Connectionist Models and Symbolic Models - Essay Example

Both these approaches have their supporters who have used them to show the various benefits which can be derived from taking different viewpoints to AI (Minsky, 1991). At the same time, there are those who have said that the approaches are complementary and do not compete with each other. To justify these statements, it is necessary to take a look at each approach to see how it works.

The symbolic approach is based on the creation of AI models that work with symbol manipulation. Computation is based on the fact that a symbol or a group of symbols can be arranged in a manner to represent information. Once the order or position of symbols is changed, the information can also be updated to reflect the changing environment. It must be noted that in this approach, a symbol can represent any type or nature of information, and the symbols can be modified if required (Bader & Hitzler, 2005). Thus a symbol here represents an idea that has certain basic premises but which can be modified by the machine based on experience or intelligent action.

This process of symbol manipulation can be very successful if the AI process is being applied to well-structured problems. That means if the symbols are predefined for the AI system, it can modify the associated parameters based on its own experience and manipulation of the symbols (Bader & Hitzler, 2005). This method seeks to replicate the higher-order thinking processes of the human mind, which has the same ability to modify and learn from experiences as they are collected in a person’s memory. However, this approach has certain failings as well since it is difficult to grasp commonsense reality with this approach.

On the other hand, the connectionist models for AI are based on the idea of using a large number of relatively simple units which can handle extremely difficult and intensive computational problems (Minsky, 1991). This approach is completely different from the symbolic approach since the technological differences between the methods mean that it is rather difficult to reconcile the two. The connectionist model can be said to come from the biological aspects of intelligence, as it is seen in lower-order human reasoning.

The beauty of this approach lies in the fact that since it is based on creating connections between simple units of information, the learning and education process for an artificially intelligent entity is simplified with extensive repetition. Neural networks are somewhat better at adaptation and can mimic information storage, retrieval and connection, much like biological entities. A connectionist approach can even handle incomplete, fuzzy, generalized or contradictory information better than the symbolic approach is able to.

Even then, the situation is not all roses for the connectionist approach because the goal with this approach is to build flexible systems which can handle a variety of situations (Stillings, 1995). This generalization makes the AI more useful for handling low-level human thinking and reduces its performance on higher-order reasoning. Since the approach is not based on symbols, short term memory is difficult to handle and learning constructs take more time (Sun, 1999).

Clearly, the approaches have more differences than they have similarities. The key difference between them is that the Symbolic approach seeks to give a machine the ability to handle higher-order human thinking while the connectionist approach tries to handle lower-order human thinking. However, they do have certain similarities primarily because they are approaching the same field and the same scientific principles (Sun, 1999).

Despite the differences between them, it is clear that the only AI projects that have met with some success are those which have gone to combine the two approaches and created machines with a certain amount of both higher and lower order thinking abilities. This is because the strengths of the two approaches are complementary, and using them in combination means that the AI construct would be more powerful, versatile and able in handling problems that it is asked to solve (Sun, 1999). Undoubtedly, future computers and decision support tools would require AI to be embedded as a part of the system, and the creation of AI which uses both approaches simultaneously is warranted.

In conclusion, connectionist models and symbolic models compete for dominance in the field of artificial intelligence, yet they have some complimentary features which allow the approaches to be combined in certain cases. I believe that the future of artificial intelligence creation will depend on finding methods and means which allow both approaches to be used simultaneously thereby creating machines that have both higher and lower order thinking capabilities. It has been repeated often enough that such a combined approach is theoretically possible, and there is no reason why we should not be able to create it (Smolensky, 1988). It only remains to be seen when we are able to successfully combine the two to create a truly intelligent entity.