Connect with us

Gaming

StarCraft II-playing AI AlphaStar takes out pros undefeated – TechCrunch

Published

on

Losing to the computer in StarCraft has been a tradition of mine since the first game came out in 1998. Of course, the built-in “AI” is trivial for serious players to beat, and for years researchers have attempted to replicate human strategy and skill in the latest version of the game. They’ve just made a huge leap with AlphaStar, which recently beat two leading pros 5-0.

The new system was created by DeepMind, and in many ways it’s very unlike what you might call a “traditional” StarCraft AI. The computer opponents you can select in the game are really pretty dumb — they have basic built-in strategies, and know in general how to attack and defend and how to progress down the tech tree. But they lack everything that makes a human player strong: adaptability, improvisation and imagination.

AlphaStar is different. It learned from watching humans play at first, but soon honed its skills by playing against facets of itself.

The first iterations watched replays of games to learn the basics of “micro” (i.e. controlling units effectively) and “macro” (i.e. game economy and long-term goals) strategy. With this knowledge it was able to beat the in-game computer opponents on their hardest setting 95 percent of the time. But as any pro will tell you, that’s child’s play. So the real work started here.

Hundreds of agents were spawned and pitted against each other.

Because StarCraft is such a complex game, it would be silly to think that there’s a single optimal strategy that works in all situations. So the machine learning agent was essentially split into hundreds of versions of itself, each given a slightly different task or strategy. One might attempt to achieve air superiority at all costs; another to focus on teching up; another to try various “cheese” attempts like worker rushes and the like. Some were even given strong agents as targets, caring about nothing else but beating an already successful strategy.

This family of agents fought and fought for hundreds of years of in-game time (undertaken in parallel, of course). Over time the various agents learned (and of course reported back) various stratagems, from simple things such as how to scatter units under an area-of-effect attack to complex multi-pronged offenses. Putting them all together produced the highly robust AlphaStar agent, with some 200 years of gameplay under its belt.

Most StarCraft II pros are well younger than 200, so that’s a bit of an unfair advantage. There’s also the fact that AlphaStar, in its original incarnation anyway, has two other major benefits.

First, it gets its information directly from the game engine, rather than having to observe the game screen — so it knows instantly that a unit is down to 20 HP without having to click on it. Second, it can (though it doesn’t always) perform far more “actions per minute” than a human, because it isn’t limited by fleshy hands and banks of buttons. APM is just one measure among many that determines the outcome of a match, but it can’t hurt to be able to command a guy 20 times in a second rather than two or three.

It’s worth noting here that AIs for micro control have existed for years, having demonstrated their prowess in the original StarCraft. It’s incredibly useful to be able to perfectly cycle out units in a firefight so none takes lethal damage, or to perfectly time movements so no attacker is idle, but the truth is good strategy beats good tactics pretty much every time. A good player can counter the perfect micro of an AI and take that valuable tool out of play.

AlphaStar was matched up against two pro players, MaNa and TLO of the highly competitive Team Liquid. It beat them both handily, and the pros seemed excited rather than depressed by the machine learning system’s skill. Here’s game 2 against MaNa:

In comments after the game series, MaNa said:

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. 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.

And TLO, who actually is a Zerg main but gamely played Protoss for the experiment:

I was surprised by how strong the agent was. AlphaStar takes well-known strategies and turns them on their head. The agent demonstrated strategies I hadn’t thought of before, which means there may still be new ways of playing the game that we haven’t fully explored yet.

You can get the replays of the matches here.

AlphaStar is inarguably a strong player, but there are some important caveats here. First, when they handicapped the agent by making it play like a human, in that it had to move the camera around, could only click on visible units, had a human-like delay on perception and so on, it was far less strong and in fact was beaten by MaNa. But that version, which perhaps may become the benchmark rather than its untethered cousin, is still under development, so for that and other reasons it was never going to be as strong.

AlphaStar only plays Protoss, and the most successful versions of itself used very micro-heavy units.

Most importantly, though, AlphaStar is still an extreme specialist. It only plays Protoss versus Protoss — probably has no idea what a Zerg looks like — with a single opponent, on a single map. As anyone who has played the game can tell you, the map and the races produce all kinds of variations, which massively complicate gameplay and strategy. In essence, AlphaStar is playing only a tiny fraction of the game — though admittedly many players also specialize like this.

That said, the groundwork of designing a self-training agent is the hard part — the actual training is a matter of time and computing power. If it’s 1v1v1 on Bloodbath maybe it’s stalker/zealot time, while if it’s 2v2 on a big map with lots of elevation, out come the air units. (Is it obvious I’m not up on my SC2 strats?)

