How AI is making video games smarter

Did you know that Microsoft’s AI-powered personal productivity assistant Cortana was named after a character in Halo, its wildly popular video game franchise? The term “artificial intelligence” was coined in 1956, just two years before a nuclear physicist invented Tennis for Two, what is credited as the first video game, and AI and gaming have been intertwined ever since.

In fact, a recent paper from gaming experts at Andreessen Horowitz posits that generative AI — like ChatGPT and its various competitors — has the potential to transform gaming more radically than any other category of entertainment, delivering high-quality, highly iterative experiences in a fraction of the time and at a fraction of the cost of traditional game development today.

But first, a little background. At the risk of oversimplifying, there are two key types of AI in gaming. The first is game AI, or deterministic AI, where every output has been scripted as a response to an input — similar to decision trees — which constrains results to what has been preprogrammed.

Think Pac-Man eating up dots while being chased by ghosts, which are actually following set patterns, or Street Fighter, where you battle the computer in the guise of what are called non-playable characters. These NPCs feel lifelike, thanks to AI, but they are also following a script. The second is generative (or nondeterministic) AI, where the outputs are constantly learning and changing, and new content can be generated in response to user prompts. With AI advancements over the years, games can truly adapt based on every action the player takes.

Given their complexity, their calculated nature and the sheer amount of data they produce, games are an ideal platform for building and testing AI models. One breakthrough moment for AI in gaming occurred in March 2016, when the computer program AlphaGo, developed by Google’s AI company DeepMind, challenged top player Lee Sedol at the strategy board game Go. AlphaGo demonstrated that a machine could become intelligent enough to beat the human champion at an ancient, highly complex game.

Fast-forward to the present day, when generative AI can help tailor game experiences to the user’s abilities, spin up original virtual worlds, eliminate predictability in games and more — all enhancing gameplay. It’s a win-win not only for players but for game developers, who traditionally have had to make trade-offs between cost, quality and speed to market.

This is where generative AI opens up transformative possibilities, according to the Andreessen Horowitz piece. It can take up to a decade to build the immersive but relatively small worlds in some of the most realistic games on the market. Compare this with the far more massive game Microsoft Flight Simulator, spanning 197 million square miles of planet Earth, delivered in three years with the help of Blackshark.ai to generate a photorealistic 3D world from 2D satellite images.

AI is making gaming development more efficient in many ways, without compromising on quality.  Startups like Respeecher and Altered provide AI voice-augmenting technology. Other startups focus on simplifying the development of art assets for games. For example, Stability AI generates beautiful images in minutes or hours as opposed to weeks, and Scenario lets game developers create their own image generators trained on the specific style of their games.

In a recent Forbes interview, Bill Gates ranks AI as one of the most critical milestones in digital technology, alongside the PC, the graphical user interface of most modern operating systems, and the internet. About 3 billion people across all demographics — almost 40% of the world’s population — are estimated to have played games in 2022. With mobile devices, emerging technologies and engaging content fueling the gaming industry’s growth, the impacts of AI will be far-reaching. Game on.