The Second World War saw massive scientific contributions and technological advancements. Innovations such as radar technology, the mass production of penicillin, and the jet engine were all a by-product of the war. More importantly, the first electronic computer (Colossus) was also developed to decipher encrypted Nazi communications during the war. After the war, the desire to achieve machine intelligence continued to grow.

As humans, we wanted to enable computers with human-like abilities. The primary motivation behind Artificial Intelligence (AI) systems has always been to allow computers to mimic our behavior, to enable machines to think like us and act like us, to be like us. Over the years, this motivation has remained virtually unchanged. However, the methodology and the mindset of how we approach AI has gone through several phases throughout the years.

While today, we rely on deep Neural Networks (NNs) (also known as Connectionism AI) to allow a machine to automatically teach itself how to solve a particular task, early AI systems were heavily ruled, and logic-based methods were used to teach computers human behavior. These systems are referred to as Symbolic AI. Symbolic AI is one of the earliest forms of AI. In his 1985 book Artificial Intelligence: The Very Idea, John Haugeland coined Good Old-Fashioned AI (GOFAI) as a reference to Symbolic AI. It dominated the computer science and AI fields from the 1950s all the way through to the 1970s. One of the first successful implementations of Symbolic AI dates to 1951, when Christopher Strachey, a British computer scientist, wrote over 2,000 instructions to teach a computer how to play the game of checkers (draughts). Today, Symbolic AI has been leveraged to solve many problems, including Natural Language Processing (NLP) applications and Expert System (ES) implementations.

This chapter aims to understand the underlying mechanics of Symbolic AI, its key features, and its relevance to the next generation of AI systems.

The first objective of this chapter is to discuss the concept of Symbolic AI and provide a brief overview of its features. Symbolic AI is heavily influenced by human interaction and knowledge representation. We will discuss how this shaped Symbolic AI. We will then examine the key features of Symbolic AI, which allowed it to dominate the field during its time. After that, we will cover various paradigms of Symbolic AI and discuss some real-life use cases based on Symbolic AI. We will finally discuss the main challenges when developing Symbolic AI systems and understand their significant pitfalls.

This chapter will go through the following topics:

• Defining Symbolic AI
• Knowledge representation through symbols and signs
• Critical features of Symbolic AI
• The different paradigms and applications of Symbolic AI
• Limitations and pitfalls of Symbolic AI

Let us get started with defining what we mean by Symbolic AI.

Symbolic AI, GOFAI, or Rule-Based AI (RBAI), is a sub-field of AI concerned with learning the internal symbolic representations of the world around it. The main objective of Symbolic AI is the explicit embedding of human knowledge, behavior, and “thinking rules” into a computer or machine. Through Symbolic AI, we can translate some form of implicit human knowledge into a more formalized and declarative form based on rules and logic.

Understanding explicit and implicit knowledge

Explicit knowledge is any clear, well-defined, and easy-to-understand information. Explicit knowledge is based on facts, rules, and logic. An excellent example of explicit knowledge is a dictionary. In a dictionary, words and their respective definitions are written down (explicitly) and can be easily identified and reproduced.

Implicit knowledge refers to information gained unintentionally and usually without being aware. Therefore, implicit knowledge tends to be more ambiguous to explain or formalize. Examples of implicit human knowledge include learning to ride a bike or to swim. Note that implicit knowledge can eventually be formalized and structured to become explicit knowledge. For example, if learning to ride a bike is implicit knowledge, writing a step-by-step guide on how to ride a bike becomes explicit knowledge.

In the Symbolic AI paradigm, we manually feed knowledge represented as symbols for the machine to learn. Symbolic AI assumes that the key to making machines intelligent is providing them with the rules and logic that make up our knowledge of the world.

Humans, symbols, and signs

Symbolic AI is heavily inspired by human behavior. Humans interact with each other and the world through symbols and signs. The human mind subconsciously creates symbolic and subsymbolic representations of our environment. It’s how we think and learn. Our world is full of fuzzy implicit knowledge. Objects in the physical world are abstract and often have varying degrees of truth based on perception and interpretation. Yet somehow, we can still knowingly navigate our way through life. We can share information and teach each other new skills. We can do this because our minds take real-world objects and abstract concepts and decompose them into several rules and logic. These rules encapsulate knowledge of the target object, which we inherently learn.

