Before we talk about types, let us talk about fear
A lot of people come to AI with anxiety already built in. They have watched films where AI becomes self-aware, decides humans are a threat and takes over. They have read headlines about AI taking every job, ending privacy or becoming impossible to control.
Some of those concerns deserve serious discussion. But a lot of the fear comes from not knowing which kind of AI is actually being talked about. There is a massive difference between the AI that recommends your next video on YouTube and the hypothetical AI that could genuinely think for itself. They share a name but almost nothing else.
This lesson exists to make that distinction clear. Once you know the three types, you will be able to hear any claim about AI, whether in a news article, a boardroom or a conversation, and immediately understand what level of intelligence is actually being described.
The three types in full
Also called Weak AI, though that name is misleading. Narrow AI is not weak at all. It is extremely powerful at the one specific thing it was built to do. The word "narrow" means it operates within a defined boundary. Ask it to do something outside that boundary and it is completely lost.
Every single AI product you have ever used is Narrow AI. ChatGPT is narrow: it processes and generates text. Face ID is narrow: it recognises your face. AlphaGo is narrow: it plays Go. None of them can cross over into the other's territory. ChatGPT cannot unlock your phone. Face ID cannot write an essay.
Also called Artificial General Intelligence or AGI. This is a system that can perform any intellectual task a human being can do. Not just one task. Any task. It could write software in the morning, diagnose a medical condition in the afternoon and compose music in the evening, all without being specifically trained for each.
AGI does not exist today. It is the subject of serious research and serious disagreement. Some researchers think we are decades away. Some think we are closer. Some think the current approach of scaling up language models will never get us there. Nobody knows. What is certain is that nothing available today qualifies.
Also called Artificial Superintelligence or ASI. This is a system that surpasses human intelligence in every dimension: creative thinking, problem solving, emotional understanding, scientific reasoning, everything. By definition it would be better at improving itself than humans would be at improving it.
This is the version of AI that appears in films and in philosophy seminars. It is a legitimate topic for long-term thinking about the future, but it exists nowhere outside of theoretical discussion. Treating it as a present-day concern is like worrying about what happens when we colonise Mars before we have sent anyone back to the Moon.
Why narrow AI is more impressive than it sounds
The word "narrow" can make Narrow AI sound like a consolation prize. It is not. The systems we have built within that narrow category are genuinely astonishing by any historical measure.
Twenty years ago, the idea that a computer could hold a fluent conversation in any language, generate a photorealistic image from a text description, diagnose cancer from a scan more accurately than most radiologists, or beat any human alive at chess, Go and most video games would have sounded like science fiction. All of that is Narrow AI. All of it exists today.
The question is not whether machines can think. The question is whether what they do is useful.
Paraphrased from decades of AI debateNarrow AI systems do not have goals of their own. They do not want anything. They do not get bored, frustrated or ambitious. A spam filter does not decide one day that it wants to read your private emails. It processes what it is given, produces an output and stops. Understanding this distinction cuts through most of the fear that surrounds AI in the media.
The legitimate concerns about AI, and there are real ones, tend to be about who controls these systems and for what purpose. Not about the systems themselves developing sinister intentions.
Separating real concerns from noise
Now that you understand the three types, you can apply a simple filter to every claim you hear about AI.
When someone says AI is going to take your job, they are talking about Narrow AI. That is a real economic discussion worth having. When someone says AI is going to decide to eliminate humanity, they are jumping several decades ahead to a hypothetical version of Superintelligence that does not exist and may never exist in the form they are imagining. Those are completely different conversations.
Where we actually are today
We are firmly in the era of Narrow AI, and it is an extraordinary time to be paying attention. The tools available now would have seemed miraculous ten years ago. Language models that can write, reason and converse. Vision systems that can see better than human eyes in specific tasks. Drug discovery tools that are accelerating medical research. Climate models running faster and more accurately than ever.
None of this requires AGI. None of it requires superintelligence. Narrow AI, applied thoughtfully and built responsibly, is already reshaping industries, creating new jobs and eliminating others, and raising genuinely important questions about how society should be organised.
Think of the different types of AI the way you might think about different stages of a vehicle. A bicycle is real, useful and already changing how people get around. A flying car is theoretically possible and researchers are working on it. A teleporter remains firmly in science fiction. The existence of teleporters as a concept does not make bicycles less remarkable or less worth understanding.
As you move through this course, every technique you learn, every model you build and every tool you use will be Narrow AI. That is not a limitation. That is where the real work, the real opportunity and the real responsibility lives right now.