Google's Pet Memory Fails to Distinguish Between Pets

Google's Pet Memory feature was tested by a smart home reviewer who has multiple pets. The feature is designed to teach Google Home who their pets are and allow connected Nest cameras to identify individual animals, rather than just alerting to motion.

The reviewer has three cats, and after two weeks of testing, the AI still couldn't tell them apart. The reviewer set up a few indoor Nest Cams in strategic spots, hoping to use Pet Memory to easily check which cats were inside. However, despite telling Google Home what their pets look like, it insisted that all three cats were 'Smokey', the first cat they told it about.

The reviewer tried giving more information - Boone is a tuxedo cat with white paws, Osa the kitten is a tabby cat, and Smokey is a large gray-and-white cat - but Google Home couldn't process the extra detail. The reviewer asked Google why the system wasn't working as intended, but the company hasn't responded.

The reviewer was also unable to use Pet Memory to make their Aqara Pet Feeder smarter, despite being able to set it up to trigger when an animal was detected by the camera. However, when they tried to automate the feeder to dispense food for specific cats, Google Home still couldn't tell Boone from Smokey.

The reviewer concludes that recognizing individual pets is harder than recognizing a pet at all, and that AI-powered cameras are not yet reliable enough to rely on for important tasks like tracking pet activity.