This approach has been our way of life since the beginning of time. Thomas Hobbes, a British philosopher, famously said that thinking is nothing more than symbol manipulation, and our ability to reason is essentially our mind computing that symbol manipulation. René Descartes also compared our thought process to symbolic representations. Our thinking process essentially becomes a mathematical algebraic manipulation of symbols. Think about it for a second. What happens when we think? We start to formulate ideas. Ideas are based on symbols that represent some other object. For example, the term Symbolic AI uses a symbolic representation of a particular concept, allowing us to intuitively understand and communicate about it through the use of this symbol. Then, we combine, compare, and weigh different symbols together or against each other. That is, we carry out an algebraic process of symbols – using semantics for reasoning about individual symbols and symbolic relationships. Semantics allow us to define how the different symbols relate to each other. They also enable us to interpret symbolic representations.

To properly understand this concept, we must first define what we mean by a symbol. The Oxford Dictionary defines a symbol as a “Letter or sign which is used to represent something else, which could be an operation or relation, a function, a number or a quantity.” The keywords here represent something else. Symbols are merely explicit references to implicit concepts. We use symbols to standardize or, better yet, formalize an abstract form. This process is also commonly referred to as conceptualization. At face value, symbolic representations provide no value, especially to a computer system. However, we understand these symbols and hold this information in our minds. In our minds, we possess the necessary knowledge to understand the syntactic structure of the individual symbols and their semantics (i.e., how the different symbols combine and interact with each other). It is through this conceptualization that we can interpret symbolic representations.

Let’s consider a newborn child, for example. At birth, the newborn possesses limited innate knowledge about our world. A newborn does not know what a car is, what a tree is, or what happens if you freeze water. The newborn does not understand the meaning of the colors in a traffic light system or that a red heart is the symbol of love. A newborn starts only with sensory abilities, the ability to see, smell, taste, touch, and hear. These sensory abilities are instrumental to the development of the child and brain function. They provide the child with the first source of independent explicit knowledge – the first set of structural rules.

With time and sensory experiences, these structural rules become innate to the human mind, promoting further psychological development. The child begins to understand and learn rules – such as if you freeze water, it will eventually become ice. Here, ice is purely a label representing frozen water. Fire is hot, and if you touch hot, it will hurt. The child will begin to understand the physical and psychological world one rule at a time, continuously building the world’s symbolic representation by learning newer and perhaps more complex syntactic and semantic logical rules. Eventually, the child will be able to communicate these symbolic representations with other humans and vice versa. As humans, we widely encourage the formalization of knowledge. Therefore, we are entirely dependent on symbolic knowledge. Some symbolic examples include the following:

• Phonograms: Any symbol (typically a letter or character) used to represent vocal sounds or linguistics (or both). Phonograms are used to describe the pronunciation of a particular word. For example, the term dog has the phonogram d/o/g (3), while the word strawberry has the phonogram s/t/r/aw/b/err/y (7).
• Logograms: Any linguistic symbol (a letter or sign) that is used to represent any complete word or phrase. Logograms do not consider the phonetics of the said word or phrase. The $ (dollar) and & (ampersand) signs are good examples of logograms.
• Pictograms: Any schematic graphical (pictorial) symbol representing an entire word, phrase, or concept. Gender symbols and graphical charts are two examples of pictograms.
• Typograms: Any symbol, typically linguistic, that represents the definition or implication of a particular word through manipulating its letters. A typogram essentially becomes a symbol that encapsulates another symbol. For example, a typogram of the word missing might be m-ss-n-g. This is because the “i”s are missing from the word.
• Iconograms: Any graphical symbol that is used to represent an entire word, phrase, or concept. Iconograms differ from pictograms because they tend to be more graphically and artistically detailed. A drawing of a flower or a view of a map are examples of iconograms.
• Ideograms: Any symbol that represents a word or concept. Ideograms are often in geometric shapes, which differ from other graphical symbols. As the name suggests, while they can define words, they are typically used to represent ideas. Examples of ideograms include a traffic stop sign or a no smoking sign.

