General Problem Solver with Artificial Intelligence (AI)
The General Problem Solver
In 1959, Allen Newell and Herbert A. Simon took a new approach to Artificial Intelligence. Their goal was to develop a computer program that could function as a universal problem solver.
Newell and Simon - Courtesy Carnegie Mellon University Libraries
In theory, their general problem solver (GPS) would be able to solve any problem that could be presented in the form of specific types of mathematical formulas that are useful in programming logic. This type of problem would include geometric proofs, which start with definitions, axioms (statements accepted as fact), postulates, and previously proven theorems, and use logic to arrive at reasoned conclusions.
The Tower of Hanoi
One of the problems GPS solved was the Tower of Hanoi — a game or puzzle consisting of three rods and a number of disks of different sizes, which can slide onto any rod.
When you start, all the disks are on one rod, ordered from largest to smallest from the bottom up. The goal is to move the entire stack to another rod in the least number of moves following these rules:
● Move only one disk at a time.
● Do not place a larger disk on top of a smaller one.
● Each move consists of taking the top disk from one stack and placing it on an empty rod or on the top of an existing stack.
The minimum number of moves to solve the Tower of Hanoi is 2n – 1, where n is the number of disks, so for three disks, the minimum number of moves is (2 x 2 x 2) – 1 = 7.
The Physical Symbol System Hypothesis
One of the key parts of the general problem solver was what Newell and Simon called the physical symbol system hypothesis (PSSH). According to Newell and Simon, "A physical symbol system has the necessary and sufficient means for general intelligent action." Such a system would be able to take patterns (symbols), combine them into structures (expressions), and manipulate them using various processes to produce new expressions.
Newell and Simon believed that human intelligence was no more than a complex physical symbol system. They thought that a key part of human reasoning consisted merely of connecting symbols — that our language, ideas, and concepts were just broad groupings of interconnected symbols. For example, when we see a chair or a picture of a chair, we associate it with the act of sitting. When we smell smoke, we associate it with fire, which is associated with danger.
Newell and Simon argued that by feeding a machine enough physical symbols, creating a sufficient number of associations, and putting rules in place for combining symbols into structures and manipulating them to create new expressions, machines could be made to "think" like we humans do. This theory forms the basis of what drives most of machine learning and artificial intelligence to this day.
Refuting the Theory: The Chinese Room Argument
Not everyone buys into the notion that a physical symbol system is necessary and sufficient for human intelligence. In 1980, philosopher John Searle argued that merely connecting symbols could not be considered intelligence. To support his argument against the idea that manipulating physical symbols constituted intelligence, he presented what is commonly referred to as the Chinese room argument.
Imagine yourself, an English-only speaker, locked in a room with a narrow slot on the door through which you can pass notes. You have a book filled with long lists of statements in Chinese, and the floor is covered in Chinese characters. You are instructed that upon receiving a certain sequence of Chinese characters, you are to look up a corresponding response in the book and, using the characters strewn about the floor, formulate your response.
Someone outside the room who speaks and writes fluent Chinese writes a note on a sheet of paper and passes it to you through the slot on the door. Following the instructions you were given, you look up a response in the book, copy the response using characters from the floor to create your note, and pass it through the slot to the person who delivered the original message.
The native speaker may believe that the two of you are communicating and that you know the language. However, Searle argues that this is no proof of intelligence, because you have no understanding of the messages you are receiving or sending.
You can try a similar experiment with your smart phone. If you ask Siri or Alexa how she's feeling, she will answer your question even though she feels nothing at all. She doesn't even understand the question. This artificially "intelligent" being is merely matching your question to what is considered an acceptable answer and delivering that answer to you.
Combinatorial Explosion: A Major Obstacle
A huge obstacle to achieving artificial intelligence through a physical symbol system is what's known as combinatorial explosion — the rapid growth of symbol combinations that makes pattern-matching increasingly difficult. Combinatorial explosion is far greater than exponential growth. The formula for exponential growth can be expressed as y = 2x, whereas the formula for combinatorial explosion is y = x! (the factorial of x). For example, if x = 20, then
exponential growth:?y = 2x = 220 = 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 = 1,048,576
combinatorial explosion: y = x! = 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 x 10 x 11 x 12 x 13 x 14 x 15 x 16 x 17 x 18 x 19 x 20 = 2,432,902,008,176,640,000
With each added symbol, the number of combinations increases dramatically. Considering all the possible combinations would require immense computational resources over a considerable amount of time.
Even with these challenges, pattern-matching has remained the cornerstone of artificial intelligence, regardless of whether it is even in the same ballpark as human intelligence.
Frequently Asked Questions
What is a General Problem Solver (GPS) in the context of AI?
The General Problem Solver (GPS) is an AI program designed by Newell and Shaw in 1957, meant to work as a universal problem-solver. It aimed to solve a wide range of problems using “means-ends analysis,” a type of heuristic.
How does GPS use algorithms for problem solving?
GPS uses a specific algorithm known as “means-ends analysis” to solve problems. This algorithm helps GPS identify the differences between the current state and the goal state and then applies operations to reduce these differences.
Why is the General Problem Solver significant in the history of AI?
GPS was the first AI program to be called a general problem-solver, intended to solve a wide range of problems. It paved the way for more advanced AI systems by demonstrating that computers could be used for general problem-solving.
Can GPS interact with the world outside its program?
GPS itself cannot directly interact with the world; however, it models problem-solving within a specific domain using pre-defined heuristics and operators to reach a goal state.
What kind of problems could the General Problem Solver handle?
The General Problem Solver could handle a wide range of theoretical problems, particularly those that can be broken down into smaller, more manageable sub-problems. However, it still faced limitations and was far from achieving human-like intelligence.
What programming language was used to develop GPS?
GPS was developed using an early form of a programming language, which was essential for its design and function as an AI program. The exact nature of the programming languages has evolved significantly since then.
How does the Chinese Room Argument relate to the General Problem Solver?
The Chinese Room Argument, proposed by John Searle, criticizes the notion that a program like GPS, which manipulates symbols according to formal rules, can be said to "understand" or possess intelligence. The argument suggests that such programs are far from achieving true artificial intelligence.
What are the limitations of the General Problem Solver?
The limitations of GPS include its reliance on predefined heuristics and its inability to deal with complex, real-world scenarios effectively. It demonstrated that while AI could solve structured problems, it was far from achieving the flexibility and understanding needed to solve unstructured, real-world problems.
This is my weekly newsletter that I call The Deep End because I want to go deeper than results you’ll see from searches or AI, incorporating insights from the history of data and data science. Each week I’ll go deep to explain a topic that’s relevant to people who work with technology. I’ll be posting about artificial intelligence, data science, and data ethics.?
This newsletter is 100% human written ?? (* aside from a quick run through grammar and spell check).
More Sources:
Interdisciplinary Researcher and Educator focused on Technology and Ethics
1 个月I do think there is still a place for GPS. I see a trade-off of pros and cons between GPS and Neural Networks. I think this why my current research is centered on hybrid models.
Zonal Head, Procurement @HPCL | AI, Data Science & Business Intelligence Enthusiast | CPP - Certified Procurement Professional (IIPMR) | Lean Six Sigma Black Belt (CSSC) |
1 个月Very informative
Automation Specialist (Python & Analytics) at Capgemini ??|| Master's in Data Science || PGDM (Product Management) || Six Sigma Yellow Belt Certified || Certified Google Professional Workspace Administrator
1 个月Excellent reading experience ??