Whisker's Litter-Robot 5 Pro Fails to Track Cats with AI

Pet technology company Whisker recently launched its high-end Litter-Robot 5 Pro for $899, alongside the budget-friendly Litter-Robot Evo for $599. The premium 5 Pro model is designed to automate waste management for up to five cats weighing up to 30 pounds. It features a dual-camera system, weight sensors, and a gas sensor, all powered by artificial intelligence. However, real-world testing reveals that the machine's advanced AI features, which require an additional $8-a-month Whisker Plus subscription, fail to perform reliably.

The core of the Litter-Robot 5 Pro's smart capabilities relies on Cat ID, a facial recognition system called PawPrint, and WasteID, which is meant to distinguish between liquid and solid waste. In practice, the computer vision system struggled to differentiate between two highly distinct cats, eventually merging their profiles and miscalculating their weights. For instance, a 13.4-pound cat was registered at a much lower weight after the system confused him with his smaller companion. Furthermore, the WasteID system frequently failed to identify waste types or log deposit weights, a problem Whisker attributed to faulty gas sensors and scale calibration.

These shortcomings make the premium price hard to justify when compared to Whisker's other offerings. The $699 Litter-Robot 4 and the standard Litter-Robot 5, which costs $100 less than the Pro, offer similar mechanical scooping capabilities without the unreliable camera features. Meanwhile, the entry-level $599 Evo model reduces costs but introduces significant tradeoffs, such as a smaller waste bin, no night light, no litter-level monitoring, and compatibility limits that restrict it to one or two cats under 19 pounds.

For smart home developers and AI practitioners, the failure of the Litter-Robot 5 Pro illustrates the difficulty of transitioning computer vision and machine learning models from controlled laboratory settings to unpredictable home environments. Relying on physical sensors to back up failing AI algorithms can create cascading data errors, such as incorrect health tracking. Until these algorithms mature, developers must ensure core hardware functionality remains robust, as consumers will not tolerate paying subscription fees for unproven predictive features.