Uncle Microsoft's efforts to keep inventing the Torment Nexus proceed apace with a patent application for an in-game advertising system that might use machine learning to read your activities, and work out when you'll be least apoplectic about being suddenly served a giant MacDonalds pop-up.
The patent in question, “Contextually Aware Management of Interactive Software Application Access", was filed in April this year, and published on 3rd September by the United States Patent & Trademark Office. Briefly, it outlines methods for making in-game advertising more bearable by working out where it fits most naturally into the experience, and perhaps giving you some means of "gaming" it. There are already a welter of existing ad delivery systems that do things like this, particularly in the gacha-eat-gacha realm of free-to-play mobile gaming, but as a package, this does feel like a grubby new horizon in automated surveillance and monetisation.
Here's the full text of the patent (PDF), via Respawn First. It's a dense gobbet of legalese, but if you can weather the repeated "In some embodiments", there's quite a lot of detail. Broadly, the hypothetical ad delivery system would interrupt and suspend play while showing you ads, but would look to do so "when the context of the interactive software application indicates adverse occurrences are unlikely or not possible to occur", such as when you're in a safe location, and without creating a "negative use experience".
Under one implementation, every time you see a promotion you might earn an invisible or visible "credit" that delays the next advert. It's possible you could accumulate several credits over time, with players able to "view another promoted content to earn additional credits". The system might give you a sense of how long the next ad might last, or whether it's wise to watch an ad given the current situation in the game. Yes, we are talking about gamifying the experience of consuming ads, restructuring the experience of the game around them.
The system's secret sauce/stroke of evil genius would be its responsiveness and capacity for nuance. It might opt not to interrupt you when you're experiencing a "rare event" – "a point of excitement and/or curiosity", possibly associated with a "specific type of player". The exact timing of ads might correspond to "progression or events in the interactive software application, such as 5 rounds of gameplay, completing one level of a video game, completing 3 online matches, defeating a non-playable character (NPC), or completing a project or task".
The system might notice when you open a save screen or dialogue box, pause the game or end a round. It might pick up on a change of model textures indicating that you've taken damage, or equipped new gear. It might accumulate the capacity to distinguish between different versions of an object in games with different art styles (the patent authors note that trees in Call of Duty don't look like trees in Fortnite). All with a view to surgically isolating the exact frame in which it's least grating to have (e.g.) your Dragon Age: The Veilguard playtime interrupted so that Pharmbio Korea Co Ltd can tantalise you with a $40 discount on colonic irrigation.
Perhaps needless to say, all of this might involve some form of machine learning "to detect or identify events in the video information, audio information [or] game state data". The machine learning model could pick up on user interface notifications, then "refine over iterations to 'learn' when visual game events are correlated with the UI notification" - for example, it might start to associate health packs with health bars getting bigger.
With one eye possibly on streamers, the system might "learn" to identify and avoid serving adverts during "popular" events such as a miraculous jumping catch in a baseball game, the usage of a terrain exploit during a speedrun, a legendary loot drop, or a fighting game player parrying an opponent's immensely fiddly supermove.
سەرچاوەی فەرمی بە زمانی ئینگلیزی: https://www.rockpapershotgun.com/microsoft-file-patent-for-a-machine-learning-system-that-would-train-itself-on-your-playtime-to-work-out-when-youll-be-least-annoyed-by-in-game-ads
