Earlier today, Cody, the least likeable writer here at NAG, tried to call me, everyone’s favourite, out with this little line:
Arielle says she’d definitely play the game again when the new mode launches. We’d just need to find her two friends somehow.
To which I have to say: Ha! I need no one! Google is the only friend I need!
For those of you who don’t understand what I’m even on about, Google is currently busy developing a DeepMind-based AI that can be the ultimate co-op gaming friend, and by that, I mean an instructible game-playing AI agent.
This news comes after Google presented a research article about a Scalable Instructable Multiworld Agent (SIMA) earlier this week.
In plain English, this is a learning AI that can follow verbal instructions and understand the virtual world it’s in.
Instead of just being another hard-coded AI opponent or bot, Google’s SIMA promises to be a more natural and human-like gaming companion.
Google has officially partnered with eight-game studios to test this model in games. Some games in this test include Valheim, Goat Simulator 3 and No Man’s Sky.
In its current state, SIMA can perform 600 basic skills, including navigation, object interaction and menu use. We’re all used to AI in games. From the NPCs littered around many of our favourites to those burdensome bots you end up saddled with in solo queue League.
However, SIMA will be the first time the AI doesn’t need source-code access or an API to function. Instead, it relies on just two inputs: the images on the screen and the instructions the user provides.