Irrespective of our demographic and sociographic differences, we can immediately recognize Apple’s famous bitten apple logo or Ferrari’s prancing black horse. Even our communication is heavily based on symbols.

Figure 2.1 depicts the Sumerian language, which is recognized as being the first human language, dating back to circa 3100 BC (source: https://www.history.com/topics/ancient-middle-east/sumer#:~:text=The%20Sumerian%20language%20is%20the,for%20the%20next%20thousand%20years.). Its alphabet comprised graphical symbols representing various nouns, objects, and actions of the time. This is perhaps the best representation of the “thinking in symbols” concept.

Figure 2.1: The Sumerian language. Image by Mariusz Matuszewski on Pixabay

Humans thrive on interaction, and formalizing and declaring representations of implicit concepts and abstract objects is crucial to universal communicative abilities. The ability to create symbolic representations of the world around us might be a differentiating trait of intelligence. Recently, scientists have also found that other animals, including primates, dolphins, and horses, could understand and utilize human symbols to interact and communicate with humans. In one of their experiments, a group of horses was shown three symbols representing “no change," “add a blanket,” and “remove a blanket.” The horses could choose what they wanted based on the weather conditions by pointing toward the respective symbol. This feat is truly remarkable and drives the point home of the power behind symbols!

Now that we’ve discussed the vital role that symbols and signs play in everyday life, how does all this tie together with Symbolic AI?

This symbolic philosophy was highly influential in the field of AI. The first examples of AI programs, such as the Logic Theorist (an AI program written in 1956 by Allen Newell, Herbert A. Simon, and Cliff Shaw that was able to prove theorems on the same level as a human mathematician) and the General Problem Solver (an AI programmed in 1957 by Herbert A. Simon, J. C. Shaw, and Allen Newell that used symbolic rule representations of problem knowledge as input to solve general tasks), involved symbolical processing in conquering the quest of achieving machine intelligence.

The concept of intelligence

Before we proceed any further, we must first answer one crucial question – what is intelligence? At face value, this question might seem relatively simple to answer. However, the term intelligence is complex to define. Intelligence tends to become a subjective concept that is quite open to interpretation.

Consider a popular TV talk show. In one of its popular segments, the host introduces two prodigy children: A and B. Child A can solve every mathematical problem in the world in record time. Child B can understand and speak every language like it’s their native tongue. The host starts by introducing the audience to child A and, for the sake of entertainment, asks the child to solve a couple of math problems, each increasing in difficulty. The child answers correctly every time, and the audience is stunned and speechless. Everyone is in awe of this child’s intelligence. Then, child B is also brought out by the host. The host asks child B to solve the same math problems as child A. Child B is not able to solve them correctly. The audience is not impressed. Is child B intelligent? To the audience, probably not. But to anyone who has witnessed the skills of child B, then probably the answer would be a strong and resounding yes.

“Everybody is a genius, but if you judge a fish by its ability to climb a tree, it will live its whole life believing it’s stupid.”

– Unknown author (commonly misattributed to Albert Einstein)

What is the takeaway here?

The definitions of intelligence, while being super subjective, essentially become a direct association and measure of the following:

• The problem we are trying to solve
• The context and environment of that problem

Although it is complex to define, humans subconsciously understand that intelligence is directly measured by how well you can do the task you are interested in. Intelligence is associated with reducing the significance and effect of our target problem. So, if we want a machine to be intelligent, it must solve a specific problem or task. But how can we teach the machine to solve a task? As we previously mentioned, early forms of AI were all about enabling computers to mimic human behavior. In short, it would allow machines to think.

Humans think in symbols. Computers operate using symbols. Therefore, computers can be thought to think.

Our journey through symbolic awareness ultimately significantly influenced how we design, program, and interact with AI technologies.

Towards Symbolic AI

Symbolic AI allows a machine to manipulate symbols about our world. The premise behind Symbolic AI is using symbols to solve a specific task. In Symbolic AI, we formalize everything we know about our problem as symbolic rules and feed it to the AI. Note that the more complex the domain, the larger and more complex the knowledge base becomes.

