Semantic Reasoning—the forgotten half of AI
From the dawn of artificial intelligence (AI), there were two main approaches to making systems knowledgeable—explicitly teach them what they need to know or have them learn from experience. According to Larry Lefkowitz, PhD, Chief Scientist, AI Practice, at Sapient, this dichotomy still holds today.
Machine learning (ML), the approach based on learning from experience, has been far more prominent in recent years. It has demonstrated great value across a range of classification and prediction tasks; hence, it is almost synonymous with AI. However, AI can do much more.
There are many tasks that require explicit reasoning. This reasoning approach uses the knowledge of the problem domain and often, of the world in general. Let us consider, for example, tasks such as assembling a project team, coordinating a response to a natural disaster, or simpler tasks such as preparing a kid’s lunch or understanding a news article. In each example, the task is more complex than selecting from a set of possible options or determining the value of a variable. In addition, the execution of each task relies on the information or knowledge of the “input data.” Significantly, executing these tasks requires knowledge to be modeled in a way that a machine can efficiently process it, i.e., as an ontology or a knowledge base.
The semantic modeling approach relies on explicit, human-understandable representations of concepts, relationships, and rules that comprise the desired knowledge domain. The knowledge and the corresponding levels of expertise and expenses for this approach’s modeling can be represented with various degrees of fidelity. Fortunately, there is no need to build these models from scratch. Often, extending existing knowledge models, including domain-specific ontologies (such as the Financial Industry Business Ontology (FIBO) or numerous healthcare ontologies) and broader knowledge bases such as Cyc, SUMO, or the DBpedia Ontology, might prove beneficial.
Click here to find out why, then, has machine learning, rather than semantic modeling and reasoning, dominated the AI mindshare.