- Content Marketing and SEO Transformation
- The Changing Job Market
- A Future-Proof Framework
- Introduction to AI Definitions and Concepts
This chapter is from the book
This chapter is from the book
This chapter is from the book
Introduction to AI Definitions and Concepts
People can mean different things when they talk about AI. So before we jump into discussing how it can be applied to content marketing and SEO, we think it’s essential to define the terms you will use. Although many other sources have a comprehensive glossary of AI terms, we’ve included a few important definitions that we’ll refer to throughout this book. If you’d like to read a bigger glossary, check out the “AI Key Terminology” web page on the U.S. General Services Administration website.8
Artificial intelligence (AI) refers (at a high level) to machines that can simulate human intelligence by using feedback loops of data to “learn” with minimum human intervention. AI is generally used to process volumes of data that are beyond the scale of what a human can analyze. AI-powered tools can help content and SEO marketers with their work.
Algorithms in very simple terms are any processes or sets of rules used to solve a problem or complete a calculation. Although an algorithm may run repeatedly with new information each time, or multiple algorithms may run together, they are finite, meaning they will inevitably reach an endpoint.
Machine learning (ML) is a broad concept that refers to computer systems drawing inferences from data. Machine learning uses one or more algorithms and statistical models to analyze data. Machine learning without natural language processing (NLP) is algorithmic AI. Machine learning with NLP is generative AI (GenAI). GenAI is what most of this book is about, so we cover it and explain NLP in more detail in the “Generative AI Tools, Assistants, and Agents” section in this chapter.
Algorithmic artificial intelligence (AAI) refers to mathematical AI systems that go beyond fixed rules, essentially “talking” to other computers to analyze information and make decisions. AAI is used in scientific and medical applications, such as detecting early signs of breast cancer that may be missed by doctors by comparing a patient’s images against large training datasets of previously diagnosed medical images. Google has been using AAI since 20019 in various aspects of their services.
Generative artificial intelligence (GenAI) is what most of us now think about when we think about AI. It’s a type of deep machine learning that seeks to generate new data (text, images, or other content) based on a set of training data. Chat GPT (text) and DALL-E (images) are popular tools based on GenAI, which is what most marketers or SEOs use when they use AI to help them in their work. We cover this in more detail in the “Generative AI Tools, Assistants, and Agents” section.
Natural language processing (NLP) is a subfield of AI that uses machine learning, and in some cases deep learning, to communicate with users in natural human language. NLP is tricky to understand because it can appear that the AI is “learning.” What the AI is actually doing is taking in new information. This can happen through supervised or unsupervised “learning,” as we discuss in the following sections. Although NLP can be used in any AI implementation, it is most commonly used in GenAI.
Neural networks in humans are the interconnected nodes and neurons in the brain that allow the brain to absorb, assimilate, and process new information. Artificial neural networks are modeled after the human brain but (currently) lack the ability to apply ethics, morality, or other non-defined decision structures to decision-making. Artificial neural networks use multiple nodes in layers to accomplish complex calculations and arrive at a decision. More than three layers of nodes is classified as deep learning.
Deep learning (DL) is a form of machine learning that uses multiple neural networks to simulate the processes that occur in the human brain. The computer can assimilate new information but cannot “learn” the way a human can. Deep learning is what makes it possible for GenAI to produce relevant, high-quality content for SEOs and content marketers.
Large language models (LLMs) are a type of deep learning model that are pre-trained on very large sets of data. LLMs are used for NLP to create text or responses that seem like they were written by a human. ChatGPT is a popular GenAI tool based on LLM technology. If you’ve used it, you may know that the accuracy of its responses is limited based on the data it was trained on and when that training data was last updated.
Supervised learning happens when information is intentionally introduced to an AI system using labeled data, meaning that input data is paired with corresponding output data. Clear instructions are given to train the AI system. However, supervised learning can still be risky if the training dataset accidentally includes personally identifiable information (PII) or the input data is incomplete. This is of particular concern in the health, education, and government uses of AI, as we discuss in Chapter 11, “AI in Regulated Industries.”