Symbolic AI leverages factual logic computation and comparison. Symbolic AI is more concerned with representing the problem in symbols and logical rules (our knowledge base) and then searching for potential solutions using logic. In Symbolic AI, we can think of logic as our problem-solving technique and symbols and rules as the means to represent our problem, the input to our problem-solving method. The natural question that arises now would be how one can get to logical computation from symbolism. We do this by defining symbol relations.

Understanding symbolic relations

Relations allow us to formalize how the different symbols in our knowledge base interact and connect. For example, let us consider a hamburger. The basic hamburger is a patty in between a bun. In this case, our symbols representing the object are BUN and PATTY. The relation would then be BETWEEN. We can define this symbolic relation as BETWEEN(PATTY, BUN).

We typically use predicate logic to define these symbols and relations formally – more on this in the A quick tangent on Boolean logic section later in this chapter.

Let us pick the task of determining whether an object is an orange or not as an analogy. When we see an orange or any other entity, we immediately dissect it and split it into its more minor constituents – its respective symbols. We use all our senses to build the knowledge base of the orange. For example, some properties of the profile that we consider might include the following:

• Shape
• Size
• Texture
• Color
• Body
• Origin

Figure 2.2 illustrates how one might represent an orange symbolically.

Figure 2.2: A symbolic representation of an orange

We observe its shape and size, its color, how it smells, and potentially its taste. We feel its texture and investigate its structure. In short, we extract the different symbols and declare their relationships. With our knowledge base ready, determining whether the object is an orange becomes as simple as comparing it with our existing knowledge of an orange. For example, we know that an orange should be orange in color. An orange should have a diameter of around 2.5 inches and fit into the palm of our hands. An orange resembles a round object with a stem emerging from its top. We learn these rules and symbolic representations through our sensory capabilities and use them to understand and formalize the world around us.

From symbols and relations to logic rules

So far, we have discussed what we understand by symbols and how we can describe their interactions using relations. The final puzzle is to develop a way to feed this information to a machine to reason and perform logical computation. We previously discussed how computer systems essentially operate using symbols. More specifically, computer processing is done through Boolean logic.

A quick tangent on Boolean logic

Any variable that can be either TRUE (1) or FALSE (0) is said to be a Boolean variable. A computer system comprises multiple digital circuits, with components with an input voltage that is either ON (i.e., state 1) or OFF (i.e., state 0).

In Boolean logic, we evaluate and compute a set of logical propositions (also called expressions) whose final output can be TRUE or FALSE. Logical propositions use three leading logical operators (logic gates) – AND, OR, and NOT:

• AND: All propositions must be TRUE for the entire proposition to be TRUE. We multiply the values of the propositions.
• OR: At least one of the propositions must be TRUE for the entire proposition to be TRUE. Then, we add the values of the propositions.
• NOT: Reverses the state of the logical proposition. If it is FALSE, it will become TRUE.

For a logical expression to be TRUE, its resultant value must be greater than or equal to 1. If the result is 0, then the expression is said to be FALSE.

There are some other logical operators based on the leading operators, but these are beyond the scope of this chapter.

Predicate logic 101

Predicate (first-order) is a formal system heavily used in multiple domains, including computing, discrete mathematics, and philosophy. We use predicate logic to define and represent expressions and statements in a standardized way. We use the symbol := to denote a definition. On the left-hand side of the symbol is what we are trying to define, and on its right is the definition. We use the symbol ˄ to represent the AND operator, ˅ to denote the OR operator, and to replace the NOT operator. Given these three core operators, we can construct other operators due to functional completeness. For example, the NAND operator combines the AND and NOT operators. There are other symbols, such as the implication symbol (⇒).The AND, OR, and NOT operators are enough for this chapter. If the concept of predicate logic is new to you, we recommend you read more about this system.

Consequently, for a computer system to understand and process our symbolic relations, we must transform them into logical propositions. We can do this by adding logical operators to our symbolic relations. We typically refer to the logical operators as logical connectives since they connect all our symbols and their respective relations. For the sake of simplicity, let us pick a dummy example to understand this better. Consider the following statement:

People watch interesting and engaging movies

Given a specific movie, we aim to build a symbolic program to determine whether people will watch it. At its core, the symbolic program must define what makes a movie watchable. Then, we must express this knowledge as logical propositions to build our knowledge base. Following this, we can create the logical propositions for the individual movies and use our knowledge base to evaluate the said logical propositions as either TRUE or FALSE.

