The Art of Artificial Intelligence: For Dummies

Original article was published by Fatimah Hussain on Artificial Intelligence on Medium


The Art of Artificial Intelligence: For Dummies

Clearing up AI from the basics: Supervised and Unsupervised learning

Robots. That’s probably the first thing that comes up in your mind when you think of AI. And for thousands of others. Or maybe not. Maybe the answer is more definitive, like computer, coding, somewhere along those lines. But what on Earth is Artificial Intelligence? What are some advancements of AI? What types of AI are there? How is it used? So many questions. Luckily for you, I got the answers.

Artificial Intelligence? What is THAT?

We hear these 20 syllable words like Artificial Intelligence being thrown around everywhere. But, what is AI? In simple terms, it’s the science of making machines more intelligent. The purpose of AI was to replicate human behavior and their learning process. But what’s the point of that? With this, machines learn to mimic human behavior to execute tasks more accurately and better. We humans will never be perfect. But, machines and algorithms continue to be more and more perfect.

blue: homosapiens; red: AI

Examples

From the first AI chatbot Eliza to the game that beat the world’s smartest player, AI is manifesting a huge social impact! As shocking as it sounds, AI is slowly penetrating in our daily routines. It’s used in several research areas such as healthcare and in autonomous vehicles. AI is even programmed in voice assistants like Alexa and Siri, youtube search recommendations, and even in sports analytics! Crazy, I know. It’s definitely not magic, so what is it?

How Does AI Mimic Humans?

Over the years, AI is able handle vast amounts of data in a way humans cannot. To do that, AI uses decision making, problem solving, and critical thinking. Using strategic analyzing and patterns, AI can give reports and insights for tasks. It is important to note the fact that AI’s main source of information is DATA, TRAINING, PRIOR EXPERIENCE/KNOWLEDGE.

Using complex algorithms and features that AI is equipped with, humans send data off to the machine. From there, the machine learns from data (the input), uses a model to form an understanding and reach a conclusion (output). This is called the training process.

In the beginning, the AI machine will suck and will be inaccurate. But over time it makes connections and becomes more intelligent. This is because of the training data and the changes in the algorithm/model. There are two main types of learning processes that AI uses: supervised learning and unsupervised learning.

Supervised Learning: The Basics

Oh great, another fancy word. Not to worry, you will be an expert in this in no time. is considered task-driven. Supervised learning essentially learns from labeled data, trains the data using a supervised AI model, gives an output, and gets provided feedback. It’s called supervised learning because an expert moderates the AI model. In addition, the algorithm gets provided an answer key that the model can use to evaluate its accuracy for training data. Here’s a visualizer:

Unsupervised Learning: The Basics

You’re half-way done! Let’s move on to unsupervised learning. As the name hints, unsupervised learning is not supervised by an expert. It’s data-driven because it works with unlabeled data. This means unsupervised learning is solely controlled by the unstructured data and the way its formatted.

To do this, unsupervised learning uses data mining. Data mining is the process of discovering hidden patterns and structures in the data (input). Additionally, it finds similarities and differences from the data, and the machine uses its insights and extracts features to calculate the ideal output. And by ideal output, I mean what the machine thinks is the ideal output. The output eventually becomes more and more accurate over time.

The Main Difference Between Supervised and Unsupervised Learning

Now you know what supervised and unsupervised learning is. They seem quite similar, so what are the key differences between the two?

organizes labelled data into respective groups
  • Supervised learning knows its output. It knows its end-goal, and using the expert’s feedback, it tweaks itself in a way that it slowly becomes more and more accurate to the ideal answer
  • Supervised learning has a training data set. Training data is what the AI model trains on before it’s given testing data.
organizes unlabelled data by clustering

Unsupervised learning doesn’t know its output. It has to decide its pathway and its way of learning on its own. Computer scientists rely on unsupervised learning for answers, since most of the time the ‘expert’ doesn’t even know the answer. Eventually, it provides an accurate assumption about the data.

  • Unsupervised learning doesn’t have a training data set. Essentially, the AI goes into the problem blind, relying on data exploration and data mining to reach conclusions.

That begs the question: when do we use supervised and unsupervised learning?

When to Use Supervised Learning

Supervised learning is all labeled data, so it makes sense to use supervised learning where data is labelled. Here are some examples:

  • Classification: In other words, discrete values. An example of a classification method is checking whether tomorrow’s weather at 8 pm will be “hot” or “cold”.
  • Regression: Regression is used to predict continuous values, like temperature at 8 pm. There are many types of regression, like linear regression depending on what type of data you’re working with. That’s an article for another time.

Essentially, the goal is to make predictions.

When to Use Unsupervised Learning

Unsupervised learning starts with raw unlabeled data, so it makes sense to use unsupervised learning where data is unlabelled. Here are some examples:

  • Clustering: This is the main form of use in unsupervised learning. Some examples of clustering is categorization, dividing by similarity, etc. For instance, you could use clustering to find the targeted audience for advertising specific ads.

Essentially, the goal is to notice pattern/structure recognition.

Key Takeaways

There’s a lot of information to unpack here. This is why I have manifested some quick takeaways:

Here’s a chart for easy understanding. You’re welcome.

Hope you enjoyed reading! Check out my LinkedIn, Instagram, and follow me on Medium. I would greatly appreciate it if you could clap the story, I always put my 100% effort and dedication when writing these articles. Stay tuned and hit the follow button for weekly updates and mind-intriguing articles!