New ideas can only arise when new information is available. That new information needs to make new connections to existing knowledge and experience. Yet, this period of learning and often doubt is antithetical to teaching, sharing knowledge, and reproducing existing information. How do you reconcile the need to grow and the need for stability?

This question was particularly important during my PhD. How do you learn about a topic that is completely new to you, namely machine learning, yet produce journal articles that are supposed to advance science and humanity, inadvertently projecting expertise? What's worse, how do you do this in a field that is growing every day?

I learned to divide the process of creation into convergent and divergent phases.

In divergent phases, I focus on research. I spread out my interests and regard different sources that can provide new insights. These phases involve an increased amount of reading specifically adjusting the time blocks in your calendar towards additional time of consumption. In the beforetimes, when the outside was still allowed, this also included a focus on meetups and hackathons.

Nowadays, this also includes online presentations, Zoom chats with those people that lift you up and inspire, and, if you're lucky, those few individuals in your bubble. These sessions with peers are essential for a simple reason. Self-guided exploration, e.g. reading, often involves similar sources, limiting the degree of divergence of this phase. Oftentimes, I would read a paper and then continue on that train of thought. Whereas presentations, conferences, and those lovely one-on-one conversations force us to explore new ideas entirely. This maximizes the exposure to new information and, therefore, potential ideas and connections one can make.

Sketch of convergence and divergence

However, and this is key, these divergent phases need to be limited. In machine learning, new information is published daily. This influx of new ideas is debilitating when trying to produce something based on that information. How do you build something new if the base keeps shifting due to new research?

Convergent phases are the Yang to the divergent Yin. These phases are focused on building and creation. During these phases, it's important to limit consumption to an absolute minimum. These phases are to process the collected information and converge on specific new insights and ideas. For this reason, I block out additional time for writing on the calendar.

These phases are important for sorting through information obtained in the divergent phase. To me, writing is an exceptional tool to systematize information and work on clarity—the act of writing and editing forces information to a coherent set of ideas. Deciding to block out most additional information for a limited time ensures that there is mental space to converge on this set of ideas.

In the end, continuous learning and content creation, be it blogs or journal articles, depends on mastering these two phases.