HN Theater @HNTheaterMonth

AlphaStar's blink stalker micro, https://www.youtube.com/watch?v=cUTMhmVh1qs&t=5888s

with capped actions-per-minute and reaction time: https://news.ycombinator.com/item?id=18992698

⬐ MaxikCZ

This is what I dont like about Starcraft. The strategy is, in many instances, second to microing the units. The fact that 1hp unit have the same effectivenes as a full hp one is just weird.

Deepmind did a demo of it: https://www.youtube.com/watch?v=cUTMhmVh1qs

This video seems to imply that latency of the neural networks is around 350ms, which is fast but within human ranges: https://youtu.be/cUTMhmVh1qs?t=1488

This was addressed in their presentation 6 months ago.

https://youtu.be/cUTMhmVh1qs?t=1460

TLDR: Both their APM and TTFA is comparable to human pros.

⬐ JyB

You are skipping over the fact that the chart is highly misleading.

⬐ ehsankia

In terms of APM, yes, which is what this FAQ addressed:

> AlphaStar has built-in restrictions, which cap its effective actions per minute and per second. These caps, including the agents’ peak APM, are more restrictive than DeepMind’s demonstration matches back in January, and have been applied in consultation with pro players.

So yeah, they've tweaked this specifically, although not much details as to how.

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 Stockfish and LeelaZero compete against one another in their own respective AI chess leagues.

[0] https://www.youtube.com/watch?v=cUTMhmVh1qs

[1] http://wiki.sc2ai.net/Main_Page

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.

https://youtu.be/cUTMhmVh1qs

I have not played much StarCraft, so I can’t judge for myself.

⬐ luckylion

Sure, it does some stuff that's fine, but that's not what makes it win (yet). It's basically playing in Archon mode (but alone) where it can see and act all over the map at the same time - humans can't do that. I'm sure it will at some point become better at strategies, too, but so far it's hard to compare because the API it has is much more powerful.

When they did change that aspect of the match, the 10:0 turned into 10:1. Would Mana have won the next 10 games? Nobody knows, but at least it would've been a lot closer. SC2 rewards micro a lot, and a computer will always be better at micro than a human (unless you make it harder for the machine). https://www.youtube.com/watch?v=3PLplRDSgpo This is a couple of years old, and at that level, it would certainly beat any pro, even without a superior strategy because micro is a leverage for unit strength.

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.

https://www.youtube.com/watch?v=cUTMhmVh1qs

Overfitting maybe.

The show off game: https://www.youtube.com/watch?v=cUTMhmVh1qs&t=2h31m22s

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.

It was live streamed, you can watch it here: https://www.youtube.com/watch?v=cUTMhmVh1qs&t=2h31m22s

Tthe first 2 and a half hours goes over the other 10 matches, which weren't played live on stream.

The APM of AlphaStar was about half of the professional player in this match.

Check out: https://youtu.be/cUTMhmVh1qs?t=3189

⬐ Narishma

But when it counts, such as during micro-heavy battles, it's much faster and more precise than a human.

⬐ fernandotakai

yup. you could have 200 apm, but as long as your clicks and button presses are perfect, you are going to win against someone with 800 but is super imprecise.

blink stalkers are basically perfect for an AI because of the precision they can blink them around.

https://youtu.be/cUTMhmVh1qs?t=3218

I thought they said 350ms.

E: https://youtu.be/cUTMhmVh1qs?t=3272

⬐ Jabbles

They also say 50ms: https://youtu.be/cUTMhmVh1qs?t=7016

Here is another link (YouTube) that allows you to go back in time if you've missed anything: https://www.youtube.com/watch?v=cUTMhmVh1qs

⬐ dplgk

32 points and not a single comment?

⬐ xbmcuser

Its a 2+ hour video most people forgot how they got to the page by the time the video ended