• 2predict, Inc.

Grab a Controller -- The Game’s about to Get Real

Updated: Jan 1

You may not be big on video games or participate in esports -- but AI in gaming is an excellent example of how deep machine learning can create innovation and enhance our lives in ways perhaps imagined, but never thought possible.

And whether or not you’ve dabbled in the gaming culture, you have to appreciate how drastically video games have evolved over the past few decades. Today, video games are fueling the rapidly growing esports market, which is expected to reach $1.79 billion by 2022.


Remember Pong? Even this primitive yet addictive game of digital ping-pong leverages rudimentary AI to keep you challenged. Programmed using pathfinding and decision tree techniques, Pong kept children of the1970s busy for hours chasing an unpredictable and pixelated dot across a boring, black screen. Fast forward to 2019, and games with high-end graphics and intricate game play scenarios are commonplace, if not expected (try out Red Dead Redemption II or Fortnight, and you’ll see what I mean).


But even these spectacular and realistic video games have yet to capitalize on the true power of AI and machine learning. In the not so distant future, AI stands to completely change the gaming experience, enhancing player performance, game design, in-game strategies and more. AI is even enabling the development of conversational assistants that help players strategize and improve as they play.


Let’s take a look at the primary uses of AI in gaming, as well as some of the possibilities developers have yet to implement.


AI Upscaling and Game Design

AI upscaling has empowered developers to bring classic video games like Doom, Final Fantasy and Grand Theft Auto up to date, to appeal to modern gamers who expect to see realistic graphics and responsiveness. Upscaling leverages machine learning: For example, machine learning algorithms can be trained to identify best in-game features of today’s popular games so developers can apply those learnings to updating the classics. By training AI models using video footage or a newer game’s content and art style, programmers can add new high-resolution art and details to older games, making them more marketable and engaging to gamers.


In addition to enhancing graphics, a focus on using AI in design has been making non-player characters (NPCs) more realistic. You may notice AI working when an NPC becomes “twitchy” or seemingly nervous as they’re trying to determine your location and next move. Such behavior is based on triggers that react to certain player actions or dialogue.


Adaptive Game Play

That brings us to a discussion of adaptive game play. Over the years, AI has become extremely good at playing video games and making them challenging, advancing to the point where it’s nearly impossible to defeat a supercomputer in a game of chess. But this is just a glimpse of what lay ahead. Once AI models the player’s behavior, it can activate the generation of new content and create new in-game experiences based on what the AI determines the player prefers.


Using a AI-based computing method called procedural generation, games can change the difficulty level in real time based on game play, to either deliver a greater challenge or reduce frustration -- all with the purpose of keeping players engaged. The popular game Minecraft leverages procedural generation to create endless possibilities for construction and landscaping in “Creative” mode.


Research in procedural generation has recently been focused on using generative adversarial networks (GANs), which are deep neural net architectures comprised of two nets contesting one with the other. GANs can be used to create content of the same type or style as existing content to generate new levels of play, making a game replayable -- or endlessly playable. Still, experts say procedural generation isn’t true AI -- it can’t generate new scenarios or characters, or enable content interaction.


Human-like Character Development

According to Nick Stat with The Verge, cutting-edge AI is starting to influence game development in real and dramatic ways. Beyond the generation of new levels and increasingly realistic imagery, AI will eventually enable true self-learning characters that have realistic personas and interact with you as a human actually would in real life.


Case in point: Siren, designed as a more effective, timesaving alternative for creating hyper realistic video game characters. Developed by Epic Games in a collaboration with CubicMotion, 3Lateral, Tencent and Vicon, this incredibly realistic virtual human was made possible by Epic’s Unreal Engine 4 technology.



Siren isn’t real, but she can sure fool you into thinking she is.


Some esports analytics platforms have even developed AI-powered coaching assistants. Used by professional gamers, these “coaches” can assess game stats and suggest better strategies, such as where to hide during game play or when to counter-attack an enemy. One example is in the game Overwatch, in which players get useful tips from Omnicoach about how to use weapons, improve mobility and protect themselves against enemy avatars. Other games that benefit from AI coaching include League of Legends, Counter Strike and StarCraft.


AI has great potential for improving the odds for Fantasy Football players, as well. For example ESPN and IBM developed Fantasy Insights with Watson, which is built on enterprise-grade machine learning models for reading, understanding and making predictions based on millions of data points from multiple sources. AI can be used to find the best, most predictive combination of stats for a player’s history, giving players a competitive advantage.


What’s Next for Gamers?

In 2017, Electronic Arts (EA) established Search for Extraordinary Experiences, or SEED, a new division of its R&D organization that leverages AI to explore new technologies and opportunities that will enable the games of the future. SEED’s first demo in 2018 featured the use of lighting techniques known as real-time raytracing, built with Microsoft’s Direct X Raytracing API and self-learning robots in a virtual factory.


And as experts such as EA’s SEED researchers continue to explore the possibilities of AI in gaming, gamers can expect to see new capabilities emerge, such as smart, human-like NPCs, the ability to predict human players’ behaviors with accuracy, and advanced levels of game personalization. AI can be used to model a human player to understand how real players interact with and experience a game -- maybe even how they feel while playing -- to further refine and improve future games.


At 2predict, we’re keeping a close eye on the application of AI and machine learning across industries.

CONTACT US

 

Let's work together and drive your business with next-level analytics solutions.

©2020  --  2predict, Inc.  --  Data Policy

  • Twitter
  • LinkedIn
  • YouTube