Google AlphaGo to the challenge of the StarCraft 2: esports God contention-AlphaGo, StarCraft 2, StarCraft 2-IT information
Conquered the chess practitioners, the Alpha Dog's next goal is to compete. Game makers in StarCraft 2, Blizzard producer WCS China final recognition, communicate with Google on the project of the man-machine war of StarCraft 2, Alpha Dog will continue to play. Although "Li Shishi" candidates has not been determined, Alpha Dog in go on the old chess project after the victory over humans, can continue in modern esports project "hung and beat" human topic to heat up rapidly. The industry believes that Li Shishi failure will not thwart human esports masters with confidence.
In go human-machine during the war, there have been rumours that Alpha Dog will challenge the next Star Trek 2, Li Shishi, after losing, the StarCraft 2 top players in contention desire stronger. Used to be called "Star Trek first" Korea Player Flash in public that he can beat the AI. Other StarCraft professionals "storm Zerg" Yellow also issued pronouncements on social networks, let him, he will show our victory.
In man-machine before the start of the war, once offered the 5:0 sweep, Li Shishi, was also confident. While in alpha dog show it by professional chess players generally undervalued after the force, "Li Shishi" Declaration of victory, seems to be a question mark.
The StarCraft 2 is a real-time strategy game, players need to collect resources, build a base, upgrade the development war machine and the eventual aim to detect and destroy enemy targets. And the game, playing Starcraft 2 needs to tell the layout, well versed in policy, which requires players to quickly and accurately controlled by keyboard and mouse on a computer screen the unit's hand-eye coordination, or simply to be Deft. Pro APM (number operations per minute) can be reached under 250-300.
Previously online for some of the StarCraft 2 player made 100 "dog" biting 20 tanks under the video. The so-called "dog" in the game "Zerg" lower-level classes, under normal circumstances, 20 tanks can kill almost 100 "small dog". However, in the video "the dogs" magically avoiding the tank shelling, rush into the tank at a minimal cost and win.
This is not cheating with a computer program so that "dog" automatically avoid shelling, but with up to 8000 + APM for every "little dog" for micro-manipulation to achieve. Video was a lament: Alpha dogs achieve such inverse control is too easy, people no hope of winning.
Blizzard and Google has not yet announced the specific contest rules, professional players to guess, to achieve fair competition, will limit the APM of the Alpha Dog, and human players are basically the same. Without physical limitations, it may not be hopeless.
Flash said, players of the time, they can see all their set up and seek countermeasures, but Star Trek, beginning to see only their own units, can not understand the layout of the opponent. In other words, if artificial intelligence and human players over the disk information peer, only through their own exploration, guess the judgment of the situation or experience, to lead their own action to determine how defensive and offensive. This "guessing in the dark", is not available through the Alpha Dog when you go under by exhaustion and judge return algorithm to solve.
The other hand, although some of the StarCraft 2 layouts and strategy, limited map also amounted to a Board, but the situation is constantly changing, never "moves". Therefore, Alpha Dog WINS, not like a computer, you should think like a real judge, it is more difficult for artificial intelligence, and more gold content of the test.
谷歌AlphaGo欲挑战《星际2》：电竞大神纷纷请战 - AlphaGo,星际争霸2,星际2 - IT资讯
征服了人类围棋界后，阿尔法狗的下一个目标是电竞。游戏厂商暴雪制作总监在《星际争霸2》WCS中国赛决赛上确认，正在与谷歌沟通关于电竞项目《星际2》的人机大战，阿尔法狗将继续出战。虽然“电竞李世石”人选还未确定，阿尔法狗在围棋这项古老的棋类项目上完胜人类后，能否在现代化的电竞项目上继续“吊打”人类的话题迅速升温。业界认为，李世石的失败并不会挫败人类电竞高手的信心。
在围棋人机大战期间，一直有传闻称阿尔法狗接下来会挑战《星际2》，在李世石落败后，《星际2》顶尖选手的请战欲望更强。曾经被称为“星际第一人”的韩国选手Flash公开表示自己可以击败人工智能。另一位《星际争霸》职业选手“暴风虫族”Yellow也在社交网络上发出豪言，让他上，他会展示人类的完胜。
在人机大战开始前，一度表示会以5∶0横扫的李世石，也曾是一样自信。而在阿尔法狗展示了它被职业棋手普遍低估的棋力之后，“电竞李世石”的必胜宣言，似乎要打个问号了。
《星际2》是一款即时战略游戏，玩家需要通过采集资源、建造基地、升级发展战争机器，最终完成侦查并摧毁敌人目标的目的。和围棋一样，玩《星际2》需要讲布局、懂策略，后者还需要玩家迅速而精准地通过键盘鼠标操控电脑屏幕上本方单位的手眼协调力，简而言之就是要手快。职业玩家的APM(每分钟操作指令数)可以达到250-300下。
此前网上有一段《星际2》玩家自制的用100只“小狗”啃下20辆坦克的视频。所谓“小狗”是游戏中“虫族”的低等级兵种，正常情况下，20辆坦克几乎可以秒杀100只“小狗”。然而视频中“狗群”神奇地避开了坦克的炮击，以极小的代价冲进坦克阵并胜出。
这并不是用电脑程序作弊让“小狗”自动规避炮击，而是用了高达8000+的APM对每一只“小狗”进行微操控实现的。当时视频一出就有人哀叹：阿尔法狗达到这样逆天的操控速度太轻松了，人类毫无胜算。
暴雪和谷歌还没有公布具体比赛规则，专业选手猜测，要做到公平竞争，一定会限制阿尔法狗的APM，做到和人类选手基本一致。人类如果没有生理条件上的限制，就未必毫无胜算。
Flash表示，下围棋的时候，彼此可以看到对手的落子，然后寻找对策，但是打星际，一开始只能看到自己的单位，无法实时了解对手的行动布局。也就是说，如果人工智能也和人类选手一样对盘面上的信息对等，就只能通过自己探索，或者以经验对局势猜测判断，来主导自己的行动，决定如何防御和进攻。这种“摸黑猜测”，是无法通过阿尔法狗下围棋时用的穷举加判断收益算法来解决的。
另一方面，《星际2》虽有一些成型的布局和策略，有限的地图也相当于棋盘，但是局面变化瞬息万变，从来都没有“棋谱”。因此，阿尔法狗要胜出，不能像一台电脑，应该像真人一样思考判断，这对人工智能是难度更大，也更具含金量的考验。