Knowledge-based AI techniques

Knowledge-based AI

In “knowledge-based AI” (also known as “GOFAI” for “good old fashioned AI”, sometimes also referred to as “symbolic” or “rule-based AI”), knowledge is explicitly represented in the computer with the help of general facts and rules.
 
From this “knowledge base”, such a system can derive corresponding outputs in response to queries by drawing logical conclusions or by systematically searching the field of facts stored with semantic links.
 
Well-known examples of knowledge-based AIs include the results of the Google knowledge graph, which the search engine giant presents at the top of its search results, see the figure below (especially the info box on the right) or the chess computer  "Deep Blue", which was able to defeat the then human chess champion Garri Kasparov under tournament conditions in 1997. 

Image: Results of a query based on Google Knowledge Graphs

Knowledge-based AI can be used wherever it is possible to precisely map a manageable area of knowledge. A key advantage is that the knowledge is explicitly available and the AI's solutions can be logically understood. In practice, however, there are limits to such an explicit representation of knowledge: Some tasks require a lot of “implicit” knowledge that cannot be formalized accordingly and retrieved in time at runtime. If we want to recognize a cat in a picture, we look for four paws, fur, whiskers, etc. However, writing down explicit rules for all this is anything but easy. What is a general description for recognizing four paws, fur, whiskers, etc. that a computer can process? In addition, we often have to deal with uncertainties in reality, so perhaps not all four paws are always visible. Although such decisions are intuitively easy for us humans, it is often difficult or impossible to explicitly represent this knowledge within the computer.  For this type of problem, where knowledge cannot be represented explicitly, the technique of machine learning is successfully applied.