Lately, whenever I read pieces about AI, I keep running into the English word taste. In Korean it is often glossed as “preference,” but every time I see that, the translation feels a little lacking. “Preference” usually means a person’s stable, personal pattern of liking something. It is the kind of thing you mean when you say, “I prefer old-school coffee to an Americano.”
But if you actually follow the contexts where taste appears, it seems to point to something beyond mere liking. In those contexts, what I understand by taste is closer to discernment, or an eye for what matters. Preference is thoroughly subjective and does not require logic. Discernment, by contrast, demands its own logic and standards. And those standards are rarely fixed. They move with the situation, the goal, and the person in front of you. Taste is a standard of judgment built up over a long time, and when that standard fires in an instant, we call it intuition.
People who love mathematics often say the appeal lies in the fact that there is exactly one correct answer. Perhaps they love it because most of the problems we face in life do not work that way. Every problem comes with several options that are close to being answers, but none of them is complete. Choose A and you gain B but lose C. Choose B and you gain D but lose E. There is no immaculate right answer, only a rough judgment about what is most suitable in this moment.
To make the right choice, you certainly need enough knowledge and experience. But that alone is not enough. If knowledge and experience are raw data, then intuition is what performs the computation. And that intuition is nothing other than the momentary activation of the taste we have built over time. There is no textbook that tells you exactly how to combine the problem in front of you, the constraints you face, the situation of your organization, the roadmap, and the operating environment in order to decide what the best choice is. Often the answer emerges from the intuition that grows out of long accumulation of knowledge and experience, and from the taste that sits on top of them.
No one now denies that AI is exceptionally good at generating and summarizing content. But the ability to create something genuinely new, or to possess taste, still seems to remain in the human domain. AI is excellent at producing high-quality output that follows standard patterns, but choosing the best move a specific situation demands still calls for human intervention.
Of course, one could object and ask whether even taste might one day be replaced by AI. In fact, statistical taste—the intuition that says “if we structure it this way, the data flow will bottleneck here”—may well be something AI can do better than humans. Pattern recognition based on data has always been one of the machine’s strengths.
But the taste that determines what direction I, our team, or our product should take feels likely to remain human for quite a long time. Those choices ultimately connect to the question, “What kind of self, team, or product do we want?” That is a matter of value judgment, and values are something human beings create.
Whenever we face a difficult choice, we look for precedents, data, or advice from other people. We ask AI. Some people even ask a fortune-teller. But at the final moment, the one who makes the choice is always an individual. Whether that choice was right or wrong can only be known after enough time has passed.
Seen that way, taste is less an innate talent than a sensibility we accumulate. By going through countless moments of choice and then reflecting on their results, we gradually learn which kinds of choices tend to lead to good outcomes in which contexts. It is not something you acquire overnight. It is sharpened slowly through repeated mistakes and repeated successes.
And even that taste takes on a different shape for each person. It wears down into a different pattern for everyone, the way a pair of hard dress shoes slowly wears according to the gait of the one who walks in them.
In the world developers work in, all data exists in objective and observable form, and the logic that handles it also exists in clear and predictable form. It is, so to speak, a closed world, with fixed variables and visible cause and effect. After swimming in that world for a long time, there are moments when relying on something as subjective and hard to grasp as intuition feels deeply uncomfortable. But the reality I actually live in is an open world where infinite variables act in real time, and in a world like that there is very little to lean on except intuition, sharpened by well-honed taste.