moinsen / Software historyDE ↗
← 2016

Event

AlphaGo gewinnt gegen Lee Sedol

Context and research status
  • Kennung: 2016-alphago
  • Zeitraum: 2016
  • Datumsgenauigkeit: Jahr, sofern im Text nicht genauer / year unless specified in text
  • Art: historische Entwicklung / historical development
  • Kategorien: Wandel und Einflussfaktoren, Sprachen und Techniken, Ereignisse, Zahlen und Fakten
  • Akteure und Rollen: Lee Sedol: Go-Spieler; DeepMind: Entwickler von AlphaGo
  • Region: Südkorea / South Korea

In March 2016 DeepMind’s AlphaGo won a Go match against Lee Sedol by four games to one. The system combined neural networks with tree search. Source

A January research paper had already described an AlphaGo system trained on human expert games and self-play that had beaten Fan Hui 5–0. The later AlphaGo Zero, described in 2017, learned without human game data; the game rules were still provided. The January paper must be distinguished from the match played in March. Research papers

The January paper describes the division of labor: one neural network suggested promising moves, another evaluated positions, and Monte Carlo tree search combined these assessments with simulated play. This describes the published research variant and is not a complete specification of the later match system. According to DeepMind, Lee Sedol won the fourth game of the March match. Paper and match retrospective

Sources

Offen

Weitere Ereignisse, Einflussfaktoren und unabhängige Einordnung sind zu recherchieren. / Further events, influences and independent context remain to be researched.