Step 1 – defining our knowledge base

The first step is to understand the problem we are trying to solve. Then, our problem becomes the world that we need to model. Finally, we can define our world by its domain, composed of the individual symbols and relations we want to model.

If we recall the original statement, we know that for people to watch a specific movie, the movie must be both interesting AND engaging. Therefore, let us tabulate and categorize the different components of the statement:

Table 2.1: Statement dissected into its respective symbols, relations, and logical connectives

This step is vital for us to understand the different components of our world correctly. Our target for this process is to define a set of predicates that we can evaluate to be either TRUE or FALSE. This target requires that we also define the syntax and semantics of our domain through predicate logic.

Our symbols here are People and movies. They are our statement’s primary subjects and the components we must model our logic around. Following this, we must define their binary relations.

• People watch movies
• A movie can be interesting
• A movie can be engaging

It is also an excellent idea to represent our symbols and relationships using predicates. In short, a predicate is a symbol that denotes the individual components within our knowledge base. For example, we can use the symbol M to represent a movie and P to describe people. We can also represent relations using predicates. A predicate relationship can have one or more arguments. Let us formally define the preceding relations.

Our first relation is that people watch movies. We can denote people watching movies using the WATCH predicate symbol. In this case, WATCH accepts two arguments: people and movies. Therefore, we can write this down as WATCH(P, M). Next, we have two other relations that accept only a single argument. These two relations define a property that a movie can possess. We can write them as IS_INTERESTING(M) and IS_ENGAGING(M).

Furthermore, the final representation that we must define is our target objective. Let us represent IS_INTERESTING(M) with I and IS_ENGAGING(M) with E.

Using first-order logic, we can define our target T as T = I ˄ E:

T is TRUE if both I and E are also TRUE

Let us also denote two movies – X and Y:

• Movie X is not interesting but engaging (therefore, X = I ˄ E)
• Movie Y is interesting and engaging (therefore, Y = I ˄ E)

Recall our target objective – i.e., to determine whether a person will watch the movie. We can formally write this inference rule as an implication as follows:

Person(P) AND Movie(M) AND IS_INTERESTING(M) AND IS_ENGAGING(M) => WATCH(P, M)

If we have a person, P, and a movie, M, where the movie is both interesting and engaging, then that implies the person will watch the movie.

Step 2 – evaluating our logical relations

So far, we have defined our domain regarding symbols and relations. The next step is to build our truth table to evaluate the validity of our expressions:

Table 2.2: Logical relation evaluation using a truth table

Based on our knowledge base, we can see that movie X will probably not be watched, while movie Y will be watched. Of course, this is a trivial example to get the message across.

Defining the knowledge base requires skills in the real world, and the result is often a complex and deeply nested set of logical expressions connected via several logical connectives. Compare the orange example (as depicted in Figure 2.2) with the movie use case; we can already start to appreciate the level of detail required to be captured by our logical statements. We must provide logical propositions to the machine that fully represent the problem we are trying to solve. As previously discussed, the machine does not necessarily understand the different symbols and relations. It is only we humans who can interpret them through conceptualized knowledge. Therefore, a well-defined and robust knowledge base (correctly structuring the syntax and semantic rules of the respective domain) is vital in allowing the machine to generate logical conclusions that we can interpret and understand.

Nonetheless, a Symbolic AI program still works purely as described in our little example – and it is precisely why Symbolic AI dominated and revolutionized the computer science field during its time. Symbolic AI systems can execute human-defined logic at an extremely fast pace. For example, a computer system with an average 1 GHz CPU can process around 200 million logical operations per second (assuming a CPU with a RISC-V instruction set). This processing power enabled Symbolic AI systems to take over manually exhaustive and mundane tasks quickly.

So far, we have defined what we mean by Symbolic AI and discussed the underlying fundamentals to understand how Symbolic AI works under the hood. In the next section of this chapter, we will discuss the major pitfalls and challenges of Symbolic AI that ultimately led to its downfall.

