Research feature · 11

How professionals review Go games with AI

A six-stage professional review workflow, large-scale studies, and 60 Read19-analyzed games explain five elite styles and the real relationship among match rate, wins, and memorization.

  1. How professionals turn AI into a training system

    Preserve human judgment, scan, deepen key positions, test branches, and compress the result into one executable cue.

  2. Why one AI teacher does not erase style

    Precision, complication, deep reading, boldness, and pressure choose different practical difficulty from one candidate set.

  3. What ‘more AI-like wins more’ measures

    Exact match, policy, win-rate loss, score loss, and phase errors are different measures.

  4. How memory, novelty, and convergence coexist

    Local patterns concentrate while games enter novel global states earlier; memorization cannot explain the full gain.

Data, papers, and interviews

  1. PNAS · Superhuman AI, quality, and novelty
  2. PNAS Nexus · Knowledge and reduced diversity
  3. How Does AI Improve Human Decision-Making?
  4. Korean Physical Society · assistance detection
  5. People's Daily · Go alongside AI
  6. Korea Baduk Association · Shin versus KataGo
  7. The Paper · Ke Jie on AI
  8. The Paper · Ding Hao plays AI
  9. Nihon Ki-in · Ichiriki interview

Before trusting a number

  • Save human judgment before opening AI
  • Read win rate, score, candidates, and ownership together
  • Record model, rules, and budget
  • Describe style with error distributions, not one-game match rate
Open the review sample