the machine quickly. To perform a different task, the machine must be loaded with a different program sequence, it cannot simply learn one. The computer might be able to watch a human perform an action and then repeat it, but if it was told to learn a bicycle for example, it would simply follow a sequence and would quickly fall off as soon as it reaches something such as a bump in the road. It is the ability to react to unknown situations that allow the baby to combine both wisdom (basic rules that govern the actions of the baby) and knowledge (of what it did last time it hit a type of road obstacle) that allow it to formulate a plan of action of what to do this time. Where the computer would calculate which other bump this obstruction is most like, and do the same again, the baby might just get off the bike.

Of course babies cannot usually ride bicycles.

This got me a 1 for attainment in computing - the highest mark available.

Written by Rawson