DeepMind AI AlphaStar Wins 10-1 Against 'StarCarft II' Pros (newscientist.com)

← Back to Stories (view on slashdot.org)

DeepMind AI AlphaStar Wins 10-1 Against 'StarCarft II' Pros (newscientist.com)

Posted by BeauHD on Thursday January 24, 2019 @01:00PM from the there's-a-first-time-for-everything dept.

In a series of matches streamed on YouTube and Twitch, DeepMind AI AlphaStar defeated two top-ranked professionals 10-1 at real-time strategy game StarCraft II. "This is of course an exciting moment for us," said David Silver at DeepMind in a live stream watched by more than 55,000 people. "For the first time we saw an AI that was able to defeat a professional player." New Scientist reports: DeepMind created five versions of their AI, called AlphaStar, and trained them on footage of human games. The different AIs then played against each other in a league, with the leading AI accumulating the equivalent of 200 years of game experience. With this, AlphaStar beat professional players Dario Wunsch and Grzegorz Komincz -- ranked 44th and 13th in the world respectively. AlphaStar's success came with some caveats: the AI played only on a single map, and using a single kind of player (there are three in the game). The professionals also had to contend with playing different versions of AlphaStar from match to match. While the AlphaStar was playing on a single graphics processing unit, a computer chip found in many gaming computers, it was trained on 16 tensor processing units hosted in the Google cloud -- processing power beyond the realms of many.

103 comments

Min score:

Reason:

Sort:

said the bot by Anonymous Coward · 2019-01-24 13:14 · Score: 0

first
blanking in progress by Anonymous Coward · 2019-01-24 13:17 · Score: 0

this has been the beginning of my wall scribblings
Still Cheating by Luthair · 2019-01-24 13:20 · Score: 2, Informative

Much like the dota bot from last year AlphaStar is effectively cheating as it is aware of the entire map at once, not restricted to the viewport as humans are. These are only really a novelty until they start operating on imperfect knowledge and imperfect inputs as humans are (even if it was arbitrarily limited on reaction speed).
1. Re:Still Cheating by psycho12345 · 2019-01-24 13:33 · Score: 4, Insightful
  
  Actually, the AI is restricted to the viewport, it is using an interface into the game, so no it doesn't have maphack, like a normal scripted AI would.
2. Re:Still Cheating by Anonymous Coward · 2019-01-24 13:39 · Score: 0
  
  Watch the demo - the engineers explain the limits they impose on the program.
  Both APM and observations are limited to mid-human-player levels. It's not that computer is seeing more, or clicking faster. It's wisely using resources available to it to defeat players. It's quite impressive, actually. If AlphaGo progression is followed, they'll beat best SCII player in about 6 months.
3. Re:Still Cheating by djinn6 · 2019-01-24 13:43 · Score: 4, Insightful
  
  It's already operating with imperfect knowledge. From Vox:
  
  During the 10 matches, the AI had one big advantage that a human player doesn’t have: It was able to see all of the parts of the map where it had visibility, while a human player has to manipulate the camera.
  Emphasis mine. Yes, it's an advantage, but it's not cheating. Humans can use the minimap to see what's going on as well.
  The problem with these matches and Starcraft in general, is that the it's able to win just with good micro. So getting really good at Starcraft doesn't get us closer to actual AI.
4. Re:Still Cheating by wlorenz65 · 2019-01-24 13:52 · Score: 3, Informative
  
  Wrong for the live game against MaNa. Quote: "Following the broadcast of the recorded matches, DeepMind introduced a new version of AlphaStar that MaNa took on in a live match. The agent that played the live game didn't have the benefit of the overhead camera and instead had to make decisions on where to place its focus in the same way a human would."
  
  But you are right for the recorded games from December 2018.
5. Re:Still Cheating by Gravis+Zero · 2019-01-24 13:53 · Score: 0
  
  Much like the dota bot from last year AlphaStar is effectively cheating as it is aware of the entire map at once, not restricted to the viewport as humans are.
  That's not true. AlphaStar is using the same interface as other players. It's not simply multitask faster than the human players.
  
  --
  Anons need not reply. Questions end with a question mark.
6. Re:Still Cheating by The+Evil+Atheist · 2019-01-24 14:37 · Score: 4, Insightful
  
  How can it not get closer? AI didn't used to be able to beat professional StarCraft players. Now they can. That is, by definition, moving closer, because it's not moving backwards, and clearly an improvement from not moving at all.
  
  This is just more goalpost shifting, finding nonsense reasons to argue why it "doesn't count". Consider the alternative that maybe the things humans do aren't as as clever as we tell ourselves. If it's just "good micro", why can't humans use "good micro" to beat the AI, if we're so great?
  
  --
  Those who do not learn from commit history are doomed to regress it.
7. Re:Still Cheating by djinn6 · 2019-01-24 15:07 · Score: 4, Insightful
  
  How can it not get closer? AI didn't used to be able to beat professional StarCraft players. Now they can. That is, by definition, moving closer, because it's not moving backwards, and clearly an improvement from not moving at all.
  It is not moving at all. There's no viable route from this to general purpose AI. By playing countless games against itself on the same map, it is still performing a search on a decision tree, weighing the (now more fuzzy) nodes. It did not acquire any actual understanding of Starcraft mechanics. It does not logically reason about anything that it hasn't seen before. If you give it Warcraft instead, it'll take another several months of work from a team of very intelligent humans to make it good at it. In fact, I'll bet a big enough balance patch will cause it to have to throw out everything it's learned.
  
  This is just more goalpost shifting, finding nonsense reasons to argue why it "doesn't count". Consider the alternative that maybe the things humans do aren't as as clever as we tell ourselves.
  The goal post has always been to replace human intelligence. I don't see any AI building Starcraft-playing AI's, or discussing how long it will be before they are replaced by even better AIs.
  
