Google has patented a new technology that should help developers test their games using AI. The company believes this system should significantly increase the speed and quality of the QA testing process.
A new patent, titled Efficient Gameplay Training For Artificial Intelligence, was filed last year and published on March 23 (via VeryAli Gaming). Google claims that modern games have become much bigger and more complex in terms of scope, so they require solutions that can automate the testing process.
According to the company, QA teams are no longer able to “scale with the complexity of modern games, leading to delayed launches and lower quality products.” That’s why Google has come up with the Gameplay Trainer (GT) system, which should allow developers to teach game characters how not just to win, but to make various complex decisions based on in-game data and in accordance with certain goals.
Describing possible embodiments of this technology, Google said GT would be able to determine areas of a game world in which a human player could get stuck. Another example is that the data model may determine wether enemies are too powerful or not powerful enough for certain gameplay sections.
A diagram illustrating an overview of an operational routine of the Gameplay Trainer system in accordance with one or more embodiments
“In certain embodiments, GT may emphasize rapid development of multiple gameplay data models for the gaming application, each associated with one or more distinct objectives in the game world,” the description reads.
So in-game characters can perform different actions based on a trained gameplay model, which can be modified for further adjustments. The model can also be improved by observing the actions of human players.
As is always the case with patents, it is unclear whether Google will ever use this technology in practice or help other companies implement it into their development pipelines.