In previous articles we watched as AI research went from the frivolity of trying to tell a human from a machine or mimicking the architecture of the brain to software mimicking the thought process.
If the first advances came from physicists, biologists, physiologists and mathematicians, the next big leap would come from an unexpected place, political science.
Table of Contents
Simon and rationality
If you studied Business Administration you probably had to cradle a fat book called administrative behavior. For what is usually the bibliography of the race, it is a book that is quite useful and interesting, although a bit dense.
The author is a gentleman he would receive the Nobel Prize in Economics for refuting one of the most beloved dogmas of Economic Science. That of the rational consumer.
Graduated in Political Science his career began studying municipal administrations and after a brief stint in the administration body of the Marshall Plan, he co-founded and taught in the Industrial Administration graduate program of what is now known as Carnegie Mellon University.
What is the point in common between bureaucracies and Artificial Intelligence? The process of decision making.
The classical economists always affirmed that we are rational decision makers. In other words, before a series of alternatives, businessmen or consumers, we will choose the option that maximizes the benefits or reduces the costs the most. The conclusion of this is that given the same series of alternatives and circumstances we will all make the same decision.
Simon downplayed the scope of that supposed rationality. He argued that the decision maker never considers all the available alternatives and that not all of us use the same criteria when evaluating them. What we do do is apply the same criteria to all problems as if it were a cooking recipe. That was the basis of heuristics or rule-based programming.
Another contribution from Simon adopted by Artificial Intelligence It is the division of goals into smaller subgoals. Reaching the subgoals makes it easier to reach the overall goal.
The first Artificial Intelligence software
With the help of Allen Newell, a Physics graduate, and C Shaw, an actuary turned computer programmer, Simon started the development of Logic Theorist, considered the first Artificial Intelligence program in history.
Although the original intention was for the program to solve chess or geometric problems, they finally used it to solve theorems of a well-known Mathematics book. Nevertheless, unlike the Turing machine, the objective was not to solve mathematical problems but to emulate the way in which humans through selective heuristics determined the next step what they had to do.
The search for the correct answer can be represented graphically as a tree-like structure.. This graph is known as a search tree.
At the root of the search tree is the initial hypothesis. Branches come from the root in which variations of the initial hypothesis are located, which are the result of applying the rules of logic to it. Other manipulations are applied to each of the branches, generating sub-branches. The process is repeated until the desired conclusion is reached.
The objective of the program of Simon and his companions was not the proof of the theorem but to find the path that would lead to that proof.. The application explored the tree according to certain pre-set rules to find the branch that was most likely to lead to the correct result. He kept repeating the process until he found the right path.
If the first attempts at Artificial Intelligence were on the side of mimicking the architecture of the brain, Simon and his colleagues went the other way. They mimicked how a computer works with people. Before beginning the coding task, a group of students joined by Simon's wife and children received cards with the subroutines and logic rules expressed in English and simulated the behavior of the program components.
Be the first to comment