The Lightwave 

Practical Insights for Skeptics & Users Alike…in (Roughly) Two Minutes or Less

Here we are: another week, another edition of The Lightwave.

Today’s focus is on Supervised vs. Unsupervised Learning, two of the main methods used in Machine Learning, which is used in training many of these AI models.

Let’s get the standard definitions out of the way:

Supervised Learning: A machine learning approach where an AI model is trained on a labeled dataset, learning to map input data to correct output labels, so it can make predictions on new, unseen data.

Unsupervised Learning: Unsupervised learning is a machine learning technique where an AI model explores unlabeled data to discover hidden patterns, structures, or relationships without predetermined categories or outcomes.

For our purposes, imagine supervised learning as cooking with a detailed recipe. Using this method, we have the chef (AI), the ingredients (the data), and the author of the cookbook (the human programmer).

Here, we're teaching an AI to recognize different types of cuisine.

In supervised learning, we'd provide the AI with a cookbook full of labeled recipes. Each recipe (data point) would come with a list of ingredients, cooking instructions, and most importantly, a label to help classify it…perhaps something like "Italian," "Chinese," "Mexican," and so on.

The AI chef studies these recipes, learning to associate certain ingredients and cooking methods with specific cuisines. After going through the entire cookbook, when presented with a new, unlabeled recipe, the AI can use what it's learned to guess the type of cuisine.

In other words, this method is like cooking by the book. The process is structured, guided, and has a clear goal.

It's perfect for tasks where we know exactly what we want the AI to learn and can provide plenty of examples.

Unsupervised Learning: Creating a New Dish

OK, now picture unsupervised learning as a chef experimenting in the kitchen without a recipe.

The AI chef is given a pantry full of random ingredients (unlabeled data) and is told to create something interesting.

In our cuisine example, an unsupervised learning approach might involve giving the AI Chef thousands of recipes without any cuisine labels.

The AI would then try to group similar recipes together based on the patterns it observes in ingredients and cooking methods.

It might create clusters for "dishes with pasta," "spicy foods," etc.

Kind of like culinary innovation. There's no predetermined outcome, just exploration and grouping based on the available ingredients. It's useful when we want to discover new patterns in data or when we're not sure what we're looking for.

In the metaphorical Kitchen of AI, many real-world AI applications blend both approaches, like a chef who follows recipes but also innovates. For instance, a meal planning app might use supervised learning to suggest dishes based on your past preferences, while also using unsupervised learning to group similar recipes and suggest new dishes you might enjoy.

Check Mate

Tomorrow we’ll look at how Supervised and Unsupervised Learning methods often work together. Specifically, we’ll focus on AlphaZero, a groundbreaking artificial intelligence program created to master certain board games—more specifically for our purposes, Chess.

See you then.

For more information, visit www.NorthLightAI.com