In the early 1980s, most AI developers moved away from Symbolic AI. Symbolic AI spectacularly crashed into an AI winter since it lacked common sense. Researchers began investigating newer algorithms and frameworks to achieve machine intelligence. As a result, Symbolic AI lost its allure quite rapidly. Furthermore, the limitations of Symbolic AI were becoming significant enough not to let it reach higher levels of machine intelligence and autonomy. In the following subsections, we will delve deeper into the substantial limitations and pitfalls of Symbolic AI.

Common sense is not so common

In a nutshell, Symbolic AI has been highly performant in situations where the problem is already known and clearly defined (i.e., explicit knowledge). Symbolic AI heavily relies on explicit symbolic representations. However, the world around us is filled with implicit knowledge. Our universe is a rather abstract concept. Translating our world knowledge into logical rules can quickly become a complex task. While in Symbolic AI, we tend to rely heavily on Boolean logic computation, the world around us is far from Boolean. Most physical symbols and relations are fuzzy. They are not static but rather based on a degree of truthiness. For example, a digital screen’s brightness is not just on or off, but it can also be any other value between 0% and 100% brightness. A person can be a little hungry as opposed to completely starving. The concept of fuzziness adds a lot of extra complexities to designing Symbolic AI systems. Due to fuzziness, multiple concepts become deeply abstracted and complex for Boolean evaluation.

Additionally, it introduces a severe bias due to human interpretability. Let’s pick a simple analogy – the color cyan. For some, it is cyan; for others, it might be aqua, turquoise, or light blue. As such, initial input symbolic representations lie entirely in the developer’s mind, making the developer crucial. Recall the example we mentioned in Chapter 1 regarding the population of the United States. It can be answered in various ways, for instance, less than the population of India or more than 1. Both answers are valid, but both statements answer the question indirectly by providing different and varying levels of information; a computer system cannot make sense of them. This issue requires the system designer to devise creative ways to adequately offer this knowledge to the machine.

Inevitably, this issue results in another critical limitation of Symbolic AI – common-sense knowledge. The human mind can generate automatic logical relations tied to the different symbolic representations that we have already learned. Humans learn logical rules through experience or intuition that become obvious or innate to us. They tend to come to us naturally, without us overthinking them. For example, a child must always be younger than their parents. We close our eyes when we want to sleep. We do not eat food that smells like it has gone bad. These are all examples of everyday logical rules that we humans just follow – as such, modeling our world symbolically requires extra effort to define common-sense knowledge comprehensively. Consequently, when creating Symbolic AI, several common-sense rules were being taken for granted and, as a result, excluded from the knowledge base. As one might also expect, common sense differs from person to person, making the process more tedious.

In the real world, there are so many different levels of abstraction, hierarchies, and underlying relationships. It’s impossible to capture all these rules entirely. To start, even humans do not know all the universe’s secrets. Let us recall the orange example from Figure 2.2. Assume we pass two fruits to the Symbolic AI program: an orange and a tangerine. With the symbolic structure and relations we had previously defined, it would be rather difficult to differentiate between them. Even a human might find this task difficult, let alone a machine that feeds knowledge through logical rules devised by a human.

Although Symbolic AI paradigms can learn new logical rules independently, providing an input knowledge base that comprehensively represents the problem is essential and challenging. The symbolic representations required for reasoning must be predefined and manually fed to the system. With such levels of abstraction in our physical world, some knowledge is bound to be left out of the knowledge base.

The test of time

Another concept we regularly neglect is time as a dimension of the universe. As we all know, time changes a lot of things. Some examples are our daily caloric requirements as we grow older, the number of stairs we can climb before we start gasping for air, and the leaves on trees and their colors during different seasons. These are examples of how the universe has many ways to remind us that it is far from constant.

A Symbolic AI system is said to be monotonic – once a piece of logic or rule is fed to the AI, it cannot be unlearned. Newly introduced rules are added to the existing knowledge, making Symbolic AI significantly lack adaptability and scalability. One power that the human mind has mastered over the years is adaptability. Humans can transfer knowledge from one domain to another, adjust our skills and methods with the times, and reason about and infer innovations. For Symbolic AI to remain relevant, it requires continuous interventions where the developers teach it new rules, resulting in a considerably manual-intensive process. Surprisingly, however, researchers found that its performance degraded with more rules fed to the machine.

