In Psychology, the mainstream thinking of researcher regards human intelligence as not a single ability or process but rather an array of separate processes. The research in AI has focused mainly on the components such as learning, reasoning, ability of perception, solving problems as well understanding language.
In AI research, learning is distinguished in a number of different ways, the simplest one being trial-and-error. For example, a simple program written to solve chess problem (mate-in-one) could try out different moves until one succeeds. The program than remembers that move and the next time it is presented with the same problem, it is able to solve it immediately. This problem solving technique, which involves memorizing of solutions to problems or words of vocabulary, is known as Rote Learning.
Though Rote Learning is easy to implement, one of the more challenging problem of implementation lies in another problem solving technique called generalization. A successful implementation of generalization usually makes the learner more adaptive toward situations not encountered before, thus, it is able to perform better in those situations. For example, a program written for rote learning would never be able to produce the past tense of the word "jump" until it is presented with the word "jumped" before, while the program which implements generalization is able to use generalizations from presented examples to learn the "add -end rule" and would successfully be able to produce the past tense of "jump". Sophisticated techniques such as generalization enable the programs to generalize complex data and thus solve complex problem based on their learning.
The process of reasoning involves drawing inferences appropriate to the situation at hand. These inferences can be classified as inductive and deductive. For example, deductive reasoning is "Fred is either at the shop or at home; he is not at home; so he must be at the shop. Inductive reasoning can be illustrated by "Since previous computer crashes were caused by virus attack, so probably this computer crash was also caused by a virus attack". this difference between the two is that in deductive reasoning, the truth of the premises ensure the truthfulness of the conclusion, while in the inductive reasoning, despite the fact that premises tends to support the conclusion, further investigation may reveal that it could be false as well.
AI Researchers have made considerable success in implementing inferences in computers, especially in drawing deductive inferences. However, being able to draw an inference does not make a program capable of reasoning as it involves drawing inferences that are relevant to the situation or task in hand. This is one of the hardest problems that are confronting the AI today.