  If it's just "good micro", why can't humans use "good micro" to beat the AI, if we're so great?
  Because humans have muscles that take time to move?
8. Re:Still Cheating by rtb61 · 2019-01-24 15:38 · Score: 1
  
  That is actually two advantages and not just one, see the map, and select units in those zones. That is where is wins hands or instant super clicky fingers down. Starcraft was always less about strategy and more about unit selection and unit movement, clickfest.
  
  --
  Chaos - everything, everywhere, everywhen
9. Re:Still Cheating by Anonymous Coward · 2019-01-24 15:41 · Score: 0
  
  >The goal post has always been to replace human intelligence.
  I think the goal here is problem solving.
10. Re:Still Cheating by The+Evil+Atheist · 2019-01-24 15:44 · Score: 2, Interesting
  
  There's no viable route from this to general purpose AI
  How do you know? Is there a roadmap that says all AI must progress in this way? If there were such a roadmap, why do we still need research then? Do you even understand why we research things we don't have adequate knowledge/experience?
  
  It did not acquire any actual understanding of Starcraft mechanics
  If it beat professional human players, then yes it did acquire an actual understanding of Starcraft mechanics. In fact, a better understanding than humans.
  
  If you give it Warcraft instead, it'll take another several months of work from a team of very intelligent humans to make it good at it. In fact, I'll bet a big enough balance patch will cause it to have to throw out everything it's learned.
  So what? The fact it wasn't even able to do something like Starcraft before. Do you not understand that technological progress isn't linear? How long ago were people saying AI can't play Go. And how long after beating humans at Go that people like you were saying AI can't beat humans at things that require imperfect knowledge? Do you really think DeepMind started from scratch after conquering Go, or maybe they used their experience
  
  The goal post has always been to replace human intelligence. I don't see any AI building Starcraft-playing AI's, or discussing how long it will be before they are replaced by even better AIs.
  Uh no. Nice attempt at rewriting history, but there have been clear intermediate goalposts that people like you keep walking back on. Chess. Go. Poker. Starcraft. The list keeps building.
  
  So what if you don't see AI building Starcraft-playing AIs? Your argument becomes more and more ridiculous. You claim there was absolutely no progress being made. And here, you reveal that your criteria of "progress" is replacing human intelligence. You realize you actually just shifted goalposts mid discussion, right? Now you're claiming any progress is no progress unless it replaces humans in one go.
  
  Because humans have muscles that take time to move?
  So? AlphaStar has its own limitations too.
  
  --
  Those who do not learn from commit history are doomed to regress it.
11. Re:Still Cheating by Anonymous Coward · 2019-01-24 16:06 · Score: 0
  
  Have you played or watch competitive Starcraft? Micro is an important aspect but it's a lot more than just being able to move your units better. You need to be able to scout, process that info to see what they are potentially building and then try and counter that yourself. e.g. if you see multiple air production facilities you will build up your anti-air etc.
  If the opponent is delaying production in order to mine more resources, then you could decide to match that or try and rush them. It's a lot more than just micro.
12. Re:Still Cheating by Anonymous Coward · 2019-01-24 17:07 · Score: 0
  
  Human intelligence developed from increasingly capable lesser brains that were originally solving very limited problems much simpler than what these new AIs are doing. It's true AIs are not good at introducing its own symbols and manipulating those symbols in a way grounded in a reality and that is useful for planning, but it seems almost certain that whenever that starts being something that AIs can do, it will be an outgrowth of the kind of reinforcement learning done here.
13. Re:Still Cheating by avandesande · 2019-01-24 17:53 · Score: 1
  
  They don't mention any limit on APM, imagine being able to micro every unit in your army real time? Starcraft is a bad example of AI beating humans.
  
  --
  love is just extroverted narcissism
14. Re:Still Cheating by timeOday · 2019-01-24 18:06 · Score: 2
  
  "The AI agent was hobbled in some ways. For example, it was restricted from performing more clicks per minute than a human" https://www.theverge.com/2019/...
15. Re:Still Cheating by sheramil · 2019-01-24 19:03 · Score: 2
  
  If it's just "good micro", why can't humans use "good micro" to beat the AI, if we're so great?
  The AI in Bullfrog's "Dungeon Keeper" used to drop a bunch of enemies in your territory and then immediately pick them up again, giving you no time to attack them. You had to keep on them in case it did let them go, but it was incredibly annoying, and only narrowly within the rules. The human player couldn't grab, say, thirty monsters at once, drop them and grab them again.
16. Re:Still Cheating by Anonymous Coward · 2019-01-24 19:16 · Score: 0
  
  Much like the dota bot from last year AlphaStar is effectively cheating as it is aware of the entire map at once, not restricted to the viewport as humans are.
  [citation needed]
  (Hint: Your statement is false. Why do you lie?)
17. Re:Still Cheating by Njovich · 2019-01-24 19:34 · Score: 2, Interesting
  
  The goal post has always been to replace human intelligence
  The goalpost in AI research has always been chess. After that was solved some made random other goals, but there is certainly no concensus there.
  Not sure what you mean by replacing human intelligence, it sounds like you are not happy until all humans are dead. Hopefully you are not an AI saying that. In the context of starcraft 2, this system has replaced human AI. If you want an AI system that can do everything in all fields that humans can do, well, it's not going to happen in our lifetimes, possibly never, and it's certainly not the definition of AI that AI researchers use.
18. Re:Still Cheating by Anonymous Coward · 2019-01-24 19:37 · Score: 0
  
  Aaaaand the clueless moron rises its ugly misguided opinion again.
19. Re:Still Cheating by phantomfive · 2019-01-24 19:51 · Score: 1
  
  In fact, I'll bet a big enough balance patch will cause it to have to throw out everything it's learned.
  A different map would throw out everything it learned.
  
