En our delivery Previously we told how Simon, a Political Science theorist together with a physicist named Newell and an actuary turned programmer named Shaw began the construction of the first artificial intelligence program known as Logical Theorist. This required the invention of the first specific programming language for Artificial Intelligence
We had left this story with the trio together with collaborators and family simulating the behavior of the different parts of the program using people and handwritten cards.
After several simulations like these, the program was implemented on a real computer. The test was successful as the software managed to prove thirty-eight theorems of one of the chapters of the book Principia Matemática by Russell and Whitehead. Even in one of the cases (and without having specific instructions to do so) he found a way to test it much more "elegant" than the authors of the book.
The first programming language for Artificial Intelligence
The fact that Simon and his team took so long to write their program is because they needed a specific programming language that had enough power and flexibility for their purposes. That language was called IPL (Information Processing Language) and it first introduced the list processing technique for programming.
IPL differed from the high-level languages of the time in that it did not require symbols to be defined beforehand and it had the ability to associate and modify symbol structures.
The so-called list processing technique consists of storing each piece of information together with directions on how to find pieces of information associated with them. By changing the indications, new associations can be built.
The "General Problem Solver"
To create their next software, Simon and Newell decided to try a different approach. At the time, a psychological investigation was circulating that invited participants to explain aloud the way in which they solved logical problems. The duo discovered that these forms were completely different from the ones used by their software so they decided to do their own version of the investigation and create software based on the methods described by the participants. The program (known as GPS for General Problem Solver) was coded based on an organization of information and heuristics independent of the tasks they were asked to perform.
This new methodology received the name of "Means-to-Ends Analysis" and consists of compare the current situation with the ideal and take actions that reduce the difference between them and then reassess until the difference is reduced to zero. This methodology allows the program to react to changes in the variables of the problem. The programmer indicates the problem and a so-called difference table in which the possible courses of action are indicated and in what circumstances they are.
GPS was able to break down a problem into subproblems and apply the backtracking approach, that is to say that if one path did not work, he would go back and follow another.
During the 11 years it was in operation, GPS solved puzzles, performed symbolic integration, and broke secret codes.
While Simon and Newell were entertaining themselves with this, a student named Robert K. Lindsay developed a program known as SAD SAM. the soft was able to extract information from sentences of the type "Juan is Pepa's son" and "Juan is Alberto's brother" and build a family tree ofeducing that Alberto is also Pepa's son (I have no idea how he would manage with the stepfamilies of today's world.
Of course, the giant of the computer industry at the time, IBM, could not stay out of research on artificial intelligence, a field that in the middle of the Cold War was already revealing enormous potential for military applications and, in the next article we will talk about his first contributions in the field.