Overview
- Meng Xiaoyi, a Hangzhou startup, has begun preorders for PettiChat, a clip‑on device that uses microphones, motion sensors and a mobile app to report pets’ vocalizations and behaviors as short human‑readable messages.
- The company says the system was trained over about two years on large proprietary datasets and uses Alibaba Cloud’s Qwen model, claiming roughly 94.6–95% accuracy and translations in about 1.2 seconds.
- PettiChat also offers bidirectional features and location tools, including GPS tracking and geofencing, and the firm has reported thousands of preorder orders and a planned commercial launch later this month.
- Animal behavior specialists warn there is no independent peer‑reviewed validation of the accuracy claims, and they stress such systems are more likely to detect broad states like stress or hunger rather than produce literal, reliable human sentences.
- The product joins a fast‑growing pet‑tech market that could expand monitoring and preventive care but raises practical questions about dataset bias, generalization across breeds and the privacy implications of continuous audio and GPS data collection.