We might teach the program rules that might eventually become irrelevant or even invalid, especially in highly volatile applications such as human behavior, where past behavior is not necessarily guaranteed. This phenomenon is referred to as concept drift or data morphism. In short, the underlying relationships of the data shift or change. Even if the AI can learn these new logical rules, the new rules would sit on top of the older (potentially invalid) rules due to their monotonic nature. As a result, most Symbolic AI paradigms would require completely remodeling their knowledge base to eliminate outdated knowledge. This remodeling process often becomes highly convoluted and tedious. For this reason, Symbolic AI systems are limited in updating their knowledge and have trouble making sense of unstructured data.

Being the first major revolution in AI, Symbolic AI has been applied to many applications – some with more success than others. Despite the proven limitations we discussed, Symbolic AI systems have laid the groundwork for current AI technologies. This is not to say that Symbolic AI is wholly forgotten or no longer used. On the contrary, there are still prominent applications that rely on Symbolic AI to this day and age. We will highlight some main categories and applications where Symbolic AI remains highly relevant.

Expert systems

Symbolic AI has been predominantly used to design and develop ESs. IBM’s Representation, Ontology, Structure, Star (ROSS) is a great example. You can refer to https://arxiv.org/pdf/1411.4192.pdf for further reading. ROSS is an expert system platform for legal research, much like an AI lawyer. Given a natural language prompt, ROSS can sift through the law, complete court cases, and other documents, and return relevant structured data and evidence based on the query. Symbolic AI is the core method behind several other expert systems, with additional examples being decision-making systems, process monitoring, and logistics.

Natural language processing

Symbolic AI was also seriously successful in the field of NLP systems. We can leverage Symbolic AI programs to encapsulate the semantics of a particular language through logical rules, thus helping with language comprehension. This property makes Symbolic AI an exciting contender for chatbot applications. Symbolical linguistic representation is also the secret behind some intelligent voice assistants. These smart assistants leverage Symbolic AI to structure sentences by placing nouns, verbs, and other linguistic properties in their correct place to ensure proper grammatical syntax and semantic execution.

Moreover, Symbolic AI allows the intelligent assistant to make decisions regarding the speech duration and other features, such as intonation when reading the feedback to the user. Modern dialog systems (such as ChatGPT) rely on end-to-end deep learning frameworks and do not depend much on Symbolic AI. Similar logical processing is also utilized in search engines to structure the user’s prompt and the semantic web domain.

Constraint satisfaction

Naturally, Symbolic AI is also still rather useful for constraint satisfaction and logical inferencing applications. The area of constraint satisfaction is mainly interested in developing programs that must satisfy certain conditions (or, as the name implies, constraints). Through logical rules, Symbolic AI systems can efficiently find solutions that meet all the required constraints. Symbolic AI is widely adopted throughout the banking and insurance industries to automate processes such as contract reading. Another recent example of logical inferencing is a system based on the physical activity guidelines provided by the World Health Organization (WHO). The knowledge base of this AI is the guidelines themselves. Since the procedures are explicit representations (already written down and formalized), Symbolic AI is the best tool for the job. The researchers were able to provide the guidelines as logical rules. When given a user profile, the AI can evaluate whether the user adheres to these guidelines.

Explainable AI

Symbolic AI is also highly interpretable. Since the program has logical rules, we can easily trace the conclusion to the root node, precisely understanding the AI’s path. For this reason, Symbolic AI has also been explored multiple times in the exciting field of Explainable Artificial Intelligence (XAI). A paradigm of Symbolic AI, Inductive Logic Programming (ILP), is commonly used to build and generate declarative explanations of a model. This process is also widely used to discover and eliminate physical bias in a machine learning model. For example, ILP was previously used to aid in an automated recruitment task by evaluating candidates’ Curriculum Vitae (CV). Due to its expressive nature, Symbolic AI allowed the developers to trace back the result to ensure that the inferencing model was not influenced by sex, race, or other discriminatory properties.

These limitations of Symbolic AI led to research focused on implementing sub-symbolic models.