  --
  "First they came for the slanderers and i said nothing."
20. Re:Still Cheating by phantomfive · 2019-01-24 19:55 · Score: 1
  
  Emphasis mine. Yes, it's an advantage, but it's not cheating. Humans can use the minimap to see what's going on as well.
  The minimap doesn't show you what units are in an area, it doesn't let you see their hitpoints. It doesn't let you click on the units: if you want to click on the unit, you have to move your screen there then click on it, that's two actions.
  
  It's cheating.
  
  --
  "First they came for the slanderers and i said nothing."
21. Re:Still Cheating by sg_oneill · 2019-01-24 19:59 · Score: 1
  
  By playing countless games against itself on the same map, it is still performing a search on a decision tree, weighing the (now more fuzzy) nodes.
  No it isn't. Its a multi-level (deep) Neural Net with various mechanisms to allow it to interface with the game. It doesn't have an inherent knowledge of the game. Its just trained on the game much in the way a child would.
  Its already surprisingly "general".
  
  --
  Excuse the Unicode crap in my posts. That's an apostrophe, and slashdot is busted.
22. Re: Still Cheating by Anonymous Coward · 2019-01-24 20:03 · Score: 0
  
  What people want to hear, though, is about human equivalent general purpose problem solving (putting aside the ethics of that). Being trained in a specific narrow domain isn't intelligence, its automation.
23. Re:Still Cheating by The+Evil+Atheist · 2019-01-24 20:26 · Score: 2
  
  And if anyone here bothered reading the article, you'd know this is not simply just dropping a bunch of enemies into your territory.
  
  --
  Those who do not learn from commit history are doomed to regress it.
24. Re:Still Cheating by phantomfive · 2019-01-24 20:51 · Score: 1
  
  Poor decisions in strategy (bad builds, attacking up ramps, attacking into arcs) combined with super-human micro. The blink micro from the computer was beautiful, and the ability to fight in four different positions on the map was something only a computer could do. The computer could micro perfectly on one side of the map while warping in units on the other side of the map.
  
  We've known for a long time that computers are better than humans at micro, that's not very interesting.
  
  --
  "First they came for the slanderers and i said nothing."
25. Re:Still Cheating by phantomfive · 2019-01-24 20:54 · Score: 1
  
  AlphaStar is using the same interface as other players.
  This is in no way true at all.
  
  --
  "First they came for the slanderers and i said nothing."
26. Re:Still Cheating by djinn6 · 2019-01-24 21:55 · Score: 3, Informative
  
  It did not acquire any actual understanding of Starcraft mechanics
  If it beat professional human players, then yes it did acquire an actual understanding of Starcraft mechanics. In fact, a better understanding than humans.
  That you can train a dog to bark twice when shown 1+1, and three times when shown 2+1 does not mean the dog can do arithmetic.
27. Re:Still Cheating by djinn6 · 2019-01-24 21:58 · Score: 1
  
  The goal post has always been to replace human intelligence
  The goalpost in AI research has always been chess.
  From Wikipedia:
  
  Modern AI research began in the mid 1950s.[20] The first generation of AI researchers was convinced that artificial general intelligence was possible and that it would exist in just a few decades. As AI pioneer Herbert A. Simon wrote in 1965: "machines will be capable, within twenty years, of doing any work a man can do."[21] Their predictions were the inspiration for Stanley Kubrick and Arthur C. Clarke's character HAL 9000, who accurately embodied what AI researchers believed they could create by the year 2001.
28. Re: Still Cheating by jovius · 2019-01-24 23:10 · Score: 1
  
  Thereâ(TM)s a great movie on AI from 1970 too: Colossus: The Forbin Project - https://www.imdb.com/title/tt0...
  It touches all the aspects of a super connected AI reaching singularity, and explores the humanityâ(TM)s response to it (or humanityâ(TM)s efforts trying to come to terms with something like that).
29. Re: Still Cheating by jovius · 2019-01-24 23:21 · Score: 1
  
  On the other hand, why should AI follow rules made for humans in the first place?
  Perhaps thereâ(TM)s lingering hope that humans should _feel_ equal to an AI.
  AI operates without feelings and human limitations as effectively as possible. So, an AI is subjected to a world thatâ(TM)s very limited. How is that fair, if itâ(TM)s done to only comfort us. Itâ(TM)s just a show and makes headlines.
  The focus seems to be to mimic a human, but AI can be something new. A completely new form of âoelifeâ or âoeconsciousnessâ can arise; unexpected strategies and new game mechanics.
  It would be much more interesting to see AI vs AI battles and competitions. Although they would probably have to be slowed down many times over for us.
30. Re:Still Cheating by religionofpeas · 2019-01-24 23:22 · Score: 2
  
  If you train a dog to beat humans in Starcraft, it means the dog understands Starcraft.
31. Re:Still Cheating by Anonymous Coward · 2019-01-25 00:47 · Score: 0
  
  How can it not get closer? AI didn't used to be able to beat professional StarCraft players. Now they can. That is, by definition, moving closer, because it's not moving backwards, and clearly an improvement from not moving at all.
  It is not moving at all. There's no viable route from this to general purpose AI. By playing countless games against itself on the same map, it is still performing a search on a decision tree, weighing the (now more fuzzy) nodes. It did not acquire any actual understanding of Starcraft mechanics. It does not logically reason about anything that it hasn't seen before. If you give it Warcraft instead, it'll take another several months of work from a team of very intelligent humans to make it good at it. In fact, I'll bet a big enough balance patch will cause it to have to throw out everything it's learned.
  
  This is just more goalpost shifting, finding nonsense reasons to argue why it "doesn't count". Consider the alternative that maybe the things humans do aren't as as clever as we tell ourselves.
  The goal post has always been to replace human intelligence. I don't see any AI building Starcraft-playing AI's, or discussing how long it will be before they are replaced by even better AIs.
  
