Colored Keyboard

Al Research the Games: Deepmind Beats Nearly AllComers

When DeepMind, Google’s AI research outfit, set out to demonstrate its latest breakthrough, it had to confront an added twist: how do you set your robot free to play games on the internet without anyone realising they’re competing against it?

The company caused a stir when it announced that its AlphaGo AI had beaten a world-class player at the ancient Asian board game Go. A few months later, it beat the world number one player.

But for the deeply strategic real-time war game StarCraft II, it had a different goal: to reach “grandmaster” standard – putting it in the top 200 players worldwide – on the game’s public servers, building its ranking the same way any human player would. That meant being matched with a steadily improving cadre of other human players, and winning against them consistently enough to be promoted.

StarCraft may seem like an odd next step, for a team that has previously taken on chess and Go, but the game has some qualities that make it interesting to researchers. It’s real-time, with millions of possible actions each second, and a vastly more complex roster than the six pieces of chess. Most importantly, it features hidden information: for the first few minutes of the game, it’s impossible to even see what your opponent is doing, let alone work out what they’re planning.

Embed from Getty Images

That means strategies have to be flexible enough to account for surprises, and need to incorporate mind-games as well. There’s also an advantage in a community where even the best players in the world can be found playing each other online, ranked according to a very public algorithm, with a ton of data flying around.

Players were told the new AI, dubbed AlphaStar, would be online, and were given the option of opting-in to play it. In order to ensure it achieved its rank fairly, it had to play its games anonymously, so that opponents didn’t spend more effort trying to trick it or break it than they did trying to win.

“There was a bit of a meme where people started asking ‘are you AlphaStar’ to others,” said DeepMind’s David Silver, one of the company’s co-founders and a lead author on the Nature paper announcing the StarCraft II victory. “We had the policy to just not chat – other than wishing people good luck, and then ‘good game’.”

The need to remain anonymous did also turn the experience from a test of raw skill into a sort of “Turing test for video games”, said Silver’s colleague Oriol Vinyals. “AlphaStar needed to play like a good human, not like a superhuman.”

Embed from Getty Images

That meant taking a different approach from previous StarCraft AIs, which tended to lean on the abilities that only a computer could have. In a game where human competitors track their “actions per minute”, a professional-level player may hit three or four hundred, while some AIs were acting thousands or tens of thousands of times over a sixty-second period. At other times, AIs were given near omniscience, with all the information available over the entire map plugged into their systems at once.

“We really wanted to have an interface that we believe was reasonable from a capability standpoint,” says Vinyals. “So we added this notion of a camera view, which is very crucial for players to control where in the map they’re actually focusing on, and we also reduced the peak actions per minute, to 22 actions in a span of five seconds.” In other words, the AI is forced to play much more like a human.

All of which is moot if the AI gives itself away by, well, playing like a robot. Luckily, it doesn’t – quite. In the first series of matches played publicly, in January, AlphaStar did exhibit one slightly mechanistic behaviour, falling prey to an almost cartoonish tactic where its opponent, the human player MaNa, moved a unit into and out of its field of view, changing its behaviour each time. It worked for MaNa to eke out the only win the humans scored over those first 11 matches.

More interestingly, the AI did develop its own understanding of the best tactical play, occasionally differing from the generally accepted practice among pros. The intricacies are a bit specialist, but reinforce the idea that simply teaching an AI to perform a task to human level can improve our understanding of the work itself.

“AlphaStar has been an amazing experience,” Oriol says. “Not because we beat most humans. But it’s more like that we were able to see what some limitations might be, to inspire research that will come, hopefully in the next few months or years and decades. Picking harder and harder problems and trying to be very good at them has been clearly the way so far.”

Girl on Computer

Who is the Pixelbook Go For?

There’s no question that the Pixelbook Go is an impressive-looking device. Multiple reviewers have praised the product’s sleek build quality and carefully considered design, in addition to its unusual twelve-hour battery life. But for a laptop that starts at $650, it offers little in the way of features. Instead of the traditional, more powerful Windows or macOS operating systems, the Pixelbook Go runs Chrome OS, a platform designed to handle Google’s Chrome web browser and do little more. To its credit, the Pixelbook Go has the hardware chops to perform this task excellently, but its constrained featureset may leave prospective buyers skeptical of its practicality, particularly considering the expansive spate of options available at this price point.

Though the company is often praised for the build quality and design of its products, including its flagship Pixel line of smartphones, Google has been known to introduce consumer products that fail to take off in the competitive personal electronics market. Take, for instance, last year’s Pixel Slate, a ChromeOS tablet starting at $599 with a premium look and feel that offers even less functionality than the company’s laptops, especially without its optional $199 keyboard case or $99 Pixelbook Pen. Even though this expensive tablet runs Android apps, many are not optimized for the Pixel Slate, leading to an unreliable user experience when dealing with third-party software. The Pixel Slate supports split-screen multitasking, for instance, but many third-party apps are not yet compatible with this feature. While the product was nonetheless praised by reviewers for what it was, it was a commercial flop, and Google seems to have shifted its focus away from ChromeOS tablets towards laptops at least for the time being.

Embed from Getty Images

While the Go is significantly cheaper than Google’s premium version released earlier this year, its price point, at essentially twice the cost of Chromebooks by other manufacturers, puts the product into a class all of its own. For that additional price, consumers are treated to impressive hardware specs that ensure the task of browsing the web, even when using multiple tabs and accessing content-heavy sites, remains fast and smooth. Chrome on the Pixelbook Go runs about as well as it does on any other laptop on the market, with the exception of Google’s own, more expensive 2-in-1 hybrid Pixelbook. And the Go features a high-quality, 1080p display, which, combined with a battery life that lasts all day and then some, renders the device perfect for extended Netflix or Youtube binge sessions.

The problem with the Pixelbook Go is the existence of laptops running exactly the same software almost as well for half the price or less. For $299, Samsung’s take on the Chromebook concept features a display of the same resolution and a similar, attractive design, with specs that are likely more than adequate for the tasks one might seek to accomplish within the confines of the Chrome web browser. And for $100 less, Asus’s Chromebook C423 features a lower-resolution screen but can handle light web browsing with ease. Consumers willing to spend $650 on a laptop are likely better off purchasing one that features a full operating system like Windows or MacOS, rather than what is in essence little more than a stripped-down version of Android. At this price point, Windows laptops with similar specifications are abundant, including Microsoft’s own Surface Laptop, and while they may not share the attention to detail of the Pixelbook Go’s build quality, the boost in functionality their more sophisticated software environments offer more than makes up for it. 

With all that being said, I still wouldn’t say the Pixelbook Go looks like a bad product. It’s simply one with an unclear market. While people who buy the Go are likely to be satisfied with their decision, the fact of the matter is that any number of better value propositions exist at and below its price point. As such, though it appears to be an acceptable device with some impressive specifications, it’s hard to recommend the Pixelbook Go to anyone when considering its alternatives.


Featured image credit: