The Zen and the Art of Machine Learning Research

Becoming an AI researcher is similar to meditation. You must not simply read alone or build alone; discipline to consistently invest time while engaging in both Reading and Building in parallel is essential.

When choosing a research topic, rather than getting caught up in recently trending technologies (Agents, Context Engineering, etc.), you should focus on the Basics. You need a process of understanding deeply enough to directly calculate Cross-entropy or visualize SVD yourself.

Research that simply raises existing Benchmark scores is likely to lack depth. The ability to find datasets that can properly validate new methodologies is a core competency of modern AI research.

Also, an open mind that isn't trapped by past methods of success or fixed assumptions is important. Since the AI field is changing extremely fast, you must let go of ego and constantly explore new possibilities.