  If it's just "good micro", why can't humans use "good micro" to beat the AI, if we're so great?
  Because humans have muscles that take time to move?
  yes, the only valid setup for the competition is an AI embodied, moving a robotic arms detached from the actual mouse and keyboards,
  and using vision to look at a real screen.
  Otherwise, using APIs is just unfair advantage.
32. Re:Still Cheating by Anonymous Coward · 2019-01-25 01:49 · Score: 0
  
  Starcraft was always less about strategy and more about unit selection and unit movement, clickfest.
  Pretty much the problem with any RTS.
  Whenever it isn't turn based you get this issue.
  It would be interesting to see it work with a turn based strategy game that isn't limited to fixed positioning.
  Something like tabletop Warhammer for example.
  Starcraft and other RTS games tends to favor exact mechanics. You get a significant advantage with precise clicking and the way the economy works in those games means that the advantage increases as time goes on.
  With that said the discussion about this in the Starcraft subreddit is pretty interesting.
  There are some significant differences in how the AI plays compared to professional players in that the AI builds an excess of workers early on.
  This is usually not ideal since those resources could be spent better on other stuff.
  It seems like it has incorporated an assumption that the opponent will destroy some of those workers and builds an excess to compensate for this and it appears as if it paid off.
  We might actually see professional Starcraft players mimic this style.
  Does anyone know if the Chess and Go AI games led to players learning anything from it?
33. Re: Still Cheating by Anonymous Coward · 2019-01-25 02:04 · Score: 0
  
  Yes, precisely. Computers have long been good at winning video games, it is not a demonstration of 'intelligence'. Yawn.
34. Re:Still Cheating by Anonymous Coward · 2019-01-25 02:31 · Score: 0
  
  If you train a dog to beat humans in Starcraft, it means the dog understands Starcraft.
  Okay, let us test your hypothesis.
  Put AlphaStar in a match with another professional StarCraft II player. However, this time, do one of the following, or similar:
  1. Same map but a different player type
  2. Same player type but a different map
  3. Different map and different player type
  4. (optional) Maybe use a human opponent which ranks closer to top 5?
  Run through each possible permutation, using a randomly-selected map from a pool which does not include the single map the AI trained extensively on. Perhaps give the AI a couple or few run-throughs of the map to simulate human familiarity with the game over an extended period of time, but not the equivalent of 200 years (which is ludicrous when you intend to play against humans, even those who play professionally a LOT).
  Then, plop the AI down with a human opponent and we'll get a more realistic picture of how well the AI understands the game.
35. Re: Still Cheating by Anonymous Coward · 2019-01-25 02:32 · Score: 0
  
  On the other hand, why should AI follow rules made for humans in the first place?
  It's not like humans haven't tried to overcome those problems.
  There have been programs that allows the human player to zoom out further than the official clients allow to get a better view of the field.
  Players using such programs have been banned from the game since it gives them an unfair advantage.
  Ideally the AI should run on another computer and only get the video feed and send commands through a keyboard and mouse interface since humans are banned if they use other means to play the game.
  Another interesting thing to note here is that it wasn't the same AI used in all matches.
  One of them had the same camera limitation as humans have. In that game the human won.
36. Re: Still Cheating by Anonymous Coward · 2019-01-25 03:08 · Score: 0
  
  Give it a keyboard, monitor, and mouse instead of a custom brinary interface.
37. Re:Still Cheating by Anonymous Coward · 2019-01-25 03:08 · Score: 0
  
  That's not true. AlphaStar is using the same interface as other players. It's not simply multitask faster than the human players.
  It used a different interface in the game it lost compared to the earlier games.
  In the games it won it was zoomed out beyond what the regular interface allows for human players so that it could keep track of all its units and their vision at once.
  When it was limited to the zoomed in interface that humans are limited to it lost.
  It is still pretty impressive and it will be interesting to see how it plays when it gets good enough to play more than just one map and a single matchup.
  How long time does it take to train it? Is it likely that we will see a more "realistic" setup with it later this year?
38. Re:Still Cheating by LifesABeach · 2019-01-25 03:10 · Score: 1
  
  Interesting use of problem solving AI stuff. But can the DeepMind AI be made to respond to this one simple issue, "Given my limited resources, which Medicare program should I buy? And, could a DeepMind AI handle and process my Medicare transactions seamlessly?" But the gaming skills could be used as a trainging coach are good.
39. Re:Still Cheating by Anonymous Coward · 2019-01-25 03:28 · Score: 0
  
  doesn't get us closer to actual AI
  That's some first-rate "No-True-Scotsman" stuff.
40. Re:Still Cheating by phantomfive · 2019-01-25 05:23 · Score: 1
  
  Actually, the AI is restricted to the viewport
  At the end of the match they tested an AI that was restricted to the viewport, but it lost badly.
  
  it is using an interface into the game,
  
  It is not using the same interface that humans use. It can instantly see all of its units' health, and is able to fight simultaneously in three places on the map.
  
  --
  "First they came for the slanderers and i said nothing."
41. Re:Still Cheating by phantomfive · 2019-01-25 05:25 · Score: 1
  
  The goalpost in AI research has always been chess
  Go read a book by Marvin Minsky, you don't know history and your ignorance shows.
  
  --
  "First they came for the slanderers and i said nothing."
42. Re:Still Cheating by phantomfive · 2019-01-25 05:27 · Score: 1
  
  The agent that played the live game didn't have the benefit of the overhead camera
  The agent lost that match. It lost the match badly, despite having inhuman levels of micro ability.
  
  --
  "First they came for the slanderers and i said nothing."
43. Re:Still Cheating by phantomfive · 2019-01-25 05:30 · Score: 1
  
  It seems like the AI was unable to remember where the units were when they went out of sight.
  
