This episode is part of “Entering Artificial Intelligence”: Lecture 3 – From Expert Systems to Deep Blue: Growth in Logical Reasoning and Search Optimization | Logical Reasoning: From Known to Unknown.
Entering Artificial Intelligence
Hello, this is Professor Wu Fei’s digital column “Entering Artificial Intelligence”. In the last lecture, we introduced the Dartmouth Conference, which marked the beginning of artificial intelligence, and the subsequent bumpy journey of AI development. In this third lecture, I have prepared for you the content titled “From Expert Systems to Deep Blue: Growth in Logical Reasoning and Search Optimization”. Albert Einstein once said, “The most significant goal in all sciences is to explain the maximum amount of empirical facts from the least number of assumptions and axioms using logical deductive reasoning methods.” Logical reasoning and search optimization are the main methods used in early artificial intelligence for problem-solving. In this lecture, I will introduce three parts: the origins of logical thought, the use of logical thought in expert systems to solve specific problems in early artificial intelligence, and finally, an analysis of Deep Blue, which defeated human chess champions, to illustrate the role of reasoning and search in artificial intelligence.
Logical Reasoning: From Known to Unknown
In 1975, Herbert A. Simon and Allen Newell jointly delivered a speech entitled “Computer Science as a Science of Empirical Exploration: Symbolic Reasoning and Search” at the Turing Award ceremony. In this speech, the two Turing Award winners argued that the objective objects and evolutionary processes existing in the real world can all be described and explained using symbols, and various “problems” can be answered through reasoning and search.
Logic is the discipline that explores, elucidates, and establishes effective reasoning principles for knowledge. It consists of two essential components: normalized knowledge and reasoning methods. The term “logic” was first introduced to China by Yan Fu, but he chose to use the term “mingxue” instead of “logic” in his translations. Later, Zhang Shizhao believed that using “mingxue” to express the ideas contained in “logic” was inappropriate, and advocated for the adoption of the term “logic” as its corresponding Chinese word, which has been in use ever since.
Logical reasoning is an important manifestation of human intelligence, with its foundation and fuel being normalized knowledge (such as concepts and relationships). The poet, philosopher, and logician Ramon Llull proposed a normalized description of knowledge called the “Tree of Knowledge” at the end of the 13th century. In the Tree of Knowledge, there are relationships of inclusion and exclusion, association and disassociation among knowledge. Inspired by this, Leibnitz pointed out in his 1666 paper “On the Combinatorial Art” that human thought can be generated by the combination of basic concepts, meaning that human thought, no matter how complex it seems, is merely a combination of simple, basic elements. Therefore, by performing operations such as combinations on the symbols in what Descartes referred to as the “alphabet of human thought” (the Tree of Knowledge), a thinking machine can be constructed to achieve computational thinking, making a leap from “theological debate” to “philosophical reasoning”.
If normalized knowledge is the cornerstone of logical reasoning, then reasoning methods are the “engine” of logical reasoning. Reasoning is the process of deriving new judgments from one or more known premises. Therefore, it is necessary to study rigorous logical reasoning methods (such as induction and deduction) to ensure that the conclusions drawn from correct premises are valid. The “syllogisms” proposed and established by Aristotle are a famous logical reasoning method. Simply put, syllogisms derive conclusions from major and minor premises. For example, given the major premise “All men are mortal” and the minor premise “Socrates is a man”, using the reasoning technique of “syllogism” allows us to arrive at the previously unknown conclusion “Socrates is mortal”. Logical reasoning builds a bridge from the known and observed to the unknown!
Once knowledge is normalized, reasoning rules can be used to simulate human thinking from the known, in a computational manner. This is akin to what Hobbes stated in his work “Leviathan”: philosophy is “knowledge obtained through true reasoning about the causes and effects of things.” What I refer to as “reasoning” is computation.
Product Name: Entering Artificial Intelligence | Fifteen Lectures on General Knowledge
Lead Person: Wu Fei
Produced by: Higher Education Press,Higher Education Audio-Video Publishing House
Collaborating Institutions: Shanghai Advanced Institute of Zhejiang University,Shanghai Artificial Intelligence Laboratory Intelligent Education Center
Production Date: January 2022
