As I drift through my winter vacation, detoxing from work, tech and world matters, I notice my little brother crawling toward a heat stove. With a week of rest behind me, my mind wanders into philosophical territory, and I can't help but map this moment to machine learning.
Learning through associations
Watching him, I'm reminded that when we're very young, we don't yet have a fully built network in our brains that would help us visually understand what is "hot" and what is "cold". When he reaches out and touches the stove, what some might call a wood-burning stove - he feels the burn instantly. It's not that he's thinking, "I shouldn't touch that because it's hot"; he's simply experiencing something new.
In that moment, his brain begins to wire associations. He sees the glowing surface, maybe notices a heat haze/mirage, and concludes "Okay.. things that glow red seem to be painful". He obviously doesn't know how to label it, until his mother comes and repeatedly tells to not come close - "it is hot!".
In machine learning terms, this process is like forming space embeddings - each sensory input is mapped into a mental space where similar experiences group together. Slowly, he starts to build a network, a kind of mental graph, where each node represents a piece of what "hot" can be.
Building a mental graph
Later on, when he encounters a heat radiator that doesn't have the obvious signs of a heat haze or glow but still gives him a burn, his mind takes note. Now he's not just associating "hot" with a glowing object. He updates his internal model: even without the dramatic visual cues, if an object radiates warmth, it might be hot. And in that moment, he's done what people like to call a "Bayesian update" - adjusting his beliefs based on new evidence.
This simple learning process, where he refines his understanding with every new touch, is very similar to how neural networks work ("yeah, no shit sherlock!" - you may comment). We feed data into a model, let it pick up patterns, and then update the "weights" of its connections as new information comes in aka Gradient descent.
We hunt for patterns
Everything about our brains is about forming and finding patterns. Life itself is full of patterns. Mathematics was once called a science of patterns, and music is composed of patterns. Even the way we perceive random things - like looking at clouds and saying, "This one looks like an elephant" - is pattern recognition at work.
We hunt for patterns because remembering them is easier than memorizing raw information. Instead of storing every tiny detail, our brains compress data into patterns, much like a well-optimized algorithm. We don't remember every individual note in a song, but we recognize the melody. We don't memorize every interaction we have with someone, but we identify their behavior through social patterns.
Machine learning mirrors this exact process. A model doesn't memorize all its training data - it finds patterns, generalizes them, and makes predictions based on what it has learned. Whether it's a neural network classifying images or our minds recognizing a familiar face in a crowd, both rely on the same fundamental principle: finding structure in chaos.
In a way, everything we do - solving problems, making decisions, even just recalling memories - is a reflection of this pattern-hunting instinct. It's how we make sense of the world.
In fact we humans may even go to the extent of finding patterns where there are none - a phenomenon called Apophenia. This is why we see faces in clouds, hear messages in songs played backwards, and find meaning in random events. Our brains are wired to find patterns, even when they don't exist. Here is Neil deGrasse Tyson debunking the mysterious square structure on Mars in a video.
Quiet moments
It's amazing how a quiet moment, in all its forms, can spark creativity. People say, "The best ideas come when you're in the shower," and it's true - sometimes that might be the only place where you find peace.
Ultimately, it's all about finding and forming patterns, so it's important to get some peace and a good night's sleep.
So next time you face a new challenge - maybe we should remember that every experience helps build your personal graph of knowledge. And sometimes, with just a small update, your entire understanding can change.