AI to Compete Against Human Tasters at French Blind Tasting Championship

On March 15, an intriguing event will take place alongside the French Blind Tasting Championship organized by La RVF. An artificial intelligence system will compete against professional wine tasters to identify a dozen wines. This experimental match could bring some surprises. Wine enthusiasts need not worry; machines are not yet replacing humans in blind tasting. However, the AI might challenge human tasters.

A dozen volunteer tasters, selected from the participants in the French Blind Tasting Championship, will face an AI for the first time. The AI, named DeepRed, is developed by the Lyon-based company M&Wine. It aims to identify the country of origin, appellation, main grape variety, vintage, and producer of various wines, including whites, reds, rosés, and sparkling wines. This is achieved through a sophisticated combination of data and artificial intelligence.

Philippe de Cantenac, the event organizer, explains that M&Wine approached him months ago to test their technology against professional tasters. He found the idea intriguing, although it remains experimental and is conducted alongside the main championship. M&Wine has arranged for a bailiff to collect a dozen wines from around the world for DeepRed to analyze. On March 15, the same wines will be presented blind to the tasters. After the tasting, the results will be compared with those of DeepRed, which will be delivered under seal by the bailiff.

The question remains whether the machine can match or surpass the human tasters. Théodore Tillement, director of M&Wine, clarifies that their goal is not to replace humans but to test the limits of their technology. DeepRed was showcased at the Consumer Electronics Show in Las Vegas, combining AI with research from the Institut Lumière Matière in Lyon and the Ingéniérie des Matériaux Polymères laboratory. The technology identifies wines using 40 mineral elements, creating a unique signature for each wine based on terroir and viticulture methods.

Like human tasters, the technology requires training to master wine tasting. M&Wine has analyzed over 35,000 wines from 50 countries, storing the data in a database. The AI then searches this database to find the best match for the analyzed wine, refining its results for accuracy. Currently, this technology is useful for wine traceability, helping traders verify the authenticity of wine in bottles. However, DeepRed is not yet infallible. It currently works only with recent vintages and struggles with blended wines. This blind tasting against skilled tasters will help improve the technology.

The event on March 15 will be a significant step in exploring the capabilities of AI in wine tasting.