Amazon explains how it uses generative AI to prevent delivery of defective products

Amazon uses technology in its order flow to prevent defective products from being delivered. The company (3) detailed the entire process, accomplished with generative artificial intelligence (AI), in a post on Monday.

The verification system is supported by the “Project PI” model; “PI” is the English abbreviation for “Private Investigator”. A. the tool uses a combination of generative artificial intelligence and computer vision technologies to track order flow and detect defective products or incorrect orders (out-of-order sizes and colors).

According to Amazon, Project PI is smart enough to detect the source of the problem, allowing the team to adopt fixes and avoid future errors.

“The system, where available, has been proven to be capable of classifying the millions of items passing through tunnels each month and accurately identifying product problems,” Amazon said.

How does Amazon’s Project PI work?

A. Amazon explains that Project PI is installed in a ‘image tunnel’. In this setup, the vehicle uses computer vision to scan products and assess their condition (type, packaging irregularities, and other details). If a fault is detected, the product is removed from the order flow for replacement.

After leaving, The product was reviewed by Amazon experts for a more in-depth conference. The condition of the object is then evaluated to see if it can be resold at a lower value, donated or thrown away.

“Accidentally shipping defective items can lead to undesirable returns, which can lead to packaging waste and unnecessary carbon emissions due to additional shipping,” Amazon argues.

In addition to its defect detection system, Amazon uses generative models (specifically a multimodal language model) to investigate the cause of negative customer reviews. The company learns from the buyer’s complaint and thus detects a previously undetected defect, preventing this from occurring again in future deliveries.