Here’s your daily tech digest, by way of the DGiT Daily newsletter, for Friday, January 25, 2019.

AI can beat us at RTS games now

Artificial intelligence software DeepMind, which Google acquired back in 2014, has been feasting on humanity in complicated board games in recent years.

But mark the date: now it’s proven again that AI can beat humans, in even complex, fast, open-strategy games.

  • First, DeepMind set AI agent AlphaGo to play Go, the ancient Chinese board game, and deemed harder for AI to “win” than Chess, given the gameplay style.
  • AlphaGo, of course, beat the best human players, which was a Big Moment for computational advancement.
  • Now, DeepMind has shown its AI muscle in StarCraft II: Wings of Liberty, besting pro-players, including Grzegorz “MaNa” Komincz in a live match streamed on YouTube and Twitch yesterday.

From board games to live action:

  • A fast-paced, real-time game like Starcraft II is exponentially harder for computers to play than a board game. Time wasted thinking or misclicking is time handed to your opponent to outbuild or outmaneurver you.
  • The AI agent, like a muman, must juggle focusing on constructing its own base, while fending off the opponent, and calculating how to take down the enemy. Games take from minutes to hours, depending on the styles of play.
  • Anyone who’s played a real-time strategy (RTS) game, such as StarCraft, Command & Conquer, or Total Annihilation (my old-school favorite), knows how tricky it is.
  • This is a first victory for AI at such a complex situation.
  • After quietly training against itself, and then playing against a pro in Dario “TLO” Wünsch, the competition went public.
  • A demonstration event was streamed live with commentators, and regular comments from the pro-player MaNa who would chat through the decisions and reactions to the gameplay.
  • But as always, there are certain caveats.

The results:

  • AlphaStar, the name for DeepMind’s trained AI in this encounter, beat MaNa hands down – winning 10 straight games, before MaNa took one back in a small win for mankind.
  • Now, AlphaStar had some unique limitations. For example, it played just one race, known as Protoss, and challengers were only allowed to play Protoss as well. So it’s not quite AI vs all-comers, just yet.
  • Experts are discussing elements where AlphaStar was hindered, or given advantages. For example, AlphaStar was limited to a human-achievable number of actions per minute (APM). In a real-time strategy game, higher APM means better controls over the growing armies and resource-gatherers at your disposal, which usually means better results.
  • However, AlphaStar actually performed dramatically less APMs than MaNa, implying more efficiency and smarter decisions at each point.
  • Experts gushed more about DeepMind’s unique strategies, how well the AI could manage large numbers of units, with commentators talking about “hypnotic” levels of precise control.
  • Another controversial point is that AlphaStar was allowed to see the entire map during most games. In the one game where MaNa did win, the map vision for AlphaStar was limited. (DeepMind suggested that this didn’t matter too much, but the AI did control up to three armies in three different locations at once, impossible for humans)
  • Being able to see what your enemy is doing, rather than guessing, would normally be a huge advantage for human players.

Machine-learning gaming does it different:

  • I was impressed to see AlphaStar pull off advanced moves and different strategies across almost every game, using a very human style of gameplay I wouldn’t have expected,” said MaNa.
  • “I’ve realised how much my gameplay relies on forcing mistakes and being able to exploit human reactions, so this has put the game in a whole new light for me. We’re all excited to see what comes next.”
  • When MaNa won, he played into a weakness that the AI hadn’t yet figured out – rapidly warping units in and out of the back of AlphaStar’s base. In reaction, every time, AlphaStar, despite being in a better position overall, would turn its entire army around to deal with the threat. That allowed MaNa time to build up a big force and take down the AI. (Skip to about 2h:39m of the stream to watch.)
  • MaNa won that single game, figuring out an AI weakness.
  • But DeepMind will learn from that mistake, in theory.

The big picture is more than StarCraft:

  • DeepMind said that these AI advances that played out in StarCraft II are another sign of how benchmark achievements should be able to help humanity: “We believe that these advances … represent a step forward in our mission to create intelligent systems that will one day help us unlock novel solutions to some of the world’s most important and fundamental scientific problems.”
  • Read DeepMind’s detailed insight and report AlphaStar here.

Here’s everything else happening:

1. Everything we know about Moto G7 series, which will consist of four phones: Moto G7, Moto G7 Plus, Moto G7 Power, and Moto G7 Play. We know a lot, basically. (Android Authority)

2. Google ‘Coral’ spotted, but what is it? (Android Authority). It runs the next unreleased Android version, and with a Snapdragon 855 chipset… if it’s not the rumored Pixel 3 Lite, what is it?

3.  Also, Apple now has patents for its future devices to act as poisonous gas and carbon monoxide detectors (9to5Mac).

4. New Asus ROG RTX-powered laptops coming Jan 29 (CNET).

5. Report: iPhone XR bestselling 2018 model in US, iPhone XS worst (Android Authority).

6. As many as 11,543 Microsoft employees got swept up in a reply-all email apocalypse (Business Insider).

7. Also, a new report alleges 50% of Facebook accounts are very likely fake (

8. A new study supports the idea that some people are “Super Poopers.” You could be one of them (Mother Jones).

9. A HBO documentary on startup’s laboratory fraud Theranos debuted last night at Sundance Film Festival. “The Inventor: Out For Blood In Silicon Valley”: Horrifying optimism, a deluded founder, the cult of personality and red flags galore (TechCrunch).

10. Record-setting ice hole drilled in Antarctica: 2km down to better understand climate change (BBC).

DGiT Daily: Your Tech Resource

In case you don’t know, the DGiT Daily delivers a daily email that keeps you ahead of the curve for all tech news, opinions, and links to what’s going down in the planet’s most important field. You get all the context and insight you need, and all with a touch of fun, and the daily fun element that you miss missing from this

Sign up right here to get the full email with much more into your inbox.

Read comments