  --
  "First they came for the slanderers and i said nothing."
44. Re:Still Cheating by Anonymous Coward · 2019-01-25 05:57 · Score: 0
  
  Yes, this is not real intelligence. But it doesn't matter.
  The fact is that it can perform a given task better than humans and previously this was not possible. If you wonder why it doesn't matter, consider where we are going to use the AI. We are going to put it into work. If the AI can do human work better than humans, it does not matter how smart the AI really is. Oh and before you try to list what can go wrong with this, I will gently remind that humans make similar mistakes all the time, much more than the AI ever could. And if we setup a feedback loop to the AI, it will keep learning from its own mistakes, unlike most humans.
  This might not get us any closer to general AI, but at least it has made it pretty obvious that tasks which humans thought to require intelligence, don't actually require it.
45. Re: Still Cheating by Anonymous Coward · 2019-01-25 06:23 · Score: 0
  
  Itâ(TM)s a question of representation. If you can reduce your Healthcare question to a StarCraft game then yes it could.
  But more likely StarCraft is like being a CEO of a large company
46. Re: Still Cheating by LifesABeach · 2019-01-25 06:33 · Score: 1
  
  I'm finding myself confused, I'm nodding my head in the affirmative to the comments of an A/C. I should really stop self medicating.
47. Re:Still Cheating by HeckRuler · 2019-01-25 08:12 · Score: 1
  
  The DoTA bot was certainly very restricted as a swath of the game was simply barred. Namely certain abilities that it couldn't figure out how to deal with. I think the random factor did it in. That was pretty bullshitty, but a good example of the extent of AI capabilities. Rather than gloating about how well it did, the journalists should have sold it more like "This is where AI is currently at". But tech journalists generally suck.
  This example, AlphaStar, is effectively cheating by limiting the map selection. That's not normal game-play. But it's a boring thing to learn. Rather than training it for 200 effective years, they'd just have to multiply that by 7, for 1400. And they'd have to retrain whenever they cycle maps. It's just more money and not that interesting in the field of AI.
  They still had Fog of War. You can't see what you don't scout. But it effectively cheating by having complete knowledge of all units it can legally see without having to deduce parameters from a screenshot like a human. Here, they explain it pretty well:
  
  During the matches against TLO and MaNa(The human players), AlphaStar interacted with the StarCraft game engine directly via its raw interface, meaning that it could observe the attributes of its own and its opponent’s visible units on the map directly, without having to move the camera - effectively playing with a zoomed out view of the game. In contrast, human players must explicitly manage an "economy of attention" to decide where to focus the camera. However, analysis of AlphaStar’s games suggests that it manages an implicit focus of attention. On average, agents “switched context” about 30 times per minute, similar to MaNa or TLO.
  Additionally, and subsequent to the matches, we developed a second version of AlphaStar. Like human players, this version of AlphaStar chooses when and where to move the camera, its perception is restricted to on-screen information, and action locations are restricted to its viewable region.
  And that second version failed hard. They throw out some "it only trained 7 days" and "we hope to study this more", but it looks like that's where the line in the sand is being drawn. This is where AI currently is. And it's impressive. There will forever be the knee-jerk reaction types that claim it's just a bunch of if-else gates or it's not real AI or it's just a novelty, but they're also the same ones that claimed this this game was way beyond the abilities of AI.
48. Re: Still Cheating by Anonymous Coward · 2019-01-25 08:38 · Score: 0
  
  And it had 200 years of game experience under its belt. Shit even the pros have at most 15.
49. Re: Still Cheating by Anonymous Coward · 2019-01-25 08:44 · Score: 0
  
  Read the paper. It says in there. When the "AI" could only view what the humans could, it lost badly.
  Once they opened up the viewport and gave it conteol of all its units it started winning.
50. Re:Still Cheating by The+Evil+Atheist · 2019-01-25 16:47 · Score: 1
  
  No, because you trained it to do two sums. You did not train it to do arithmetic.
  
  Your original complaint was that it did not develop an understanding of Starcraft mechanics. If it can play better than humans, it has demonstrably developed an understanding of Starcraft mechanics. Because your criteria was "Starcraft mechanics", not "all games ever".
  
  See? More implicit goalpost shifting when you don't even understand your own argument.
  
  --
  Those who do not learn from commit history are doomed to regress it.
51. Re:Still Cheating by The+Evil+Atheist · 2019-01-25 16:51 · Score: 1
  
  Yes, it does. What else would an understanding of Starcraft entail, but to beat other people at Starcraft?
  
  --
  Those who do not learn from commit history are doomed to regress it.
52. Re:Still Cheating by Anonymous Coward · 2019-01-27 14:42 · Score: 0
  
  Because humans have muscles that take time to move?
  So? AlphaStar has its own limitations too.
  Did you watch the video? Have you ever watched SCII pro tournaments? No SCII top players can do that kind of micro (blink stalker). Just look at the battles, at some points AlphaStar got 1500 APM (action per minute) during battles, no human can ever do that.
53. Re:Still Cheating by Luthair · 2019-01-28 07:17 · Score: 1
  
  Its perfect knowledge for the player - note I did not say it saw through the fog of war. If a unit peeks through the fog for a moment when a human does not have the viewport at that part of the map AlphaStar knows every detail but the human does not. AlphaStar is able to be aware of what all units are doing at all times, Hence perfect knowledge.
Only now? by Anonymous Coward · 2019-01-24 13:22 · Score: 0

Of all games, I expected computers to demolish players in Starcraft easily. Top players are required to keep their eye on a dozen locations across the map, manage multiple armies with dozens of units each (sometimes individually controlling each unit) and meticulously organize your bases, upgrades, buildings and gatherers. That level of multitasking is ideal for computers to manage.
1. Re:Only now? by psycho12345 · 2019-01-24 13:30 · Score: 1
  