Unsupervised learning occurs when the AI system is provided with information that is not clearly labeled with input–output pairs and the AI is expected to make its own conclusions about how to connect the information. This commonly occurs when LLMs are updated with more recent crawls of content. ChatGPT is an excellent example of an AI that “learns” unsupervised. This opens potential legal and ethical issues that we go over in Chapter 6.
Hallucinations occur when an AI delivers an inaccurate or incorrect conclusion. This happens because the data it’s working with is limited, incomplete, or incorrect. This can commonly occur with unsupervised learning. Hallucinations are a nightmare for marketers and brands if AI generates content that misleads consumers, makes harmful claims, or violates regulations. We’ll cover ways to reduce the impact of hallucinations throughout this book.
If you’re already overwhelmed with these glossary terms, that’s OK. The graphic in Figure 1.5 may help. Now that you have an overall understanding of several important concepts in AI, let’s talk in more detail about generative AI, which is particularly relevant for content marketing and SEO.
)
FIGURE 1.5 How AI works, simplified for marketers
Generative AI Tools, Assistants, and Agents
The next evolution of AI is already happening as we write this book. AI agents are the new buzzword. To fully understand AI agents, we need to explain GenAI a little more.
Generative AI (GenAI) Tools
Generative AI is what makes AI tools accessible to the non-technical person. It takes computer language and “translates” it into human language using LLMs. GenAI tools create new content, such as text, images, audio, video, or code, based on an instruction (“prompt”) given to them and an expected output style (text, image, code, and so on). Examples of these that you may already use include ChatGPT, Claude, DALL-E, Sora, and Suno. Most of the information we cover in this book pertains to GenAI tools.
AI Assistants
AI assistants like Apple Siri or Google Assistant are popular tools that use GenAI. They also use personal data about you and your habits to help you complete tasks. You can ask your assistant to tell you your schedule or make a dinner reservation. Here’s an example assistant prompt:
“Hey Siri, make a reservation for dinner at 7 p.m. today at Corner Bistro.”
The AI assistant may use your OpenTable app to make the reservation.
AI Agents
AI agents are the next evolution in AI capabilities because they can use multiple GenAI tools together to perform a task or make a decision. The key difference between an assistant and an agent is that you must tell an assistant what you want. An agent can use other information to determine the best response.
In the dinner reservation example, you can use an AI agent to complete a more esoteric task.
“Hey agent, I want to go to dinner on Friday.”
The agent can use multiple tools or even other agents to autonomously make you a reservation at Corner Bistro and respond with something like
“I made you a reservation at Corner Bistro because it has great reviews and is near your office. The best available time was 7:15 p.m.”
The key is that if you ask an agent a question, it has more than just the defined ways it can respond. For example, if you ask the agent to define a complicated process, it may choose to respond with an image of a flowchart of the process. Agents can also “learn” to start tasks on their own, to work with other agents to complete multi-step processes, and to take proactive measures without being told. Table 1.1 shows a comparison of an AI assistant with an AI agent.
TABLE 1.1 AI Assistants vs. AI Agents
AI ASSISTANTS |
AI AGENTS |
|---|---|
Reactive |
Proactive |
Respond to a specific prompt |
Determine how to achieve a provided goal |
Complete simple tasks |
Make decisions, complete complex processes |
With a diverse range of business applications, it’s no wonder the market for AI agents is expected to skyrocket. The Boston Consulting Group reports that AI agents are becoming more prolific across technology applications, growing with a 45 percent compound annual growth rate over five years.11
You may be familiar with AI agents for social media bots like Sprout Social’s Smart Inbox or business process agents like NVIDIA’s Eureka or Microsoft’s Copilot Studio.
What’s truly remarkable is that tools like these can autonomously interact with users, make real-time decisions, and improve their responses to accomplish tasks or goals more effectively. However, because AI tools and agents rely on other datasets rather than on novel ideas, they can still make costly mistakes for a brand, which we’ll explore later in this book.
