In 1943, neurophysiologist Warren McCulloch of the University of Illinois and mathematician Walter Pitts of the University of Chicago wrote a book called Neural Networks and Automata. They said that each neuron in the brain is a simple digital processor and that the brain is a type of computer. Likewise, connectionism is a method in cognitive science that tries to explain mental phenomena with the help of artificial neural networks (ANN).

What is connectionism?

Connectionism tries to explain how intelligence works by using artificial neural networks. Furthermore, connectionism is a theory of cognition that says that learning happens when the strength of connections changes from experience. This theory is on the idea that signals can be sent to different places simultaneously through connections that can be numerical.

Connectionists and Classicists

The classical view that human cognition is similar to symbolic computation in digital computers has been the most common in the past forty years. In classical theory, information is of strings of symbols, just like how we store data in computer memory or on paper. On the other hand, the connectionist says that data is non-symbolic in the weights. The classicist thinks that thinking is like digital processing, where strings are in order based on what a (symbolic) program says to do. The connectionist sees mental processing as the gradual and changing activity growth in a neural net. Each unit's training depends on how strong its connections are and how much its neighbours are doing.

What is Connectionist AI?

People focused on the symbolic type of AI for the first few decades of its history, but now people are more interested in a newer model called connectionist AI. It makes a model of how AI works based on how the human brain works. A single neuron is shown in this model by something called a "perceptron."

A system made with connectionist AI gets smarter as it is exposed to more data and learns the patterns and relationships. In contrast, humans write the code for symbolic AI. An artificial neural network is a type of connectionist AI. Each has hundreds of artificial neurons or processing elements that work as single units. They are in layers, with weights acting as links between the layers.

When to choose connectionist AI?

When people have a lot of training data to feed into the algorithm, connectionist AI is a good choice. Even though this model gets smarter the more it is used, it needs a solid base of correct information to start learning. This kind of AI is often used in the healthcare field, especially when there are a lot of medical images that humans have checked for accuracy or annotated to give context. But it can't often explain how it came up with a solution. So, people shouldn't choose it as their only or main option if they need to explain to someone else why the AI came to the conclusion it did. For example, researchers could use connectionist AI to decide the fate of a person accused of murder. In that case, most people would think it was cruel and unfair to rely on AI in that way without knowing why the algorithm came to that conclusion.

Conclusion

The primary paradigm has branched several different approaches since the 1980s when connectionist research was in its "Golden Age". From 1980 to 1995, connectionism seemed to have grown up and toned down its goals. Some of the benefits of the connectionist approach are that researchers can use it for a wide range of tasks. Its structure is similar to that of biological neurons, it doesn't need a lot of innate systems, and it can break down naturally. Some of the problems are that it's hard to figure out how ANNs process information or account for mental representations of different parts, making it hard to explain things at a higher level. Also, the problem you are trying to solve has a lot to do with which algorithm you should use.