Amazon has announced Project PI AI, which is a practical approach utilizing generative artificial intelligence technology.
The company explained that the Project PI AI system relies on artificial intelligence technology that combines generative AI techniques with computer vision capabilities, aiming to reduce the number of faulty products shipped to customers.
The project examines each product as it passes through the scanning tunnel, using advanced algorithms to detect issues like damaged packaging, color or size inaccuracies, and expired items.
Upon discovering a defect in the product, it is prevented from being shipped to the customer and returned to Amazon’s team for verification. A decision is then made on what to do with the product, such as reducing its price for resale through the Second Chance platform, donating it, or finding another use for it.
Project PI AI is used in several Amazon centers, where the company inspects millions of products passing through the scanning tunnels daily. The system examines each item moving through the tunnel searching for defects using various methods.
The optical character recognition model is used to verify the expiry dates of products, aiming to prevent sending any expired products to customers.
The computer vision models, trained on images of products from Amazon listings and other images sent to customers, analyze colors to ensure they match customer requests. The system also detects any obvious physical damage in the item, such as a twisted book cover or torn packaging.
By detecting defective products before shipping them, Amazon works on improving the customer experience by enhancing sustainability efforts through reducing returns and minimizing carbon emissions.
The Amazon’s Project PI AI helps in identifying the root causes of negative customer experiences.
When returning products, Amazon utilizes a large linguistic model that includes reviewing customer feedback and analyzing images captured by Project PI AI to understand the problem’s cause.
Project PI AI leverages its past experiences, providing Amazon selling partners with information about defective products to avoid repeating mistakes in the future.
