What is AI?

📄 What is AI?

🎯 Learning objectives

• Understand what AI really is (and what it isn’t)
• Know the difference between narrow AI and AGI
• Know why the AI explosion is happening right now

When most people think “AI” they might picture Terminator, HAL 9000, or C-3PO – thinking machines with consciousness and their own will. But that’s science fiction. The AI we have today is something entirely different – and at the same time much more practical and powerful than many people realize.

AI is a term used broadly, but what do we actually mean? Here we clarify the definition and the most important difference from how humans think.

What is AI, really?

Artificial intelligence (AI) is systems that perform tasks that traditionally required human intelligence – such as recognizing faces, understanding speech, making decisions, or solving problems.

But here’s the key: AI mimics intelligent behavior, but does not think like a human.

Think of a calculator. It can solve mathematical problems that would take you longer to work out. But the calculator doesn’t “understand” math – it just follows instructions extremely fast. AI works in a similar way, but for much more complex tasks.

AI isn’t new – but something has changed dramatically in recent years. Three interacting factors explain why the AI explosion is happening right now and not 30 years ago.

Brief history

AI isn’t new. As far back as the 1950s, researchers were dreaming of thinking machines. But three things over the past 10–15 years have made the AI explosion possible:

1. Data – The internet has created unimaginable amounts of data (text, images, video, audio). AI systems learn from examples, and now there are millions of times more examples than ever before.

2. Computing power – Modern processors and specialized hardware (GPUs) can perform billions of calculations per second. What would have taken years in the 1990s now takes minutes.

3. Algorithms – Researchers have developed smarter methods for training AI, especially something called deep neural networks.

The combination of these three factors has transformed AI from a theoretical curiosity into a practical revolution.

There’s an important distinction between the AI we have today and the AI often portrayed in film and literature. Understanding the difference is essential for having realistic expectations.

Narrow AI (ANI) vs AGI – what's the difference?

Narrow AI (Artificial Narrow Intelligence, ANI)

This is all the AI in use today. “Narrow” means the system is an expert in one specific task or a narrow domain:

• An AI that recognizes faces cannot drive a car
• A chess AI cannot translate languages without separate training
• An AI that generates text cannot (without specific training) analyze X-rays

Every system is specialized – extremely good at its own thing but completely helpless outside its domain.

Artificial General Intelligence (AGI)

This does not exist yet. AGI would be an AI that can:

• Learn any task (like a human)
• Understand context across different domains
• Reason abstractly and generalize knowledge
• Apply experience from one area to solve problems in another

Current state: We have extremely impressive narrow AI. AGI is still a research question – no one knows whether we’re 5, 15, or 30 years away from achieving it.

Narrow AI can do very different things. By dividing systems into three categories we get a clear picture of what AI actually does with information – and where you encounter each type.

Three main types of AI systems

Analytical AI – analyzes and categorizes

Looks at existing data and makes decisions or predictions.

Examples: Gmail’s spam filter, Netflix recommendations, medical diagnostic tools, credit scoring systems.

Generative AI – creates new content

Creates something that didn’t exist before: text, images, audio, video, or code.

Examples: ChatGPT, DALL-E and Midjourney, GitHub Copilot, music-generating AI.

Robotics and automation – acts in the physical world

Connects AI to physical systems that can move and affect the environment.

Examples: Self-driving cars, industrial robots, delivery drones, robotic vacuum cleaners.

Types are often combined – for example, a self-driving car uses analytical AI for perception and robotics for steering.

To understand how AI works we need to start with how traditional computer programs work and why machine learning is fundamentally different.

How does AI learn?

Traditional programming and machine learning

Traditional programming – a programmer writes exact rules: if the temperature is above 25 degrees show “It’s warm”, between 15 and 25 show “pleasant”, below 15 “cold”. Every scenario must be defined in advance. This works great when the rules are clear.

But some tasks can’t be ruled into existence – recognizing a cat in a photo, deciding whether a text is positive or negative (irony!), or interpreting emotion in a voice. There’s no simple rule that always works.

Machine learning – instead of explicit rules, we give the system thousands or millions of examples and let it find the patterns itself.

Example: cats in photos

• Traditional approach (practically impossible): list rules for whiskers, ears, “cuteness”…
• Machine learning: label hundreds of thousands of images as “cat” / “not cat”, train the model, test on new images.

No human wrote all the rules – the model discovered patterns from data.

Bias – when AI learns our prejudices

AI is often trained on human data. Historical inequalities and stereotypes can then be amplified in the model’s behavior if not measured and addressed. Keep that in mind when interpreting results.

Summary

• AI mimics intelligent behavior through computation – it does not think like a human.
• All modern AI is narrow AI (ANI). AGI does not yet exist as a practical product.
• Data + hardware + algorithms drive the pace of development we’re seeing now.
• AI can be analytical, generative, or connected to robotics.
• Machine learning learns from examples rather than handwritten rules.

Test your knowledge

3 questions · 100% correct to pass · Review your answers when done

Start quiz →