Using LLMs for productivity without replacing the human originality

The ultimate goal of engineering is to enhance our lives, making them more enjoyable and livable. Milestones such as the invention of cars, washing machines, dishwashers, mobile phones, the internet, and industrial machinery have all significantly shaped the world we live in today. These innovations have also influenced the way we think and act. Recently, we've witnessed another transformative development: Generative AI.

Generative AI has fundamentally changed how we work and interact with others. It's being used in a wide range of tasks — from generating ideas and learning complex topics to drafting articles, writing resumes and cover letters, developing apps, and creating media. In many ways, Gen AI has boosted productivity and streamlined processes.

However, despite its impressive capabilities, Generative AI is not, and cannot be, a substitute for human thought. As humans, we possess a broader perception, the ability to generate ideas beyond existing knowledge, and the capacity to understand and synthesize abstract concepts. This distinction between artificial neural networks and natural neural networks is what makes us uniquely human.

There is a growing trend of relying on large language models (LLMs) to do tasks that require critical thinking and human judgment. For example, people often use Gen AI to write cover letters, even though these should be personalized and reflective of an individual’s unique fit for a job. Similarly, while Gen AI can assist in writing code, the true skill in coding lies in understanding the problem, being aware of the user, and considering the environment in which the code will run.

One of the pitfalls of over-relying on AI is that we often accept its output without giving it a second thought—or even a first one. This can lead to generic results that fail to meet their intended purpose. For instance, an API developed using Gen AI might overlook critical elements, such as returning a paginated response when the data set becomes too large or ensuring that user authorization middleware is properly called from another module. No matter how much "prompt engineering" you do, AI will never fully align with the nuanced and evolving business requirements.

The primary responsibility of an engineer is to build systems that are scalable, secure, and reliable. This is why AI will never fully replace engineers. While AI can undoubtedly change how engineers work—making coding faster and debugging easier—humans are still needed to resolve bugs, design complex systems, and ensure overall quality. It's also worth noting that the role of engineers might shift, with fewer needed to build certain things, but those who remain will need to engage in deeper, more critical work.

As a society, we should leverage Gen AI to enhance what already exists: improving the performance of slow systems, simplifying complex implementations, and creating clearer documentation for critical code. Gen AI can make your essays more fluid and help clarify ambiguous documents, but it shouldn't replace your thought process—rather, it should assist in refining and curating it.

As a software engineer, I use tools like Copilot to handle repetitive coding tasks, break down functions, or draft clear documentation. However, the original logic and critical thinking behind my work are entirely my own. Even when job hunting, I avoid using AI to draft messages; instead, I filter them using Claude.

Your thoughts are your essence, your way of writing code is your essence. Problems aren't generic, but LLMs are, there is no innovation with LLM.