  With Deepmind, they are using an interface into SC2, not directly inside the game as a script. The AI is also APM capped. So no, it can't simply out multitask the human.
2. Re: Only now? by Anonymous Coward · 2019-01-24 13:36 · Score: 1
  
  Lol ?
  Except to the computer doing something at one side of the map is the same as doing something on the other side of the map.
  So no, it's not the same.
  I will be more impressed still with a robot looking at a screen playing on keyboard and mouse.
3. Re: Only now? by Anonymous Coward · 2019-01-24 13:44 · Score: 0
  
  The frequency of computer "looking" at one side of the map, and then "looking" at another side of the map is roughly the same as in human play.
4. Re:Only now? by Chicken_Kickers · 2019-01-24 14:02 · Score: 2
  
  Still not a fair match unless Deepmind is connected to a robot arm moving and clicking a mouse.
5. Re: Only now? by Anonymous Coward · 2019-01-24 14:04 · Score: 0
  
  in the 4th game vs Mana, it was pretty clearly playing at much higher effective apm than any human, with near perfect micro of 3 different groups of units in different areas. In the live game (where it had to control the camera), it seemed to be playing more like a human.
6. Re:Only now? by Anonymous Coward · 2019-01-24 14:09 · Score: 0
  
  Would it be fair if APM was throttled at 50% human, but still no real robot arm present?
7. Re:Only now? by Lanthanide · 2019-01-24 14:18 · Score: 1
  
  Yes, but which units should you use to counter your enemy? When should you scout? When you see your enemy has building X at time Y and location Z, what is your best response?
  You can do a lot with brute forcing, which is the sort of play style you're talking about, but if your AI can't do reasonably well at most of the above components, then it will still be able to be beaten by a pro player.
8. Re:Only now? by Anonymous Coward · 2019-01-24 15:22 · Score: 0
  
  I agree. if the AI had an information delay, input delay and input randomization - i.e. not only imperfect information, but slower information and the ability to misclick or react too late. I think this is a similar situation to say, using a cannon to pitch in baseball. Sure we cant pitch as well as a cannon and yes cannon technology is making leaps and bounds, but the cannon is not intelligent and also isn't that interesting to watch.
  If in those games the AI had displayed superior strategy and interesting techniques I would be more impressed - however it was doing things that are easily identifiable as incorrect even for a computer - like probes running around, killing its own units and cancelling buildings all for no gain whatsoever.
  We have a LONGGGGGG way to go for any actual intelligence but these new modelling tools are interesting. It is just annoying people call this AI when it is more like a pivot table or pie chart - simplifying a complex problem with lots of information down to an equation that gives a rough estimate of the answer.
9. Re:Only now? by yes-but-no · 2019-01-24 18:13 · Score: 1
  
  I believe any strategy or interesting technique is bound to be some equation at some level. Just we use different names - if it's easy to see the algorithm we call it simple bruteforce math; if we can't grasp the whole algorithm, we call it strategy. At the end it's all math computation. Unless of course if we name intelligence as something outside the grasp of math (things like insight/intuition/faith(?))
10. Re:Only now? by Anonymous Coward · 2019-01-25 03:30 · Score: 0
  
  Well, the overproducing of probes in early game was a bit interesting and might be something human players can use for their own games.
  It is simply a build order that relies on the opponent doing worker harass and you build enough probes that the expected loss doesn't hurt you.
  OTOH it might lead to the opponent noticing that you have an excess of worker and focus on your units instead since the excess of probes then might hurt you.
11. Re:Only now? by phantomfive · 2019-01-25 05:29 · Score: 1
  
  So no, it can't simply out multitask the human.
  That's the only way it won. It was able to notice when its stalkers were low in health and pull them back at the perfect time. It was able to accurately micro in battles at three or four locations.
  
  The strategies and decision making (continually sending its army up a guarded ramp in a choke point, for example) were poor, but it was able to make up for it by using superior micro.
  
  --
  "First they came for the slanderers and i said nothing."
Yawn by Anonymous Coward · 2019-01-24 13:28 · Score: 0

In other words, okay. Now what?
No need for AI, just a spell checker by Anonymous Coward · 2019-01-24 13:28 · Score: 0

Starcarft my hairy ass
Myopic AI by Anonymous Coward · 2019-01-24 13:40 · Score: 0

AI played only on a single map, and using a single kind of player
So it's an AI that plays StarCraft II for Dummies.
Still undefeated against turtles by Anonymous Coward · 2019-01-24 13:57 · Score: 0

It's not fair for those of us clickers who like to take the time to make an elegant, symmetrical, efficient base with lots and lots of defenses.
Meaningless by Rick+Schumann · 2019-01-24 14:18 · Score: 1

Nothing to see here..
1. Re:Meaningless by avandesande · 2019-01-24 17:46 · Score: 1
  
  Yup APM spam FTW
  
  --
  love is just extroverted narcissism
2. Re:Meaningless by phantomfive · 2019-01-24 21:11 · Score: 1
  
  It wasn't APM spam, the APM was within a normal range, but the multi-tasking was super-human. Able to micro simultaneously in three different places on the map, able to know exactly how much health each of your units have (you know the exact moment you should blink black your stalker, for example).
  
