First written in July 2025.

My hobby was used to reading management books about how private and public organizations make contributions in their field. I think many organizations that do things related to information and support share a timeless goal of helping people to understand their private and public worlds because that is what information and support can help. Before the digital era, languages, papers, newspapers, radios, or TVs were the great medium for that, in the digital era we have witnessed bidirectional communications infrastructure - the internet, and many websites on top of the internet, information retrieval systems for navigating these websites, question answering systems that just returns the answer, and intelligent virtual or physical assistant that provides information and various support. Just like computer architecture, the coarse layer in each needs its own research (e.g. linguistics (for language), mechanical engineering (for printing newspapers), electrical engineering (for radios, TVs, and internet infrastructure), mathematical and machine learning science (for information retrieval, QA systems, and intelligent agents)).

Currently, my focus is on the last layer of this information and support architecture, but as a researcher, I hope we can focus on things that don’t change, such as helping people with information and support to understand the world, instead of focusing the current implementation of each coarse layer. For instance, to be freed at current implementation, I hope our research at each layer can be uniquely foundational to move forward to right future topics, but still fall into the broad topic of information and support for humans. I hope our study in each layer is conducted by leveraging innovations from the different scientific components in other layers or the same layer (I think well-known examples are, Information retrieval model in the web-scale collection and feedback, Representation learning model for Natural language processing, Using accelerator for greedy training of neural networks) to develop further.