Time: Mar 23, 2025 08:30 PM Eastern Time

Welcome to an engaging Show and Tell session hosted by Research Triangle AI! In this session, Guoliang will introduce Daoqi, a fascinating variant of Weiqi (Go), along with a powerful AlphaZero-like AI that he trained from scratch. He will demonstrate the AI in action, share insights into its implementation and training process, and explore the interesting technologies used to host the game online.

Agenda:

• Introduction to the game (Daoqi - a variation of Go)
  https://daoqi.club/
• Demo of the Daoqi AI (based on a famous Go AI program - KataGo)
• How is the AI program trained? AlphaGo -> AlphaZero -> KataGo -> DaoqiAI
• How is the web interface built?
• Q & A

What is Daoqi?

Daoqi is a board game evolved from Go (Weiqi). Similar to Go, both players use black and white stones, placing them on the intersections, with victory determined by the amount of territory controlled. However, Daoqi's board design is unique, with each line connected end-to-end, logically forming a torus. This borderless and cornerless characteristic means that Daoqi lacks the traditional corner and edge patterns found in Go, with all battles occurring in the central area. This design requires players to have a stronger sense of the overall situation, as each stone may simultaneously face attacks from all directions.

Speaker：Guoliang Cao, AGA 6D, Co-inventor of Daoqi

Guoliang has been a passionate Go (Weiqi) enthusiast since high school, achieving a highest rank of AGA 6D, which is comparable to Amateur 3D in China. Beyond playing, he enjoys exploring new ideas and innovations in the game. In 2003, he independently invented Daoqi, a variation of Go—though many others have independently conceived similar versions at different times.

Since then, he has dedicated significant time to developing computer programs that enable people to play Daoqi. In 2005, he organized an online tournament to promote the game. Beginning in 2022, Guoliang adapted the renowned open-source Go AI program KataGo to play Daoqi. Through training and refinement, he successfully developed an AI that evolved from a complete novice to a superhuman-level Daoqi player.