Currently

General Interest

  • Algorithms and economics.
  • Intersection of discrete and continuous math.
  • Mathematical and algorithmic foundations of learning.
  • Type theory for math discovery and computing practice.

with emphasis on theoretical foundations and provably guaranteed frameworks using proof and logic techniques.

Particular Curiousity

  • The internet revolutionized humanity’s connectivity, AI revolutionized information flow, but rarely do we cooperate. Socialists have ambitions for Collective Intelligence, a paradigm of AI which enables coherent cooperation among the crowd, but no technical progress of practical relevance is witnessed. In the same way Shafi and Silvio’s zero-knowledge proofs enabled a new kind of secure communication, collective intelligence could lead to a new kind of economic interaction. See Generative Social Choice.
  • Combinatorics is rooted in algorithmic toolboxes, but in the data driven world, statistics, algebra, analysis and topology are the mathematical underpinnings. Topological Combinatorics within the math community studies connections between Topology and Combinatorics. Spreer’s works are ambitious. Can we use Topological Combinatorics to reduce continuous algorithms to discrete ones, or to understand continuous algorithms through our combinatorial toolbox?
  • AI safety and understanding machine learning became very critical after their unexpected advancement. Moitra aims for understanding machine learning by its algorithmic aspects; Learning-augmented algorithms retains robustness of algorithms when predictions are inaccurate; Neuro-symbolic designs a hybrid of data-driven and symbolic-proven components. Nonetheless, theory and logic are not catching up with the rapid progress of practitioners. Are there near-term intermediate approaches to bridge theory and practice?
  • Functional programming is a language that enables correct coding by definition. It is closer to a logician than a computer scientist who is used to machinery in Turing machines. Attempts like Lean and Terry Tao’s project explore how CS and LLM could contribute to mathematical discovery. Type Theory forall and Categories for AI investigate Logic contributing to computing and AI. Nowadays it is believed natural language is the future of computing, but what about formal languages?

Philosophy.

What is Theoretical CS?

A good CS theory is abstract from specific applications, yet explains or influences computing practice. CS theory may not solve useful problems but the conceptual idea and concepts are motivated by practice. Solutions tied to specific concrete problems loses the potential and power of abstraction. CS theory irrelevant to practice may advance math but then it won’t be a theory for computing. Tim’s beyond the worst-case analysis is inspiring, and I like Oded and Avi’s 1996 essay.

Theory vs Application

A theorist’s mission is to progress the intellect of humanity. It’s about enlightening new conceptual ideas, and resolving technical challenges. Theory could lead to applications as number theory preceeded cryptography. Applications could stimulate theory as OpenAI engineers have built what we theoretically thought is not doable. It seems progress is driven by an interplay among diverse fields of inquiry and approaches. A good theorist does not get enclosed within a bubble, but progresses the intellect, following whatever angle is more promising. I don’t see boundaries between math, theoretical CS, and practical CS. Roger Penrose’s Fashion, Faith, and Fantasy is full of wisdom.

Community

I care about humanity’s future and economic growth by means of leadership. Science’s intellectual progress is a social endeavour, and hence it is natural for a theorist to care about people, teaching, and communication.

Service.

I do moderate TCS’s awesome list, getting the chance to meet students all over the world.

Good Memories.

Family.

Mum & Dad (Lawyer), Brother (Athlete Coach), Sister (Passed away aged 14).