Google's Pet Memory Fails to Distinguish Individual Animals in Nest Camera Tests
AI-generated image · Bay Street Wire
A review of the Gemini for Home feature reveals that the AI struggles to identify specific pets by name, undermining promised smart home automations.
Google recently introduced a Pet Memory feature for Gemini for Home, designed to allow Nest cameras to recognize specific pets by name. As The Verge first reported, the feature is intended to move beyond general animal detection, allowing the smart home to adapt to specific pets through personalized notifications and automations.
In a two-week test conducted by The Verge's Jennifer Pattison Tuohy, the AI failed to distinguish between three different cats. Tuohy attempted to use the system to identify which cat wanted to enter the home, verify if all pets were inside before dark, and automate feeding based on which animal was at the feeder. However, the results showed that the system could not tell the difference between cats named Boone and Smokey. Tuohy reported that the Google Home system identified every cat seen by the Nest cameras as Smokey.
This development follows a progression of animal detection technology from Google. The Verge notes that the Nest Cam IQ, released in 2017, could differentiate between people and pets. Google subsequently expanded animal detection to other Nest cameras and later integrated AI-generated text descriptions to detail what the cameras observed without requiring the user to load a video clip.