Stop Learning Patterns, Start Solving Problems - Lessons from Biology and Engineering

Throughout history, humanity’s approach to understanding and creating has shifted dramatically. From categorizing the natural world to designing complex systems, there has always been a tension between rigid frameworks and creative exploration. This shift is especially evident in two seemingly different domains: biology and engineering. Both fields have evolved from emphasizing predefined patterns to embracing problem-solving as a dynamic, imaginative process. This evolution holds valuable lessons for anyone navigating the complexities of modern innovation.

The Era of Categorization: Biology’s Early Days

In the 18th and 19th centuries, biology was driven by a need to classify and categorize. Scientists like Carl Linnaeus created taxonomies—hierarchical systems that sorted organisms into kingdoms, phyla, genera, and species. This approach made sense for its time. The world was vast and uncharted, and organizing it into neat boxes was the first step toward understanding it.

While taxonomy was foundational, it had limitations. It treated nature as static and segmented rather than interconnected and dynamic. This mindset began to shift with Charles Darwin’s theory of evolution. Instead of viewing species as fixed entities, Darwin revealed them as evolving, adapting systems. This sparked a revolution: biology was no longer just about classification; it became about understanding processes, relationships, and systems.

The shift continued with the rise of genetics, molecular biology, and systems biology. Researchers began focusing on how things worked, not just what they were. Patterns were still useful, but they became tools, not endpoints. This flexible, process-driven approach unlocked profound insights into the complexity of life.

The Parallel in Engineering: Learning to Build

A similar story is unfolding in engineering education and practice. Many engineering courses today emphasize predefined patterns and methodologies. Whether it’s software design patterns, system architectures, or AI frameworks, students are often taught that there are “N distinct ways” to build a system and that learning these patterns is essential.

While these patterns can provide helpful starting points, they risk becoming constraints. Rigidly adhering to predefined templates limits creativity and discourages problem-solving. Real-world problems rarely fit neatly into predefined categories, and innovation often requires breaking away from existing molds.

From my experience working on AI agents, I’ve found that the most valuable insights come not from memorizing patterns but from confronting challenges directly. When my team and I started building AI systems, we didn’t begin with a library of patterns. Instead, we faced the problems head-on, experimented, and let solutions emerge organically. The patterns we discovered were born out of necessity and tailored to our specific use cases. They weren’t taught; they arose naturally.

Why Problem-Solving Beats Pattern Learning

The key takeaway from both biology and engineering is this: predefined patterns can inspire, but they shouldn’t constrain. Here’s why problem-solving is a more effective approach:

• Creativity Thrives Without Limits: When you’re not bound by established patterns, you’re free to explore unconventional solutions. This often leads to breakthroughs that wouldn’t arise within rigid frameworks.
• Patterns Are Context-Specific: What works in one scenario might not work in another. Discovering solutions tailored to your unique challenges ensures better outcomes.
• Learning by Doing: Problem-solving fosters deeper understanding. Instead of memorizing abstract patterns, you gain hands-on experience that stays with you.
• Flexibility in Innovation: When new challenges arise, a problem-solving mindset equips you to adapt, while a pattern-driven approach may leave you stuck searching for the “right” template.

Conclusion

For students in AI, I would recommend not spending too much time learning patterns or design recommendations. Instead, focus on solving problems and understanding the foundational principles and mathematics behind a subject. When you engage directly with challenges, the patterns will arise naturally.

I’ve experienced this myself when learning about different types of recurrent neural networks (RNNs). Early on, I spent too much time trying to memorize the various types and their applications. But when I shifted my mindset and started thinking about real-world problems, the patterns became obvious. For example, in translation tasks, it’s clear that the architecture naturally forms an N-to-N schema. By focusing on applications and problem-solving, you’ll gain a deeper understanding and develop solutions that fit your unique use cases. So, let go of rigid frameworks, and let curiosity and creativity guide your learning journey.

Medium link: https://medium.com/@timothee.guedon/stop-learning-patterns-start-solving-problems-lessons-from-biology-and-engineering-d663c35a834c