Artificial intelligence: goals and possibilities

It is difficult to clearly and precisely define Artificial Intelligence as it is difficult to define intelligence alone: which is something that even today we cant measure. Intelligence tests or IQ tests that are taken today are far from being accurate and fair. This said, in an advanced way we can define intelligence as something that possesses a conscience of their own and who can process certain logic and apply it to advance their understanding of this logic. Logic is therefore defined as the ability to think and to arrive with new ideas, meaning the majority of people can use these ideas to change society in different ways.

This example can be easily thought of differently but the principle remains. In any event, in no one way will we be able to define intelligence as long as we only know what’s going on in our heads; this means to define artificial intelligence we are going to have to take a look around us.

In our everyday lives, the fact that you can talk with other people requires a certain amount of intelligence: Artificial intelligence can therefore be defined as sitting in front of an artificial entity (a computer), to talk to it and to receive answers that correspond to the questions that were asked (even if the computer gives the answers based on a set of conditions that it “thinks” are correct). There is a test that exists to do precisely this: the Turing test.

Created by the person of the same name in the second half of the 20^th century, it consists of a person sitting in front of a terminal and the computer asking questions to determine if it’s a man or woman. Once having concluded the answer it then tests its man simulation programs. The test is simple: Put a man in front of the computer and ask him to communicate. He doesn’t know to whom he is communicating; maybe it’s a machine, or another man (or woman). The goal is to make a distinction between the two entities. If the “guinea pig” can’t work out the difference, then it’s over: the computer successfully passes the Turing test.

You can although reassure yourself that this phase has not yet arrived. One of the most powerful (and rare) chatterboxes (a discussion machine), ALICE, wins prizes for efficiency, but still cant imitate a human. ALICE is built on a giant database where all imaginable questions have been inputted along with their respective answers. But the system does have a dead end: not only does it take a long time to search the answers, but we also see problems when she doesn’t know what to answer. The problem lies in that she is incapable of properly learning, which is the mechanism of real “artificial intelligence”.

And there is really little need at the moment to be able to program gifted Artificial Intelligence programs, most of the programs that are available today are limited and are incapable of advancing themselves. They only have the ability to learn what their creator decides to teach them. At this moment, we are only capable of providing an inalterable database, this making it limited. Imagine if your company’s phone book was like this; it would be useless as the information would be 10 years out of date! This would make it old, out of date and totally useless in an industry that changes as quickly as that of IT.

So, we have a semi-definition of what artificial intelligence is; A system that can learn and simulate life. Now we need to know how to implement this, both from the point of view of a designer and of a user and how this can benefit both camps.