Contrasting to Symbolic AI, sub-symbolic systems do not require rules or symbolic representations as inputs. Instead, sub-symbolic programs can learn implicit data representations on their own. Machine learning and deep learning techniques are all examples of sub-symbolic AI models.

Sub-symbolic models can predict some target objectives after extracting patterns from their input. Their training process is more significant than that of the manual symbolic process. With specific techniques, such as NNs, the developer does not even have to process the input data!

Sub-symbolic AI models can be scaled to more significant tasks and datasets effortlessly. Furthermore, sub-symbolic systems learn polytonic relationships, allowing for retraining and updating their previous knowledge. As such, sub-symbolic systems work well with non-stationary datasets. We tabulate the main differences between symbolic and sub-symbolic models as follows:

Table 2.3: A comparison between the symbolic and sub-symbolic paradigms

Comparing both paradigms head to head, one can appreciate sub-symbolic systems’ power and flexibility. Inevitably, the birth of sub-symbolic systems was the primary motivation behind the dethroning of Symbolic AI. Symbolic AI quickly faded away from the spotlight. Funnily enough, its limitations resulted in its inevitable death but are also primarily responsible for its resurrection.

As we got deeper into researching and innovating the sub-symbolic computing area, we were simultaneously digging another hole for ourselves. Yes, sub-symbolic systems gave us ultra-powerful models that dominated and revolutionized every discipline. But as our models continued to grow in complexity, their transparency continued to diminish severely. Today, we are at a point where humans cannot understand the predictions and rationale behind AI. Take self-driving cars, for example. Do we even know what’s going on in the background? Do we understand the decisions behind the countless AI systems throughout the vehicle? Like self-driving cars, many other use cases exist where humans blindly trust the results of some AI algorithm, even though it’s a black box.

Symbolic AI provides numerous benefits, including a highly transparent, traceable, and interpretable reasoning process. So, maybe we are not in a position yet to completely disregard Symbolic AI. Maybe Symbolic AI still has something to offer us. Throughout the rest of this book, we will explore how we can leverage symbolic and sub-symbolic techniques in a hybrid approach to build a robust yet explainable model.

Symbolic AI is one of the earliest forms based on modeling the world around us through explicit symbolic representations. This chapter discussed how and why humans brought about the innovation behind Symbolic AI. The primary motivating principle behind Symbolic AI is enabling machine intelligence. Properly formalizing the concept of intelligence is critical since it sets the tone for what one can and should expect from a machine. As such, this chapter also examined the idea of intelligence and how one might represent knowledge through explicit symbols to enable intelligent systems.

Knowledge representation and formalization are firmly based on the categorization of various types of symbols. Using a simple statement as an example, we discussed the fundamental steps required to develop a symbolic program. An essential step in designing Symbolic AI systems is to capture and translate world knowledge into symbols. We discussed the process and intuition behind formalizing these symbols into logical propositions by declaring relations and logical connectives.

Finally, this chapter also covered how one might exploit a set of defined logical propositions to evaluate other expressions and generate conclusions. This chapter also briefly introduced the topic of Boolean logic and how it relates to Symbolic AI.

We also looked back at the other successes of Symbolic AI, its critical applications, and its prominent use cases. However, Symbolic AI has several limitations, leading to its inevitable pitfall. These limitations and their contributions to the downfall of Symbolic AI were documented and discussed in this chapter. Following that, we briefly introduced the sub-symbolic paradigm and drew some comparisons between the two paradigms. These comparisons serve as a foundation for the rest of the book.

The following chapters will focus on and discuss the sub-symbolic paradigm in greater detail. In the next chapter, we will start by shedding some light on the NN revolution and examine the current situation regarding AI technologies.

You can browse the following material for further reading to complement this chapter:

• Vellino, Andre. (1986). Artificial intelligence: The very idea: J. Haugeland (MIT Press, Cambridge, MA, 1985); 287 Artificial Intelligence. 29. 349–353.
• Cassirer, E. (1953). The philosophy of symbolic forms. New Haven: Yale University Press.
• For further information on Boolean logic and predicate logic, you can use these links:
  • https://computer.howstuffworks.com/boolean.htm
  • https://www.tutorialspoint.com/discrete_mathematics/discrete_mathematics_predicate_logic.htm