Artificial intelligence is a large umbrella of all kinds of different smart learning systems and we're going to talk later on in a different guide on how machine learning and data science specifically fit inside of the AI ecosystem. But this guide is simply going to be an overview on what AI is and how it's used in the real world.

The AI definition from tech target is the AI or artificial intelligence is the simulation of human intelligence processes by machines especially computer systems. And I think that is a great way to describe what artificial intelligence is. We are essentially trying to mimic human intelligence while also taking advantage of the power and robustness of modern computer systems.

We have a great real-world analogy on how this operates imagine that you have a baby or a toddler. You can teach that toddler how to start recognizing different things you can show it pictures of dogs and cats and it will eventually be able to pick out and tell you which one is the right one. The human brain was made to do that we have this cognitive ability to learn.

Now the issue is that we are limited in bandwidth. So you take that same toddler that turns into a kid turns into an adult no matter how good they get at analyzing that dog and that cat on the screen. They're not going to be able to do that for a million image searches on Google.

That's where machine learning and artificial intelligence comes in is what we're attempting to do with AI is mimic human intelligence so we can make it scalable. The world of AI has come down into a couple different categories and part of what I'm going to be discussing is going to be practical and part of it's going to be theoretical just due to the nature of the current state of AI. So there are two types in AI, we have strong and weak AI weak.

AI is all we have right now and what weak means it doesn't mean that it's bad it simply means that it's limited. It means that we're still having to some degree. We're having to build these agents ourselves and they are limited in certain ways they may be incredibly powerful such as the Google search engine AI is able to learn and adapt and it is an incredibly powerful tool. However strong AI is a theoretical concept that's propose that will happen at some point in the future and that is where a system can learn anything.

So a strong AI would theoretically be a system that could become a search engine. It could go give product recommendations it could do pretty much anything that you can imagine. And because of the robustness of the computer and just the power behind that it would be able to adapt without us as humans and as programmers telling it what to do.

Now strong and weak AI are the two broad categories but we actually have an even more detailed categorization of artificial intelligence. We have four different types and the first two are available and it's what we're going to go through in this course and throughout the entire machine learning track. We're going to be covering type 1 and Type 2

Now, type 3 and 4 have not been invented yet. They're simply theoretical concepts that people are trying to work towards. But right now they are not reality. The first type of artificial intelligence it's just called Type 1 is a reactive machine. So what a reactive machine does is it can take a list of stored procedures and then you can have it run through simulations and then it will react to the circumstance given its programming. A great example of reactive programming is the Watson computer the supercomputer that has artificial intelligence from IBM and it can play chess and it can play chess very well.

However, it's reactive which means that the developers at IBM put Watson together and it pre-programmed all kinds of different probabilities. And so what it can do is when it sees a movement on the chessboard. It runs through a simulation and it runs through this full list of probabilities and then it says OK I'm going to move my knight to this position because I think this is going to give me the best chance to win.

The difference between type 1 and Type 2 is type 1 the reactive type of AI doesn't have any kind of memory so therefore it doesn't really get into machine learning because it doesn't learn. It was programmed with a set of probabilities and then it runs through those simulations to give an output.

And that leads us perfectly into what type 2 is, type 2 is called limited memory algorithms and so in AI right now the majority of algorithms that we're going to be working with and that fit in the realm of machine learning are these limited memory types of systems. Now this can range anywhere from a self driving car all the way through the Google search engine recommendation system. So this is going to be really the sweet spot of artificial intelligence right now because these systems are able to learn about the world and then adapt based on what they learn.

Now the reason why this is called the limited memory type is because these systems are still limited, they're limited in what they can learn. Because even though a self-driving car may be very powerful and it may be able to take in road signs and traffic lights and be integrated with GPS and mapping systems there still is a ceiling with what it's able to do and so because of that it is limited. But that is as far as artificial intelligence has gone up to this point.

Moving on to Type 3. And once again type 3 has not been invented yet but it is a theoretical idea. And that is the theory of mind and what this means is that that AI agent would be able to understand that other beings had intentions. So extending our self-driving car example it would be as if your car would be able to start thinking about the intentions of the other drivers on the road and would adjust dynamically.

Now right now self-driving cars in a sense have the ability to look at what other drivers are doing but the system doesn't really process the idea that there are humans who have thought processes are simply reacting to how they are programmed in order to stay safe on the road.

The fourth and final type is self-awareness. Now, this is by far the most sci-fi version of AI but this is the theoretical idea that a machine would become completely self-aware. This means it would be able to have its own thoughts its own intentions. You would be able to program it but then it would go live it's completely own life just like a normal human being.

This is a concept that has been hotly debated since 1956 when the term artificial intelligence came into being. And so this is something that we are most likely pretty far away from if it is ever even possible but when it is that would be a complete change for the entire world and would most likely be a new stage in the step in evolution.

So those are the four types of artificial intelligence. And that is a very broad realm of technologies that all fit inside of AI in the next guide we're going to walk through how machine learning and data science fit inside of the AI ecosystem.