What Is AI and Why Does It Need Rules?

What Is AI and Why Does It Need Rules?

Imagine walking into a classroom on the first day of school. The room is filled with students from different backgrounds, each bringing their own experiences, questions, and ways of learning. Some students excel at math but struggle with reading. Others are creative problem-solvers but need help staying organized. Now imagine that one of these students isn't human at all—but a robot. This robot student can read thousands of books in seconds, solve complex equations instantly, and even write poems. Yet despite these incredible abilities, this robot student still needs a teacher. It still needs rules. It still needs someone to guide it toward being helpful rather than harmful.

That robot student is artificial intelligence—AI for short.

But here's the mystery that puzzles many people: if AI is so smart, why does it need rules? Why can't it just figure things out on its own, the way we hope to as we grow older? The answer lies in understanding what AI truly is, how it thinks (or rather, how it processes information), and why even the most brilliant minds need boundaries to ensure they're working in everyone's best interest.

This chapter takes you on the first leg of an adventure into the world of AI governance—the system of rules and guidelines that keep artificial intelligence safe, fair, and helpful for everyone. By the time you finish reading, you'll understand not only what AI is but also why governance isn't a restriction on AI's abilities. Rather, governance is what makes those abilities valuable to humanity.

The Nature of Intelligence: What Makes Something "Artificial"?

Before we can understand AI governance, we need to understand what we're governing. The term "artificial intelligence" sounds like a contradiction—how can a machine be intelligent? To answer this question, we need to think carefully about what intelligence actually means.

Human intelligence is remarkable. When you see a cat run across your living room floor, your brain doesn't just register "four-legged creature moving quickly." Your brain draws on years of experience: you know cats have soft fur, make distinctive sounds, and behave in certain predictable ways. You can predict that the cat might knock over your water glass because you've seen cats do this before. You can even imagine how the cat feels—perhaps startled or playful. This ability to take raw information, combine it with learned experience, and generate useful predictions and responses is what we call intelligence.

AI attempts to replicate certain aspects of this process. However, here's the crucial difference: human intelligence emerged naturally over millions of years of evolution. It developed because it helped our ancestors survive. AI, on the other hand, is constructed by humans—it doesn't emerge from life and survival. It has to be deliberately designed, trained, and continuously monitored to ensure it behaves appropriately.

How AI Systems Actually Work

Modern AI systems work through a process called machine learning. Rather than being explicitly programmed with step-by-step instructions (like "if you see a cat, then do this"), machine learning systems learn patterns from examples. Imagine you wanted to teach a child to recognize cats. You wouldn't say "a cat has four legs, pointed ears, whiskers, and fur." Instead, you'd show the child many pictures of cats and say "this is a cat, this is a cat, this is not a cat." Over time, the child learns to identify cats by picking up on the subtle patterns that distinguish them from other animals.

Machine learning works similarly. An AI system called a neural network—which gets its name because it's inspired by how biological brains work—is fed millions of images labeled as "cat" or "not cat." The system adjusts its internal parameters to minimize the difference between its guesses and the correct answers. After processing millions of examples, the network becomes remarkably good at recognizing cats, even in images it has never seen before.

This process of learning from examples is called training. The data used during training—these millions of labeled images—forms what researchers call a training dataset. The patterns the AI learns are stored in numerical values called weights and biases. These numbers, multiplied and combined through countless layers of mathematical operations, determine how the AI responds to new inputs.

The complexity here is staggering. A state-of-the-art language model might have over 100 billion parameters—individual numerical values that were adjusted during training. When you ask such a model a question, your text is converted into numbers, processed through these billions of parameters, and converted back into words. The model has learned to produce responses that sound like human language because it has seen billions of examples of human text during training.

The Fundamental Difference: Narrow vs. General Intelligence

Not all AI is created equal. The AI that recognizes cats in images is an example of narrow AI (sometimes called weak AI)—it can do one specific task extremely well but cannot apply that knowledge to other domains. Ask a cat-recognition AI to translate languages, and it will fail completely. This AI has no understanding of what a cat actually is, no consciousness, no desires or fears. It simply learned statistical patterns in image data.

In contrast, general AI (sometimes called strong AI or artificial general intelligence, AGI) would be a system capable of understanding, learning, and applying knowledge across any domain the way humans do. Such a system would be able to recognize cats, translate languages, play games, understand jokes, and much more—all using the same underlying intelligence. As of this writing, general AI does not exist, though it remains an active area of research and speculation.

This distinction matters enormously for governance. Narrow AI systems are designed for specific purposes. A recommendation algorithm learns to suggest videos you might like. A medical imaging AI learns to identify tumors in X-rays. Each system has a clear domain, and governance can focus on ensuring the system behaves well within that domain. General AI, if it ever becomes reality, would require a fundamentally different governance approach since it could potentially be applied to any task.

For the purposes of this book, when we talk about AI governance, we're primarily concerned with narrow AI systems—the specific tools and technologies that are already woven into the fabric of our daily lives. But understanding that these systems are not truly "intelligent" in the human sense is essential. They're powerful pattern-matching tools that can be incredibly useful or surprisingly dangerous, depending on how they're designed, trained, and governed.

AI in Your Daily Life: Recognizing the Invisible Assistant

Now that we understand what AI is, let's look at where it hides in your everyday life. You might be surprised to learn just how often you interact with artificial intelligence—often without realizing it.

The Voice in the Box

When you ask a smart speaker like "Hey Assistant" or "Hey Alexa" to play your favorite song, an AI system is working behind the scenes. First, the system uses a specialized AI called automatic speech recognition to convert your spoken words into text. This is harder than it sounds—people speak in different accents, mumble, or talk while music plays in the background. The AI has learned to handle these variations by training on millions of hours of recorded speech.

Next, natural language understanding parses your request to figure out what you actually want. "Play my favorite song" might mean different things to different people. The AI needs to understand that "favorite" refers to your personal preferences, stored somewhere in its memory, and that "song" means a musical track rather than a rington