MYCIN was an expert system that used AI to identify bacteria causing severe infections, such as meningitis, and to recommend treatments. MYCIN employed a simple inference engine, and a knowledge base of around 600 rules. Shown here are Streptococcus pneumoniae bacteria that can cause meningitis.

1959

KNOWLEDGE REPRESENTATION AND REASONING

“For a system to be intelligent,” writes computer scientist Nils Nilsson, “it must have knowledge about its world and the means to draw conclusions from, or at least act on, that knowledge. Humans and machines alike therefore must have ways to represent this needed knowledge in internal structures, whether encoded in protein or silicon.” Nowadays, much of the attention on AI seems to be on machine learning and statistical algorithms for applications such as image recognition. Nevertheless, logic-based knowledge representation and reasoning (KR) still has a big role to play in many areas.

KR is the field of AI concerned with representing information in such a way that computer systems can efficiently use the information to make medical diagnoses and legal recommendations, as well as to facilitate intelligent dialog systems such as Siri on the iPhone or Alexa on the Amazon Echo. As just one example, a semantic network is sometimes used as a form of KR to represent semantic (i.e., meaning) relationships between concepts. These semantic networks often take the form of graphs, with vertices representing concepts and edges (i.e., connecting lines) indicating the semantic relationships between them. KR also has application in automatic reasoning, including the automated proving of mathematical theorems.

Some of the early work in AI KR includes the General Problem Solver, a computer program developed in 1959 by Allen Newell (1927–1992), Herbert Simon (1916–2001), and colleagues to analyze goals and solve simple general problems (e.g., the Tower of Hanoi). Later, the Cyc project—started in 1984 by Douglas Lenat (b. 1950)—employed numerous analysts documenting various areas of commonsense reasoning to help AI systems perform human-like reasoning (e.g., the Cyc inference engine employs logical deductions and inductive reasoning). Today, AI researchers in the area of KR deal with many issues, including ensuring that the knowledge base can be updated as needed to allow efficient development of new inferences. The researchers are also concerned with how uncertainty and vagueness can best be addressed in KR systems.