The project continues and AlphaStar will grow stronger, naturally, but the team at DeepMind thinks that some of the basics of the system, for instance how it efficiently visualizes the rest of the game as a result of every move it makes, could be applied in many other areas where AIs must repeatedly make decisions that affect a complex and long-term series of outcomes.

Source link

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Gaming

Why Transformers now look like a big bunch of gears and car parts

Published

on

Enlarge / How did one of the rarest 911s end up becoming a Transformer?

Stef Schrader

“I didn’t know what car Mirage was going to be at first,” said Steven Caple Jr., director of Transformers: Rise of the Beasts. “Where I’m from, in Cleveland, Ohio, I’d never even been in a Porsche before,” he continued. “My actual first introduction to Porsche was Bad Boys I, so shout out to Michael Bay—that’s all I really had.”

Caple admitted in a panel during Austin’s South by Southwest festival that the star car of the beloved action film Bad Boys inspired him to make Mirage a classic Porsche in the upcoming film. Mirage is a bit of a rebel himself, and the callback to the classic buddy-cop movie just felt right.

Fortunately, extraterrestrial Autobots won’t be tempted to pull over in any sketchy places to debate the merits of in-car snacking, but this does mean they have bigger nemeses that necessitate transforming into giant robots to handle. It can be more complicated than you’d expect to make a cool Porsche into an Autobot film star, though—in fact, Porsche has a whole team that helps Hollywood studios get just the right car on the silver screen. Here’s how it all comes together.

Character development

It starts with a character. Filmmakers have a certain look and vibe in mind when a new Transformer is “cast,” so to speak. Mirage is a bad boy with an attitude, and the film, set in 1994, is meant to be a sequel to Bumblebee. That made Caple think of the 1994 911 Turbo from Bad Boys.

“I was born in the ’80s, and I was a kid in the ’90s… this is the era when I grew up,” Caple explained. “This movie is like a time capsule to me.”

“You get to ’94, and everything started to change—from the wardrobe to the culture to the music to the cars,” he continued. “You start to step away from square-bodied cars and say, ‘hello curves.'”

You probably have to be pretty into your Porsches to spot that this is a 3.8 RS and not a 911 Turbo.
Enlarge / You probably have to be pretty into your Porsches to spot that this is a 3.8 RS and not a 911 Turbo.

Stef Schrader

The “casting” choice of the 964-era 911—a car that was dramatically smoother and more streamlined than any 911 before it—is a callback for the current Transformers series, given that Bad Boys was Michael Bay’s feature-length directorial debut. Yet Mirage has always been portrayed as an upper-crust member of Autobot society, so it makes sense that the Transformers team picked an even rarer 964-generation Porsche to portray him: a 1993 911 Carrera RS 3.8.

“When I was designing the character, it started there,” Caple said. “I talked to Owen [Shively] and the team at Porsche and said… he’s going to be an outlaw. He’s going to be a rebel. Going to be flashy. Very confident, but smooth.”

That’s when Porsche suggested looking into the 911 Carrera RS 3.8.

The Carrera RS 3.8 uses the same wider body shape as Bad Boys‘ 911 Turbo, but it was a homologation specially produced to legalize the Carrera RSR race car with a host of lightweight parts and a hardcore aerodynamic package designed for track domination. Porsche only ever made 55 RS 3.8s, according to Total 911, making it an exceptionally rare ride. In other media and toys in the past, Mirage has been a Ferrari and a Formula 1 car, so an ultra-rare Porsche feels like a solid fit.

While many of us associate the Transformers series with the heavy use of CGI, the filmmakers still need to source real cars to use for many of the shots—and Porsche has a whole team dedicated to helping filmmakers place just the right car into film and television projects.

Owen Shively, from that early ideation conversation Caple mentioned, is the CEO of RTTM Agency, Porsche Cars North America’s exclusive representative when it comes to entertainment partnership requests like this. When Porsche needs someone to arrange a specific car for a new film or TV project, Shively’s agency is where they turn.

Continue Reading

Gaming

Link as MacGyver: Tears of the Kingdom demo shows new material-crafting abilities

Published

on

For a major game that was first announced nearly four years ago and is set to launch in less than two months, we’ve seen remarkably little gameplay footage from The Legend of Zelda: Tears of the Kingdom beyond some very sparse trailers. Nintendo set out to partially fix that today, releasing a new video in which Zelda producer Eiji Aonuma showed off some of Link’s new abilities in a guided 10-minute gameplay presentation.

The most impactful new ability on display was called “Fuse,” which lets Link put together two disparate objects to create a new one with a brand-new effect. In the simplest example of this, Aonuma fused together a basic tree branch (which breaks incredibly easily even during simple fights) with a rock, creating a makeshift hammer with a lot more power and durability.

