PCMag editors select and review productsindependently. If you buy through affiliate links, we may earn commissions, which help support ourtesting.

DeepMind AI的人类星际争霸2的球员s on Battle.net

For research purposes, the AlphaStar AI will partake in a 'small number of games' on Battle.net, which hosts StarCraft II, but only for a limited time.

ByMichael Kan

My Experience

2017年10月以来我一直与个人,覆盖a wide range of topics, including consumer electronics, cybersecurity, social media, networking, and gaming. Prior to working at PCMag, I was a foreign correspondent in Beijing for over five years, covering the tech scene in Asia.

Read Full Bio
StarCraft DeepMind

StarCraft II players will have a chance to prove that humans can beat a machine.

A cutting-edge AI from DeepMind will be playing the game anonymously in public matches. "As part of ongoing scientific research into artificial intelligence, players can now opt-in for the chance to be matched against experimental versions of DeepMind's StarCraft II agent, AlphaStar," the official Starcraft Twitter accountsaid(Opens in a new window)on Wednesday.

The AI will partake in a "small number of games" on Battle.net, which hosts StarCraft II, but only for a limited time. The matches will also occur on the competitive ladder in Europe.

In 2017, the AI from DeepMind famouslybeathumanity's top player in the Chinese game of Go.The Alphabet-owned company has now set its sight on getting the AI to master the more complex game of StarCraft, which involves mining resources and building units to attack and destroy the opposing player.

The result is AlphaStar. In December, the AIdominateda pair of top-ranked StarCraft players. However, the matches were restricted to playing against the AI on a single map and using only the Protoss alien race.

Blizzard said AlphaStar will nowtake on(Opens in a new window)human players on multiple maps using any of the three races in the game, the other two being the Zerg and the Terrans. However, the AI will only be playing one-on-one matches.

Gamers who face off against theartificial intelligencewill also receive no warning that their opponent is in fact a computer program."Having AlphaStar play anonymously helps ensure that it is a controlled test, so that the experimental versions of the agent experience gameplay as close to a normal 1v1 ladder match as possible," Blizzard's post explained. "It also helps ensure all games are played under the same conditions from match to match."

The results will be released in a peer-reviewed scientific paper, along with the replays of AlphaStar's matches. "If you would like the chance to help DeepMind with its research by matching against AlphaStar, you can opt in by clicking the 'opt-in' button on the in-game popup window. You can alter your opt-in selection at any time by using the 'DeepMind opt-in' button on the 1v1 Versus menu," Blizzard said.

The company has anFAQ(Opens in a new window)with more information about the matches and how to join them. DeepMind trained AlphaStar by first teaching it to emulate human players from previously recorded StarCraft matches. The company then came up with a novel way to refine the AI further; it created a whole league of AlphaStar AIs, and had them all battle against each other. The self-training enabled the artificial intelligence to develop some rarely seen strategies and continually improve them.

Still, AlphaStar's prowess has come under scrutiny, not all of it good. Some critics believe the AI hasunfairadvantages when competing against humans, like the ability to digest more raw computer data from the matches, when a person can only rely on what they see from game's graphics.

In response, Blizzard's post says AlphaStar will perceive gameplay using a "camera-like view" similar to how a human sees the game. "This means that AlphaStar doesn't receive information about its opponent unless it is within the camera's field of view, and it can only move units to locations within its view. All limits on AlphaStar's performance were designed in consultation with pro players," the company added.

AlphaStar will also be capped to an unspecified number actions per minute and per second to help ensure the AI competes on a level playing field.

PCMag LogoGoogle's AI DeepMind Learning How to Play Hearthstone, Magic: The Gathering

What's New Now<\/strong> to get our top stories delivered to your inbox every morning.","first_published_at":"2021-09-30T21:30:40.000000Z","published_at":"2022-08-31T18:35:24.000000Z","last_published_at":"2022-08-31T18:35:20.000000Z","created_at":null,"updated_at":"2022-08-31T18:35:24.000000Z"})" x-show="showEmailSignUp()" class="rounded bg-gray-lightest text-center md:px-32 md:py-8 p-4 mt-8 container-xs">

Get Our Best Stories!

Sign up forWhat's New Nowto get our top stories delivered to your inbox every morning.

This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to ourTerms of UseandPrivacy Policy. You may unsubscribe from the newsletters at any time.


那nks for signing up!

Your subscription has been confirmed. Keep an eye on your inbox!

Sign up for other newsletters

PCMag Stories You’ll Like

About Michael Kan

Senior Reporter

2017年10月以来我一直与个人,覆盖a wide range of topics, including consumer electronics, cybersecurity, social media, networking, and gaming. Prior to working at PCMag, I was a foreign correspondent in Beijing for over five years, covering the tech scene in Asia.

Read Michael's full bio

Read the latest from Michael Kan

Baidu
map