Artificial Intelligence (AI)

There are two main approaches to Artificial Intelligence: Symbolic Artificial Intelligence, which is inspired by Computer Science, and Sub-Symbolic Artificial Intelligence, which is inspired by Neurobiology.

Symbolic AI

Symbolic AI, also known as Good Old Fashion Artificial Intelligence (GOFAI), focuses on an information processing metaphor manipulating symbols using some engine to achieve the outcome. Outputs are obtained by processing information using a symbolic (rule) structure. Sub-Symbolic AI, also known as Connectivism, focuses on nodes, links, activation, and weights to emulate human neural activity. The outcome is achieved by the computations of these Artificial Neural Networks. Outputs are obtained by parallelly processing information using simple mathematical calculation realized by the nodes.

Symbolic AI is often preferred for dealing with abstract problems and is more useful for explaining people’s thoughts. It is generally easier to explain, easier to debug and easier to control. But introspection is needed to encode the knowledge. It is a “not so big data” approach to Artificial Intelligence (AI). The Decision Model and Notation (DMN) is an international open standard used by Trisotech for modeling and automating Symbolic AI.

Sub-Symbolic AI

Sub-Symbolic AI is often preferred for perceptual problems and is more useful for relating it to neuroscience. It is generally easier to scale up, requiring less knowledge upfront, more robust against noise and offering better performance. But its inner workings are often opaque, locked away in a “black box” that is hard to peer into. It is a “big data” approach to Artificial Intelligence. The current resurgence of interest toward Artificial Intelligence is mostly driven by Sub-Symbolic AI approaches such as Deep Neural Networks (DNN) and Machine Learning (ML). The Predictive Model Markup Language (PMML) is an open standard used by Trisotech for integrating Sub-Symbolic AI into business automation.