Enlarge / Tree branch + boulder = makeshift hammer.

Unlike in Breath of the Wild, where Link had to hunt for the most powerful weapons, the focus here will be on creating those weapons from component parts, Aonuma said. Fusing a long stick with a pitchfork can give you a longer attack range, for instance, and fusing various materials to arrows can create useful side-effects like freezing powers or a homing capability.

Fusion can also be used more defensively; Aonuma attached a mushroom to his shield, leading to an enemy-blinding explosive dust cloud when he blocked an attack. “Even if you struggle with combat, you can take down enemies using similar methods,” Aonuma said.

Attaching a fan to some logs to make an ersatz airboat.
Enlarge / Attaching a fan to some logs to make an ersatz airboat.

The ability to put different objects together extends beyond combat, too. A new “ultrahand” ability (named in an apparent nod to an antique Japanese Nintendo toy) let Aonuma stick together three logs to craft a raft and add two self-powered fans to create a kind of airboat. Aonuma noted that the car and flying machine Link rode in previous Tears of the Kingdom trailers were similarly constructed from individual parts in the game world and not inserted wholesale by the developers.

Other abilities shown in the brief gameplay demonstration highlighted Tears of the Kingdom‘s verticality. An “Ascend” ability, for instance, lets Link climb right through any ceiling above his head to the floor above, a method that even works in caves or with small ridges overhanging from cliff faces.

To reach the “sky islands” floating high above Link, Aonuma also showed off a “Recall” ability to reverse the movement of a large asteroid-like block that had fallen from the sky, using it as a kind of elevator. Aonuma said this was just one of many ways to get to these far-off islands in the sky.

Link glides through the sky, getting a panoramic view before diving into a river on the ground.
Enlarge / Link glides through the sky, getting a panoramic view before diving into a river on the ground.

But enemies can also use wind-based weapons to blow Link off the edges of these sky islands, forcing him into a paratrooper-without-a-plane-style skydive. In the demo, Aonuma guided Link’s descent as the ground quickly approached, ending with a dramatic dive into a river.

Aonuma stressed that he would “run out of time” if he discussed all the new features in Tears of the Kingdom, which he said has now completed all development work. Based on what was shown today, though, it looks like the game should lead to the kind of player experimentation and clever puzzle solving that could give Tears of the Kingdom a long life with tinkerers and content creators well past the “end” of the game’s story.

Listing image by Nintendo

Continue Reading

Gaming

Wii U and 3DS eShops close down later today, risking hundreds of unique games

Published

on

Enlarge / Your last chance to buy individual games for the Wii U and 3DS—assuming you have credits in your wallet account—arrives tonight.

Only a few hours remain for anyone who wants to buy games from the eShops for Nintendo’s Wii U and 3DS.

As it promised more than a year ago, Nintendo is shutting down those digital storefronts tonight at 8 pm Eastern, after previously halting the ability to add eShop funds in May 2022. After today, you can still download or re-download any titles you’ve previously bought from those shops, though that can obviously change in the future. Closing the eShops means that roughly 1,000 digital-only games will no longer be accessible, according to research by VGC, including 335 Virtual Console games that aren’t available through the Nintendo Switch Online service.

These kinds of sweeping moves, while perhaps understandable from a business perspective, pose a serious danger to the preservation of many games in the systems’ libraries. As Ars detailed earlier this month, video game preservationists are hamstrung by laws and regulations around remote access to DRM-protected work, even if it’s kept by research-driven organizations. Nintendo is one of many organizations that, through the Entertainment Software Association, lobbies to prevent libraries from offering legal access to archived games.

Today’s shutdown also marks the end of Nintendo’s Virtual Console, which allowed for the purchase of individual games from Nintendo’s catalog without a subscription. The Console was killed on the original Wii in 2019, and Nintendo does not intend to offer it on the Switch, noting in an FAQ about the eShop shutdown that it currently has “no plans to offer classic content in other ways” (since removed, but archived here). It has, however, offered a website on which you can “Bring back your gaming memories” of 3DS and Wii U titles you’ve purchased and played.

The Completionist’s explanation of how he spent nearly a year buying every available 3DS and Wii U game before today’s shutdown.

If you’re wondering what it would take to buy every game on the eShop, somebody already did that. The Completionist, aka Jirard Khalil, downloaded 1.2TB of Wii U and 267GB of 3DS games, taking 328 days and costing $22,791. “Since the industry started, we run a daily risk of losing games forever,” Khalil tells viewers in the video. “That’s why this matters.” Khalil is donating the consoles and storage containing all the games to the Video Game History Foundation and will host a Preserved Play fundraiser for the foundation April 15-16 on his Twitch channel.

Continue Reading

Trending