These are the types of AI you need to know in 2024

As children we thought of AI only as far as robots and science fiction. But now it's constantly in the news and being talked about as if it is one big thing. But it's not. There's a lot of different types of AI and it's important for anyone (let alone anyone in a tech job) to at least know what those types are and why the are different. So let's break down the ten main AI types you should definitely know about in 2024.

1. Machine Learning (ML): The Pattern Detective

When we talk about "Artificial Intelligence" lately we're mostly talking about some form of Machine Learning. As IBM say "machine learning is a branch of artificial intelligence and computer science which focuses on the use of data and algorithms to imitate the way that humans learn, gradually improving its accuracy."
The big point here is that machine learning systems get better over time. And this is powerful because we don't have to keep reprogramming them. I'm sure you've seen examples of Machine Learning systems being fed a ton of data (like pictures of cats and dogs). It figures out the patterns that mean "cat" vs. "dog" and gets super good at spotting them in new pictures.

2. Deep Learning (DL): Think Brain-Inspired Computing

Deep Learning is like Machine Lerning on steroids. It is a subset of machine learning that uses artificial neural networks with multiple layers to learn complex patterns and representations directly from data. If that sounds a little complicated, just think of these networks as something that is inspired by the structure and function of the human brain.
The key think here is that there are layers that the data goes through. Some layers handle certain classifications, some make more abstract and higher-level extractions, and the final layers are those that make the kind of output we have designed. This is how machines master games like Go, recognize your face (even for things like dating!), and understand what you're saying.

3. Natural Language Processing (NLP): Your AI Chat Buddy

Imagine teaching a computer to talk like a human? This sounded crazy even a decade ago (sorry Siri), but this is where we are at right now. Natural Language Processing is what powers the Large Language Models (LLMs) you keep hearing about in the news. This methodology helps machines make assumptions about what we write and say - powering those friendly chatbots, language translation apps, and even tools that figure out if customers are happy or grumpy from their reviews. Rather than just raw data, they are designed to analyse what we squishy humans say and write.

4. Generative Adversarial Networks (GANs): The Imagination Battle

You might not have heard too much about GANs recently, but they were all the rage a few years ago before LLMs took the spotlight. Think of GANs like an AI art duel! In this case, we can imagine two brain-like systems competing. One gets REALLY good at making stuff (fake images, music, you name it), the other gets REALLY good at telling real from fake.
The idea of making these networks battle each other (that's the adversarial bit) is that this effort helps create a continually improving system. While LLMs are all the rage right now, expect to hear a lot more about GANs for a while yet.

5. Computer Vision: Robot eyes in the Sky

Computer vision gives machines the power to see! Or at least that's the hope. The most noise about computer vision is coming from a certain car company that has been promising self-driving cars for a very, very, very long time. But there's also robots in factories and even those adorable ninja robots that are absolutely going to be haunting my nightmares tonight.
While it's probably not the smartest idea to build your entire car company around "computer vision is the only thing we need to work out roads and obstacles" for self driving cars, computer vision is essential in other systems like facial recognition and even medical imaging.

6. Expert Systems: Like a Rulebook for Problem-Solving

Let's get into some more focused use cases now. Expert Systems are a branch of artificial intelligence that focus on emulating the decision-making abilities of human experts within a specific domain. We can think of them as working off a big knowledge base and sets of rules.
The other side of this method is to use an "inference engine" to apply this knowledge. The inference engine is a component that processes the rules and facts against a user's query to provide conclusions or make some recommendations. Need to diagnose a weird medical issue? Ask an expert system for clues. They're great at recommending and planning too.

7. Robotics: AI + Moving Parts = Awesome

Here's where we get closer to either the T-800 or Johnny Five. Whatever our future fate looks like at the hands of our robot overlords, those robots get WAY smarter with AI inside. Think of Computer Vision for navigating, Machine Learning for learning cool new moves, Expert Systems for determining rules (like whether to hug or crush a human) and all of it working together - that's how they clean our houses and maybe someday will build stuff alongside us! Or destroy us! Yay!

8. Rule-Based Systems: Oldie but Goodie

Okay, they may seem simpler than fancy Machine Learning, but these rule-based babies are still around! They're those straightforward "if this, then that" decision-makers - useful when every step needs to be super clear. That also means that they are cheaper, and ideally are designed for very specific scenarios where variation isn't likely. Think of things like sorting machines, or… hey stop yawning. They can be exciting despite all the attention going to the killer robots.

9. Fuzzy Logic: Dealing with the 'Gray Areas'

Life ain't always "yes or no" or "black or white". Sometimes we need to be able to consider the various shades of gret. This is where Fuzzy Logic comes in. Think of weather forecasts saying it's going to be "kind of warm", or those smart washing machines figuring out just the right cycle for your messy clothes.

This is an interesting part of Artificial Intelligence because of the way the process works. The first step is to "fuzzifying" the inputs. This means taking crisp values like a temperature reading and mapping them to fuzzy sets representing linguistic terms. So things like "cold," "warm," and "hot." These sets of definitions have overlapping rules, but there are still sets of rules to apply to them, and an engine works out what to do. The fun part is that there's a process that "defuzzifies" back into a crisp value, like a specific fan speed setting.

10. Hybrid AI: The Ultimate Team-Up

The coolest stuff happens when different AI systems combine forces. While there's so much hype about LLMs and the various dramaas happening at LLM companies, the reality is that all forms of AI are evolving and crossing over in continuously changing ways. It's hard to keep up with but it helps to understand the basic types defined here to be able to look past whatever headline of the day, and think about what's really involved.
For example, a medical diagnosis tool might use expert systems for basic stuff, NLP to understand a patient's notes, AND machine learning to spot weird patterns in test results.