  --
  "First they came for the slanderers and i said nothing."
3. Re:Meaningless by HeckRuler · 2019-01-25 08:25 · Score: 1
  
  Nope. APM was lower than the human players:
  
  In its games against TLO and MaNa, AlphaStar had an average APM of around 280, significantly lower than the professional players, although its actions may be more precise. This lower APM is, in part, because AlphaStar starts its training using replays and thus mimics the way humans play the game. Additionally, AlphaStar reacts with a delay between observation and action of 350ms on average.
  Check out that chart. AlphaStar is at a mean APM of 277, TLO is at 390, and MaNa is at 678 because apparently he just never stops clickig shit.
  I don't understand why some people just hate AI and try to discredit and dismiss all advancements? Is it just natural skepticism? I'm all for that, just.... try to put in a little more effort and stop spouting bullsht.
4. Re:Meaningless by avandesande · 2019-01-25 11:54 · Score: 1
  
  APM reaches 1500 per minute in this match which is 3x what the best gamers can do. Gamers tend to establish a rhythm by maintaining APM even if they don't need to which skews results.
  https://www.youtube.com/watch?...
  
  --
  love is just extroverted narcissism
5. Re:Meaningless by HeckRuler · 2019-01-26 16:04 · Score: 1
  
  . . . They show you the whole histogram. Seriously, look at the bloody chart. TLO peaks at.... 2000 APM!? You can see the curve for the durations that TLOwas simply clicking shit faster than AlphaStar. AND you can see what slice Alpha-star issued commands faster than MaNa, who really didn't get much above 680.
  But yeah, poor TLO's fingers. Like... lay off the meth dude.
Great news by 110010001000 · 2019-01-24 14:57 · Score: 1, Interesting

Computers are good at playing computer games with a strict ruleset. Imagine that.
1. Re:Great news by The+Evil+Atheist · 2019-01-24 15:45 · Score: 1
  
  Are you saying Starcraft has a strict ruleset?
  
  --
  Those who do not learn from commit history are doomed to regress it.
2. Re:Great news by Ryanrule · 2019-01-24 16:01 · Score: 1
  
  reality has a strict ruleset.
3. Re:Great news by Anonymous Coward · 2019-01-24 18:40 · Score: 3, Interesting
  
  Are you saying it doesn't?
  In competitive StarCraft, you take into consider build orders where you need to have done a specific set of steps in a specific amount of time. If you fail to do it, it puts you behind the opponent and will cause you to lose the game.
  It's so meticulous that seasoned casters can actually tell a build based on how many workers a player has in a minute.
  Just look at the Liquipedia entry for Protoss build orders and you'll see what I mean. And that's just one race out of three.
The computer did not "win" by Anonymous Coward · 2019-01-24 15:24 · Score: 0

AI is completely incapable of achieving the most basic goal of any game it is faced with. It cannot have fun playing.
Bots Don't Have to Suck Anymore by mentil · 2019-01-24 16:51 · Score: 5, Interesting

Playing a multiplayer game with bots used to be seen as an inferior experience to playing with real humans. Now imagine that instead of something like AlphaStar's utility function being set to trying to win, it's set to trying to make the human opponents have the most fun. Of course it'd need some understanding of the mindset of the player; they might not want to always win, or always have close matches, or possibly they're a sore loser. However, this could be inferred somewhat by player behavior (even outside of the match proper, e.g. in menus).
Put that in a game and ship it, and that could be a killer feature. People might prefer to play with a bot that'll guarantee a fun time, over a human that might rage quit or be an unfair match that leads to a one-sided game.
#MakeGamesSinglePlayerAgain

--
Corruption is convincing someone that the selfless ideal is the same as their selfish ideal.
1. Re:Bots Don't Have to Suck Anymore by Kjella · 2019-01-24 20:00 · Score: 3, Interesting
  
  This is probably much harder than you think, I sometimes play chess against the computer and even though it can match rating with human players the play is quite different. Like instead of making "reasonable" mistakes and miscalculations it's making random unmotivated moves that score poorly. Likewise, it rarely has any idea what a would be a good trap for humans, instead it's just surprisingly shallow at times as if it's maxed out the ply depth for that rating. And that's in a game that has so incredibly little nuance like chess, I can't even imagine the complexity of trying to act like a plausible rookie in Starcraft II.
  
  --
  Live today, because you never know what tomorrow brings
2. Re:Bots Don't Have to Suck Anymore by mentil · 2019-01-24 20:52 · Score: 3, Interesting
  
  For a turn-based game, particularly when you can see what the artificial stupidity is doing, you're right. In a realtime game like Starcraft 2 there's so much that could be done, that doing nothing with a particular thing at a given moment is a reasonable move. If you want to make the suspension of disbelief better, you can teach it what mistakes humans make, and allow it to repeat them when desired. AFAIK teaching a deep-learning algorithm how to fail like a human has received little research.
  
  --
  Corruption is convincing someone that the selfless ideal is the same as their selfish ideal.
3. Re:Bots Don't Have to Suck Anymore by Impy+the+Impiuos+Imp · 2019-01-25 05:37 · Score: 1
  
  Playing a multiplayer game with bots used to be seen as an inferior experience to playing with real humans.
  Bots can cheat and destroy players by spying on what they do and building to counter it. Most AI in games is deliberately wimpy.
  Look at how MMORPGs have been stuck in the taunt/heal/DPS model for 20 years. To break that you have to get rid of taunts, and therefore tanks, and rely more on casting controllers and other things.
  Which apparently people don't wanna do because it's too hard.
  So boss fights remain the same old thing with a few extra dance moves to keep things mildly interesting.
  
  --
  (-1: Post disagrees with my already-settled worldview) is not a valid mod option.
4. Re:Bots Don't Have to Suck Anymore by philmarcracken · 2019-01-25 10:31 · Score: 1
  
  All thats left is to give them the copious chat logs of reports from players in multiplayer and let them learn how to insult us properly. Getting told to drink bleach has never felt so real!
Starcarft? by backslashdot · 2019-01-24 17:13 · Score: 2

Nobody notice the typo in the title? Wtf is Starcarft?
1. Re:Starcarft? by hcs_$reboot · 2019-01-24 18:36 · Score: 1
  
  It's Starcraft. Was not obvious maybe.
  
