Niantic is using AI to make Peridot’s fake pets feel more real

Here’s how it works. Starting with today’s update, Niantic’s algorithm will analyze real-world objects captured in game with your phone’s camera and then convert those objects into words to be parsed by a large language model (LLM) — in this case, a customized version of Meta’s Llama 2 program.

Additionally, each pet, called Dots, will have their personality profile — which is similar to the different natures of pokémon — also fed to the LLM. Peridot will then ask the LLM how a dot with its specific personality type would interact with the objects around it. Its answers will be how Dots respond in the game.

According to Niantic, the use of AI will enable the game to make greater use of underused animation assets, ones that the developers had a difficult time programming situations to use them for. “Thanks to the new system leveraging generative AI, those animations are now part of a much greater number of possible reactions,” Niantic said in its press release.

Overall, Niantic hopes this use of AI will make the pets in Peridot more lifelike — but hopefully not too lifelike. There’s only so much cat puke I want to clean up in a day.