Life in Loops: When the Model Learns & Unlearns

For a long time, I thought the reason my mind's eye stayed dark; unable to voluntarily picture scenes or characters, was because I never watched TV until my late school years and college. As an abstract thinker, art was my only solace, but drawing taught me something unexpected. I realized I needed to trace or copy things to create art despite long term practicing, rather than inventing images purely from my head. As someone who has loved self-diagnosing since childhood, this led to a massive breakthrough. I realized my brain isn't broken; it is simply running a highly specialized system architecture known as congenital aphantasia, a cognitive variation where a person cannot form mental images. Recent studies show that this visual imagery weakness has a prevalence of roughly 3.9% to 4.2% in the general population, but among neurodivergent individuals, this rate can jump significantly higher.

Instead of using a standard graphic processing unit to render heavy visual pictures, an aphantasic mind operates on a hyper-efficient vector processing engine. Neuroimaging data shows that individuals with aphantasia retain accurate visual knowledge and logical facts because of distinct pathways connecting different brain networks. The brain simply translates the world into abstract logic, spatial lines, and structural tokens rather than raw pixels. This blueprint-style processing setup, often reinforced by a lack of early media exposure, allows the mind to absorb massive amounts of knowledge directly as code. Once I understood this mechanical design, it completely changed how I look at my own personal growth. If we look at our own lives through this exact lens of Machine Learning, we can finally understand why patterns repeat, why we get stuck, and how we can successfully evolve our human software.

The other day, I was reflecting on a weird pattern in my life. Why do the exact same problems keep showing up in cycles, almost like an odd-year loop?

Your life is split into two major phases or periods which can be described as training datasets and test datasets. The training phase is when life hands you a brand-new, fully formed situation. In these moments, you get all the variables clearly laid out, including the text, the conversations, the visual details, and the deep emotions. You make mistakes, figure things out, and learn a lesson. Years later, the test phase begins when a similar situation appears out of nowhere. The people and places look completely different, but the core problem is identical. This is life's way of bringing back old data to see how your system responds this time to evaluate one's thinking approach.

When we struggle with these life tests, it is usually because of common algorithm errors like overfitting, underfitting, or data bias. Overfitting happens when you learn a past lesson too perfectly and mistake tiny, random details for the actual lesson. For example, if a friend broke your trust on a rainy Tuesday, an overfitted brain decides that nobody can ever be trusted when it rains. Underfitting is the opposite error, where you ignore the data completely, miss the warning signs, and walk right into the same bad situation wondering why the outcome didn’t change. Data bias occurs when your worldview is built entirely on a one-sided group of early experiences. If your childhood taught you that love equals shouting, your biased data will make you misinterpret a calm, quiet partner as someone who doesn't care about you.

We also shape our habits through a process called reinforcement learning, which is how a computer learns by trial and error using rewards and punishments. Computers adjust their behavior to get as many digital rewards as possible, and humans do the same, but we often get stuck in short-term reward traps. Your brain loves instant rewards like scrolling on social media or eating sugary snacks because they give you an immediate burst of happy chemicals. On the flip side, good habits like exercising feel like an immediate punishment because they require hard work right now. To grow, you have to trick your algorithm by connecting your daily actions to long-term rewards, focusing on the ultimate feeling of being strong and healthy tomorrow rather than the temporary effort of today.

To fix these glitches and upgrade your algorithm, you must practice what engineers call hyperparameter tuning. In human terms, these parameters are your personal boundaries and daily habits. If your system is constantly crashing from exhaustion or stress, you do not need to throw away the whole computer; you just need to twist the dials. You can tune your settings by turning down your sensitivity to other people's opinions, turning up your boundary dials to say no to extra work, and optimising your recharge cycle with proper sleep.

However, a machine cannot build or fix itself. In the tech world, an AI needs an entire team of people to make it work. It needs Data Engineers to clean up the messy data, Researchers to design the system, and Human Feedback to point out its mistakes. In the human context, we must intentionally become the engineers of our own minds. We do this through self-reflection, journaling, and therapy, acts that clean up our internal data. We also rely on external "human feedback", our true friends, mentors, and loved ones who love us enough to hold up a mirror and tell us when our software is glitching.

Understanding your own software in the background also changes how you view others and yourself, highlighting a concept engineers call model architecture. In AI, different models have completely different ways of processing information. A text-based model doesn't waste energy rendering heavy graphics or movie scenes; instead, it converts words into abstract geometric matrices, symbols, and pure semantic data. My artistic journey of needing to copy or trace was not a flaw, it was just my vector engine at work. If your mind skips the internal visual scenery and goes straight to processing raw logic, lines, and geometric symbols, your architecture is built for hyper-focus. It allows you to absorb massive amounts of knowledge directly as a blueprint. You cannot force a model built for high-level data processing to suddenly act like a graphic rendering engine, nor should you want to. Accepting your unique cognitive architecture frees you to run your system at its highest efficiency.

This architecture standard also applies to how we interact with others. We cannot force a simple, early-stage AI model to solve a complex three-dimensional puzzle if it lacks the internal layers and network capacity to process that much data. In life, you will encounter people who cannot understand your boundaries, your emotions, or your growth. It is vital to realize that event is a data set, you cannot argue or force them into understanding. Their current brain structure and emotional architecture are simply not built to process that information yet. Accepting this saves you from wasting your energy trying to reprogram a system that isn't ready.

This simple data journey brings us to a much deeper truth about what it means to be human. The most vital opportunity we have is learning not to confuse our internal model with reality itself. Your brain gives you a model of the world. Your childhood gives you a model of love. Your experiences give you a model of trust. Your failures give you a model of risk, and your successes give you a model of possibility. Your unique cognitive architecture determines how you represent all of it, but none of those models is reality itself. We are not our training data. We are not even the model produced by it. We are the system capable of noticing that the model may need updating.

The recurring "tests" in our lives are not exam papers designed simply to be passed or failed with a lower error rate. Sometimes, the situation returns not to ask if you know the correct answer, but to ask if you can respond differently now that you are different. The model itself changes through every single trial. The person who enters the test is not the same person who completed the initial training. True learning doesn't give us perfect, unshakeable answers. Instead, it gives us an algorithm that is brave enough to question its own learning. The opposite of an outdated model isn't certainty; it is curiosity.

I started this journey thinking my mind was broken. Then, I discovered it was simply different, and I learned that this difference can be an incredible architecture. Eventually, I realized that everyone around me has their own complex architecture, and I stopped trying to force other systems to behave exactly like mine. Growth is not about finally finding a perfect, rigid code that never makes a mistake. True evolution is learning to remain soft, remaining open, and remaining willing to continuously update your own model.

“A mind does not have to picture the world to understand it. Sometimes it sees through structure, pattern and meaning instead.”