Welcome to TNW’s Beginner’s Guide to AI. This multi-part feature should give you a very basic understanding of what AI is, what it can do, and how it works. The guide contains articles on (in published order) neural networks, computer vision, natural language processing, algorithms and general artificial intelligence.
Among the most common misconceptions about machine learning technology is the idea that video games dating from the 1970s and 1980s had built-in “artificial intelligence” capable of interacting with a human user.
If you are curious but in a hurry, the “AI” video game, in the traditional sense of the term, is not what people in the modern age refer to when they talk about artificial intelligence. “Bots” in an online multiplayer game, enemies in a first-person shooter, and processor-controlled characters in old-school Nintendo games are not examples of artificial intelligence, they are as smart programming tips.
Artificial intelligence, in the form we’re discussing here at Neural, includes machine learning systems such as the core neural networks behind Alexa, Siri, and Google Assistant. Adobe uses AI to predict what you want a fix to look like, Google uses it to find you a cheap flight, and Twitter uses AI to determine which ads to serve you.
But, at the risk of confusing, video game developers often use AI to create video games. The Unreal Engine, for example, uses AI to enable real-time graphics rendering. AI is generally not used to control everything that interacts with the player, as that would generally be a poor solution to most programming issues that developers encounter.
When gamers think of “AI” in a game, they are probably not imagining a set of image recognition algorithms. They think of enemies controlled by the processor who can recognize player actions and react. We’ve seen enemies of the processor take cover, call for save, and react in kind when players use new tactics. Again, this is usually not done with artificial intelligence.
There’s not a lot that an agent can do in a video game. It is therefore generally more cost effective and simpler to simply code an agent to perform certain tasks than to train a neural network to control the agent.
Perhaps in the future, as games continue to grow in size and functionality, it will make sense to create AI-powered agents to explore video game worlds in tandem with gamers. One of the most popular forms of robust AI system development involves letting models get lost in the worlds of video games. StarCraft and Super Mario Bros. are some of the most popular game worlds for machine learning research.
But the purpose of such research has nothing to do with the development of video games. Researchers observe AI models in game worlds because they are often physics-based, which helps AI learn how the real world works.
While there are a few exceptions, we can generally assume any reference to “AI” in the game world that refers to CPU control over agents – i.e. enemy orcs in the game. Shadow of war or AI companions in Fallout 4 – is not a real artificial intelligence. While the developers of both games likely used AI for a myriad of functions in their creation, the games themselves lack specially designed AI to control NPCs, Agents, Monsters, Allies, or the villains.
Published August 10, 2020 – 21:16 UTC