Samsung Bespoke refrigerators receive major software update integrating Google Gemini AI

Samsung has released a substantial software update for its Bespoke refrigerator line that integrates Google's Gemini AI to significantly expand the appliance's cloud-based object recognition capabilities. This upgrade increases the database of identifiable food items from approximately 100 to more than 2,000, allowing the system to distinguish between specific brands and varieties, such as differentiating Diet Coke from Coke Zero.

The new functionality relies on combining Samsung's existing on-device processing with Google's cloud models, though this necessitates an active Wi-Fi connection for the features to operate. While the system now tracks expiration dates and identifies specific quantities, reviewers note that the accuracy is not always perfect, with occasional inconsistencies in tracking and instances where the algorithm may misidentify items based on incomplete packaging labels.

Beyond inventory management, the update introduces 'Reliability AI', a feature designed to monitor the refrigerator's components and assist in fault diagnosis. This capability allows support agents to remotely adjust settings or provide repair guidance without requiring a physical visit to the home. For instance, agents could remotely reduce water flow to an ice tray to resolve issues with stuck cubes, provided the user has granted express consent for repair personnel to access device health metrics.

Voice control capabilities have also been enhanced through the integration, allowing users to query device settings, filter replacement dates, and troubleshoot issues via voice commands using the existing Bixby interface. The system can now generate more detailed automated shopping lists based on inventory levels and suggest recipes using ingredients currently stored within the appliance.

Despite these improvements, the update retains some limitations characteristic of current AI models. The system has been observed to occasionally hallucinate, such as misidentifying a bandage as a vegetable, and may struggle with niche ingredients if the label text is partial. Furthermore, while the backend leverages Google's models, the user interface continues to rely on Samsung's Bixby assistant without explicit callouts to Gemini within the device itself.

Samsung has offered AI-powered features on Bespoke refrigerators for several years, but previous iterations were limited in scope and often required manual data entry. This latest release marks a significant step forward in delivering useful machine learning for smart home appliances, transforming the technology from a promising demonstration into a more practical tool for household management.