AlphaGo

4 atoms · 跨 4 天 · 首见 2026-05-15 · 最近 2026-06-22

三色: 🟦 fact 3 · 🟥 take 1 · stance ▲0/▼1/◆1

来源: X 4

时态: fresh:4

标签: 好观点:2 · 好思考:2 · 好数字:1

叙事 (narrative) (1)

• 2026-06-21 · parameter count for impressiveness
  AlphaGo wouldn't have been very impressive at ~50k params
  tags: 好观点 · → daily

事实 (fact) (3)

• 2026-05-20 · 使用的 PUCT 算法评分逻辑
  PUCT scores each move with two competing terms: how good a move looks based on your exploration up till now, and a novelty bonus that rewards moves you've not visited much.
  tags: 好观点·好思考 · → daily
• 2026-06-22 · can do well in one domain but cannot generalize
  same logic for alpha go, alpha zero, they can do well in one domain, but can’t generalize
  tags: 好思考 · → daily
• 2026-05-15 · training cost
  In 2026, you no longer need DeepMind's resources to train a strong Go AI - you can vibe code all of it yourself for just a few thousand dollars of rented compute.
  tags: 好数字 · → daily
  value: qty=few thousand dollars · date=2026

时间轴 (近 20)

• 2026-06-22 · fact · can do well in one domain but cannot generalize · → daily
• 2026-06-21 · narrative · parameter count for impressiveness · → daily
• 2026-05-20 · fact · 使用的 PUCT 算法评分逻辑 · → daily
• 2026-05-15 · fact · training cost · → daily

← 实体目录 · 系统日志