To evaluate whether a computer could one day think, it is necessary to define what is meant by the process of thinking. Thought amongst living beings is not limited to just humans, it is the act of arriving at reasonable conclusions based on a number of preconditions and environmental factors. Thought can also be defined as the manipulation of information. Often thought is based on past history and it almost always takes into account some reasonable store of knowledge. This is why older humans are capable of a higher level of thought than younger humans; they have a wider store of knowledge to call upon, as well as a more considerable past history. The act of thinking requires some evaluation of all of these inputs and a fairly realistic output, where realism is measured in how accurately the output would relate to that of an average human thinking. A computer in its current form cannot think as a human would.

Currently, the methods (that we have covered, but supposedly near enough to the state of the art methods) for producing a thinking computer are somewhat flawed. Artificial intelligence techniques aim to produce realistic output but are usually limited to certain scenarios. An overview of the methods currently available:
a)
Search. This method of action involves evaluating either all or some of the possible paths to a goal. When a suitable path is found, it (the agent in question) will proceed down that path. This method works well for games where there are a reasonable number of possible moves at each stage, but even then can require a vast set of data to provide it with the actions to take at the beginning and the end of the game. When concerned with real life data this set of data would be extremely large. The number of possible start states to achieve a goal is many and varied. Search would not be possible to allow varied thought, not with the current amount of computational power. It may be possible for specific problems.
b)
Learning algorithm. This method (also known as machine learning) involves using a neural network to infer from a set of training data what ideal outputs should be. It works well using a few classes (these are the different possible outputs). Given a suitable algorithm, it is possible to classify general input which may have not been seen before to an output which reasonable models the decisions made by experts. This is similar on a largely diminished scale to what humans do on a daily basis. But representing the real world mathematically (which is what this tries to do) is difficult – there are too many variables to take into account. This is usually based purely on a specific example or for a closed system.

The situation changes if you take into account human feelings. Feelings are the subjective experience of emotions – emotions which are sometimes irrational. Emotions are triggered by changes in the environment, and are usually the function of some instinct instilled in the average human by evolution. They manipulate our thought in varying ways depending on how beneficial the situation would be for the human. If dangerous is present, our emotions will cause us to flee. Certain actions, such as eating will cause us to become happy. These are not strictly necessary for a computer, because it cannot preserve its own being (assuming it is in fact just a computer and not part of a larger construct more capable of its own preservation). Thus it is unnecessary to give a computer emotion, and the process of thought should not be concerned with feelings. If the computer was to be part of a separate being, then it may become necessary to provide it with the ability to feel. The emotions necessary to drive this being may come from whatever feedback mechanism is present to allow interaction with its requirement. These could provide part of a framework for providing basic instinct to ensure its own survival (although, ideally within Asimov's Three Laws).

Thought is a non-specific name for describing a process of manipulating and evaluating information to achieve at a conclusion. It does not specify that the result has to be sensible nor does it specify that this action would mimic what a human would do. Indeed, most animals with some degree of intelligence will think, but this will not necessary be in any way similar to how humans tend to think. So it true, a computer can think but just not at the same level as the average human being. If the definition of thought is limited to just a specific problem at hand, similar to game playing and machine learning, then a computer has a rudimentary, incomplete, form of thought.

Thought can also encompass other facets of the consciousness involving the imagination and formation of concepts (i.e. a generalized understanding about a specific item or object or possibly even an abstract concept based on a collection of other concepts). Whilst computers may be able to reason using these basic strategies, there exists no suitable way for them to effectuate even a simple imagination, comparable to a human child (although it is often said that children have a substantially more vivid imagination to adults). Likewise, the formation of concepts is restricted by the current solution's scope. It is computationally infeasible, at the status quo, to have a situation where a simulation of thought would encompass such a general scope such as to even provide the sheer amount of information that would be necessary to allow genuine concepts to arise naturally.

A simple test of intelligence (and thus the ability to formulate thoughts) is the Turing Test, which envisages a situation in which you have a human judge communicating with each two entities. One is human, the other robot. Both try to appear as if they are human. If the judge cannot reasonably distinguish between them, then that computer has passed the test – i.e. it can be regarded as intelligent. This could be used to demonstrate the process of thought, albeit in one restricted aspect (it is possible to pass this test and still not be able to imagine or plan).

All in all, whether a computer can think or not depends very much on how thought is defined. It is a subjective and abstract concept – something which is an attempt to verbally qualify our own consciousness. It is difficult to define this precisely because it may vary just between humans. There is no specific scientific description which can be rigorously stated. This issue aside, assuming thought is multi-layered, consisting of several independent processes, a computer cannot think – at the moment. It may be possible to construct a thinking computer if sufficient resources were available at some point in the distant future, but for the moment they are not. A computer can be used to solve problems and play games using what can be regarded as incomplete thought – a process that does not fully represent thought in its true form.