Chinese search engine giant Baidu is notably leaning toward artificial intelligence (AI) and self-driving vehicles. Its autonomous driving platform, Apollo, recently unveiled a major update, which allows everyone using its platform to develop cars that can drive autonomously on simple urban roads.
Baidu is also taking efforts to promote its AI technology. Its recent attempts involve a case helping convenience stores manage their fresh food stock more efficiently, thus reducing waste and increase profit.
Since last September, Baidu has started supporting AI service provider Ping++ to work with convenience store chain Today on a fresh food project utilizing Baidu’s deep-learning platform PaddlePaddle and its click-through rate (CTR) model.
Liu Yongfeng is the senior project manager of PaddlePaddle and the person-in-charge of the PaddlePaddle and Today convenience store project.
According to Liu, Baidu launched PaddlePaddle in 2013. And the technology supporting the CTR model was used to forecast customized search results and advisements for different users based on data analysis results of their previous online behavior.
“Since we have built the model and it has worked, we began to wonder whether we can make similar forecasts if data from other industries are added and analyzed,” Liu said.
After the technology’s successful application in online search and advertising, Baidu then realized it could use such data analysis and forecasting methods in other areas such as retail.
In the convenience store project, Baidu supported Ping++ which made use of the PaddlePaddle platform to analyse the sales data from hundreds of stores over the past one to two years as well as 70-plus other dimensions that might affect food sales including weather, location, and festivals.
“It took about one day to train the computer to learn the relation between sales and other affecting factors and only milliseconds to forecast future sales,” Liu added. “Among all the factors, we found out it’s the weather that affects sales most.”
Having learned the relationship between sales and affecting factors, Baidu then brought the model to dozens of store for testing. The model helped store managers forecast fresh food consumption, including meal boxes and rice-ball, and prepare inventory accordingly.
“In the testing stores, their profits have increased by 20% and food waste has dropped 30%,” Robin Li, Baidu’s founder and CEO said last December. “That means we can avoid wasting food thanks to AI technology.”
However, Baidu doesn’t directly provide such AI service to convenience stores. It only supplies relevant technology to other business service firms such as Ping++. These companies utilize Baidu’s technology to provide direct services to businesses.
Baidu hopes the convenience store case will enlighten more traditional industries to explore what AI can do for their businesses and target customers.
“Offline retailers, manufactures and other traditional sectors have accumulated tons of data in the big data area, but they have no idea how to manipulate these data,” Liu said. “Actually they can use the data to forecast when their equipment might breakdown, or to make more efficient procurement combinations. We hope the fresh food case can help them reflect how their data should be used.”
（Top photo from Baidu.com)