Chapter 1

Grasping the Basics of Generative AI

IN THIS CHAPTER

Learning about the different versions of AI

Considering the interaction between AI and humans

Discovering how AI understands prompts

Can you imagine a world where machines can learn, create, and think like humans? This is the realm of generative AI (GenAI), where technology and creativity come together. Some types of AI learn from experience, while others follow strict rules.

In this chapter, we look at AI systems that need guidance, like students in a class and those who learn on their own. We also discuss AI that makes entirely new content instead of just organizing data. This chapter explores the diverse world of AI.

Understanding the Different Flavors of AI

Each kind of AI has its own special function and way of working, just like tools in a toolbox. In the following sections, we look at these different types of AI to understand what they're like and how they work. We start with two main types:

• AI that learns from data, which we call machine learning (ML)
• AI that follows specific rules

Both types of AI have their own strengths, making them suitable for different kinds of tasks. Understanding this will help you get a clear picture of how AI is changing our world, from health care to manufacturing and beyond. Each type of AI brings something valuable to the table, showing just how diverse and useful these technologies can be.

Using AI that learns from data

ML can acquire knowledge and get smarter over time. It works by training on large amounts of data, finding patterns in it, and then making decisions based on what it finds.

This kind of AI is always changing. It gets better as it gets more data to learn from. For example, think about a system that recommends music. It looks at the songs you liked before and what other people who like the same music as you do also enjoy. Then it suggests new songs for you.

Another common area where ML excels is facial recognition. By reviewing many photos of a person's face, PXL Ident (www.pxl-vision.com/en/pxl-ident) can learn to recognize new photos of that person. Figure 1-1 shows an example of this application.

FIGURE 1-1: PXL Ident performs facial recognition, a common type of ML.

The ability to learn and change makes ML very powerful and useful. It can perform tasks like creating personal recommendations, organizing your phone's photo albums, or helping self-driving cars make decisions.

We can further break down ML into two specific types. These types differ in the way we teach AI:

• Supervised learning: The AI learns from data that already has answers. It's like giving it a quiz with an answer key. For example, when AI works on recognizing images, it gets tons of pictures that are already named, like cat photos labeled "cat." This way, the AI learns to pick out similar images on its own.
• Unsupervised learning: In this type of ML, the AI doesn't get any answers up front. It looks at the data, like customer buying patterns, and tries to make sense of it by itself. It's like solving a puzzle without the picture on the box as a guide. In business, this type of AI helps figure out which customers may like certain products, even though no one has sorted these customers into groups before.

ML is great because it can learn and change. It's like a quick learner that gets better the more it practices. This makes it perfect for jobs where things keep evolving or need a personal touch. For example, in health care, ML helps with diagnosing diseases. It looks at medical images, like X-rays or magnetic resonance imaging (MRI) scans, and learns from many examples. Over time, it gets very good at spotting signs of different health conditions.

Using follow-the-rules AI

Follow-the-rules AI doesn't learn from data. Instead, it follows a set of instructions we give it. This means that it doesn't change or get better over time. It's useful for tasks that are done the same way every time. This kind of AI is reliable for critical jobs where mistakes could be dangerous. Imagine a nuclear power plant. Here, rule-based AI helps monitor everything, making sure all systems are working correctly. It does the same thing every time, which is really important for safety. In a factory, rule-based AI checks products for any defects. It uses specific guidelines to examine each item, making sure everything meets the standard. This keeps the quality of the products consistent, which is super important for the business and the customers.

A good example of follow-the-rules AI is email spam filters. The filters have a set of rules, such as looking for certain words, to decide if an email is spam. This method is straightforward and always follows the same steps. It is great for jobs that require consistency and follow specific rules or guidelines.

Follow-the-rules AI is the go-to for tasks that require steady and unchanging performance.

Needing a Teacher versus Learning On Its Own

How AI learns is really important. However, not all AI learns the same way. There are two types of AI learning:

• Supervised learning: Supervised learning needs guidance, which is kind of like having a teacher. It learns from examples that already have answers.
• Unsupervised learning: With unsupervised learning, AI figures things out on its own. It doesn't have answers up front - it has to sort through data by itself.

Knowing the difference between these learning styles helps you understand AI better. It shows you how AI can either follow a set path or discover new things, depending on how it's taught.

Considering supervised learning

Supervised learning in AI works something like having a teacher. This kind of AI gets data that is already labeled or has clear definitions. Think of this data as a textbook with all the answers. The AI learns from this "textbook" to understand patterns and make choices about new, similar information.

For example, in medical diagnosis, supervised learning is highly useful. AI systems get trained with many medical images, like X-rays or MRI scans, that doctors have already diagnosed. The AI studies these images and learns how to spot various health conditions. Then, when it sees new patient images, it can suggest what the diagnosis may be. This helps doctors diagnose more quickly and accurately.

In the world of finance, banks use supervised learning, too. They train AI on data about transactions, some of which are marked as fraudulent and others of which are marked as safe. When the AI checks new transactions, it looks for signs that match known fraud. If it spots something suspicious, it alerts the bank. This way, the AI helps stop fraud before it causes any harm.

In both these cases, the AI relies on its training from labeled data to make smart decisions. It's a bit like a student who has studied a lot and then applies that knowledge to new problems. This kind of AI is great for tasks where you need reliable and accurate results based on clear examples it has learned from.

Dipping into unsupervised learning

With unsupervised learning, AI systems learn from data that does not have clear instructions or labels. Imagine AI as an explorer going through data without a map. It looks for patterns and figures out the structure of the data all by itself. The goal is not just to find the correct answer but to explore and uncover how the data is organized.

One area where unsupervised learning is highly useful is in retail market segmentation. In this case, AI examines customer data, like what they bought, their preferences, and where they're from. However, it doesn't have predefined groups. The AI figures out its own ways to group customers based on the data. This helps businesses understand their customers better and create marketing strategies for different groups. It's a smart way to increase customer happiness and boost sales because the offerings are more tailored to each group.

Unsupervised learning is also important on social media platforms. The algorithms look at what users do - for example, the posts they like or share - to spot trends and common themes. Using this info, the AI can adjust what each person sees in their feed, making sure it shows posts they're more likely to find interesting. This makes the social media experience better for users because they get content that is more relevant to them. In both retail and social media, unsupervised learning helps AI understand and respond to people's preferences in a more personalized way.

Recognizing differences and their impact

The main difference between supervised and unsupervised learning in AI is about whether the data has labels. Supervised learning has a clear structure. It uses data where the outcomes are already known. Think of it like having a guidebook. It's great for specific tasks like sorting things into categories or making predictions.

Unsupervised learning, on the other hand, is more like an adventure into the unknown. It works with data that doesn't have labels. The AI has to figure out the patterns and structures in this data by itself. It's kind of like exploring a new place without a map. This...