DeepMind StarCraft II Demonstration
Why Hacker News recommends it
HN commenters link it as the primary source for AlphaStar's matches and design, citing its explanations of training, APM and reaction latency, and its commentary on the AI's play.
Recommended by 2 people across 1 different months, 2019 to 2026.
What it is
A 2h22m livestream from DeepMind's London office in which StarCraft II casters Artosis and RotterdaM present AlphaStar, DeepMind's StarCraft II agent. It replays AlphaStar's 10 pre-recorded matches against pro players TLO and MaNa, then covers the live showcase match where MaNa beat AlphaStar, with DeepMind researchers explaining the work.
Cited in 1 Wikipedia article, including AlphaStar (software).
What people said
Recommends rococode · 2019-01 · Hacker News · on “Humanity Has Hope: Pro-Gamers Win 1 of 11 StarCraft Matches Against DeepMind AI”
It was live streamed, you can watch it here: [this video] Tthe first 2 and a half hours goes over the other 10 matches, which weren't played live on stream.
Recommends deburo · 2019-01 · Hacker News · on “Humanity Has Hope: Pro-Gamers Win 1 of 11 StarCraft Matches Against DeepMind AI”
One step at a time sounds like a better strategy. Watch this and learn a bit more about why they're doing Starcraft AI. The Deepmind guys first appear around 32:00. [this video]
Cites strangecosmos · 2019-01 · Hacker News · on “Ask HN: What can AlphaStar tell us about imitation learning for autonomous cars?”
I just watched the live-streamed match where MaNa beat AlphaStar and the professional commentators seemed to think that AlphaStar had good strategy. MaNa had even copied something AlphaStar did in a previous match. [this video] I have not played much StarCraft, so I can’t judge for myself.
Cites kregasaurusrex · 2019-07 · Hacker News · on “DeepMind’s StarCraft II agent will play anonymously on battle.net”
Having watched the livestream from January[0], the AlphaStar agent was able to create its own individual agents to play against and then learn from in the same way AlphaZero learned from chess & baduk/go historic games. In AlphaStar's case this learning period was described as experiencing the equivalent of hundreds of years of aggregate gameplay. This allows an agent to adjust to novel strategies not seen before because Starcraft II's number of possible game states exponentially increases as the game progresses; compared with chess and go that have large albeit fixed board states. There's also a competition for individuals who build their own AI engines[1] for Starcraft II much like how Sto
Cites noego · 2019-07 · Hacker News · on “DeepMind’s StarCraft II agent will play anonymously on battle.net”
This was addressed in their presentation 6 months ago. [this video] TLDR: Both their APM and TTFA is comparable to human pros.
Cites chillacy · 2019-07 · Hacker News · on “DeepMind’s StarCraft II agent will play anonymously on battle.net”
This video seems to imply that latency of the neural networks is around 350ms, which is fast but within human ranges: [this video]
Cites dschuessler · 2026-09 · Hacker News · on “Brood War Bench”
Somewhat related: In 2018, Google DeepMind had already created AIs that were capable of beating professional gamers in StarCraft 2 (the sequel to Brood War): [this video]
Critical rightbyte · 2019-01 · Hacker News · on “Humanity Has Hope: Pro-Gamers Win 1 of 11 StarCraft Matches Against DeepMind AI”
Overfitting maybe. The show off game: [this video] The bot is getting it's main force lured back to it's base by hit and runs for it's resource producers. Repeating the same mistake again and again. Later it acts confused and rather have it's main base wrecked than having a high risk fight with it's main force vs the players main force, even though it would auto-lose not doing so. I've got a feeling these bots have a huge API advantage in "mechanical" skills versus humans. Eg. it has to be way faster to retrieve the "x ability used by y" event than seeing an animation unfold on the screen, leaking intention to bots earlier then human players.