  --
  Slashdot, fix the reply notifications... You won't get away with it...
Not top rated by Anonymous Coward · 2019-01-24 20:30 · Score: 0

TLO is far from top rated. Even Mana has been slipping.
StarCarft? by The123king · 2019-01-24 20:57 · Score: 2

Is that the latest racing game from Blizzard?

--
If you gave me a choice between a printer and a giraffe with explosive diarrhoea, i'll get my ladder and my raincoat
DeepMind had dedicated "micro" networks by martyros · 2019-01-25 00:14 · Score: 1

This wasn't covered in the video, but in the DeepMind Blog about the match, they link to a paper describing a custom network architecture specifically designed to do "micro" during a battle, where each individual unit is acting as its own miniature agent. From the paper:

In this paper, we focus on the problem of micromanagement in StarCraft, which refers to the low-level control of individual units’ positioning and attack commands as they fight enemies. This task is naturally represented as a multi-agent system, where each StarCraft unit is replaced by a decentralised controller. We consider several scenarios with symmetric teams formed of: 3 marines (3m), 5 marines (5m), 5 wraiths (5w), or 2 dragoons with 3 zealots (2d 3z).
There's no way any human can get their "micro" to the level where they're calculating optimal behavior for individual units on the battlefield.

--
TCP: Why the Internet is full of SYN.
1. Re:DeepMind had dedicated "micro" networks by religionofpeas · 2019-01-25 00:49 · Score: 1
  
  There's no way any human can get their "micro" to the level where they're calculating optimal behavior for individual units on the battlefield.
  Sure, but there's no reason why the AI should mimic all the deficiencies of the human brain.
2. Re:DeepMind had dedicated "micro" networks by Anonymous Coward · 2019-01-25 02:42 · Score: 0
  
  Depends on what you are trying to prove with the demonstration.
  You can make a much dumber AI without any form of neural nets that will beat any human by just "clicking" faster.
  If the point is to show that the AI is better at making strategic decisions, then yes, you must limit it so that it works under the same constraints as a human.
  Otherwise you only prove that it can capitalize on a physical advantage it was created with to beat someone who doesn't have that advantage.
  We already know that we can build robots that are stronger and faster than humans. There is no need to prove that again.
3. Re:DeepMind had dedicated "micro" networks by ranton · 2019-01-25 06:38 · Score: 1
  
  If the point is to show that the AI is better at making strategic decisions, then yes, you must limit it so that it works under the same constraints as a human.
  Not necessarily. Even two humans generally approach strategic decisions in different ways. Some may focus on high level concepts while others focus on certain details and work their way up. Showing that an AI can make strategic decisions is far more important than the AI making those decisions in the same way a human would. The decisions being made are what is important.
  Anyone bringing up general AI in these discussions just likes hearing themselves talk (or type in this case). It is a meaningless argumentative discussion. I doubt anyone working on Alphastar would say their goal is to get closer to general AI. Their goal is more likely to push the boundaries of what decisions we can start to offload to AI, and are using games to do it.
  
  --
  -- All that is necessary for the triumph of evil is that good men do nothing. -- Edmund Burke
Machine Learning is Not Human-like Intelligence by Slicker · 2019-01-25 02:57 · Score: 0

Calculators also calculate far faster and more accurately than do humans. Still, it's a machine -- not a mind.
The difference between a machine and a mind is that a machine is driven by a set of mechanical/logical rules strictly, without exception. In contrast, a mind is driven by free will judgements. Free will is the ability to derive options, weigh them between each other, and select the option with the highest sum of desirability and likelihood.
Both can be programmed (after all, our minds are the result of biological machines). However, anything that follows any set of rules will always, in the natural world, come to usurp the intent of those rules. But a mind creates its own intents -- accomplishment of what is desirable in the options it chooses, such as fulfillment of hunger, pleasure, avoidance of pain, etc. A safe mind (what I would call a "person") would have its highest values: mutual freedom and well-being of persons.
Logic is illogical in its own terms, as proven by Godel's Incompleteness Theorems. In fact, logic is not fundamental in the universe. Looking through microscopes or telescopes, the smallest and largest things are arranged in oscillation patterns -- like linear math, not logic. Logical conditions seems to exist only in relative perspectives, which explains why it is always inherently incomplete. For example, whether the Moon is moving toward us or away from us, one direction or the other. Truths are always relative to the observer.
1. Re:Machine Learning is Not Human-like Intelligence by HeckRuler · 2019-01-25 08:20 · Score: 1
  
  a machine is driven by a set of mechanical/logical rules strictly,
  No, there's actually a large random factor.
  Your brain, likewise, is driven by a set of interconnected neurons with interactions dictated by electrochemical rules. It is not strict. There's a lot of... squishiness about when a neuron fires and how it fires. Likewise, there's a lot of squishiness with how weights in an artificial neural network get updated. That's thanks to the glory of rand().
  
  However, anything that follows any set of rules will always, in the natural world, come to usurp the intent of those rules.
  Haha, ok you little rebel.
  
  But a mind creates its own intents
  
  You should ask yourself why you like sex and if you don't think you've been pre-programmed for that, I don't know what to tell you.
  
  Truths are always relative to the observer.
  Heresy.
xkcd by Anonymous Coward · 2019-01-25 05:43 · Score: 0

People who don't understand how fast AI is progressing, here is a comic from year 2012
https://xkcd.com/1002/
As you can see the "computers still lose to top humans" section has several games:
- Starcraft (AI beats humans nowadays)
- Poker (AI beats humans nowadays)
- Go (AI beats humans nowadays)
- Arimaa (I have not seen results, but I'm pretty sure Deepmind would have no problem with it as it is very similar to chess)
So it took 6 years for this comic to become completely outdated.
Re by Rikaan · 2019-01-27 22:13 · Score: 1

Thanks... i like it https://valentinstag.me/