With AI-based technology, Amazon's Make In India innovations going global

After its success in the Indian market, global e-commerce giant Amazon is now taking few of its India-specific innovations around artificial intelligence (AI) and data analytics to some of its global markets. 

Globally, the Seattle-headquartered company is using AI technologies for numerous applications such as early detection of cancer, driving e-commerce sales through its smart speakers, while it is making efforts to add drones to its armoury of delivery methods. 

In India, the company is betting heavily on technologies such as data analytics, AI, and machine learning (ML) to understand and predict consumer behaviour and build a strong e-commerce business. 

“There’s a lot of interest from global teams in the work that we’re doing on regional preferences and catalogue quality like detecting mismatches between titles and images, and so on. Those are now being deployed in other markets,” said Rajeev Rastogi, director for ML at Amazon India. 

Amazon is using ML to make sense of complex Indian addresses, reduce delivery failures, and cater to diverse regional tastes in the country. For example, a customer searching for a product like ‘sari’ in Bengaluru wants a different type of the item than somebody else is searching for it in Kolkata or Delhi. The search engine shows the users the saris that are popular in their region. 

Besides using ML to improve catalogue quality like detecting mismatches between titles and images, the company is mulling its use to translate the reviews of the products in local languages. “If customers in France say something about a (particular) product, we would love to show that to customers in the UK or India as well,” said Rastogi.

Amazon, which already has over 100 million registered users in India, last September unveiled a Hindi version of its website and app with an aim of attracting the next 100 million Indians to shop online. It is also looking at creating more such platforms in other Indian languages as well where “ML would play a role there.” 

AI aims to build machines that can simulate human intelligence processes, while Stanford University describes ML as “the science of getting computers to act without being explicitly programmed.” 

Amazon’s smart speakers, with voice-activated AI assistant Alexa, are another big opportunity for the company. In India, Alexa had acquired over 20,000 skills such as telling cricket score, conducting quizzes, and doing utilities to serve Indian customers since October 2017 when it launched the Alexa-powered Amazon Echo line of speakers in the country. 

Rastogi said that Amazon is now looking at tapping the population based in rural areas. “Many of the people in rural areas...are more used to an interface where they talk to a shopkeeper as opposed to shopping online,” he said. So getting voice-based commerce (where) somebody could just talk to Alexa and do shopping could be the Next Big Thing. 

According to investment bank RBC Capital Markets, Amazon is expected to generate an additional $10 billion in sales globally from Alexa-related revenue by 2020. 

While AI is expected to drive some of the most important services, there is a growing concern that it could replicate the prejudices that humans have about gender and race. Last year, Amazon had to shut down an AI recruiting tool that showed bias against women. When asked about the steps Amazon was taking to tackle such biases, Rastogi, an alumnus of IIT-Bombay and The University of Texas at Austin, said that biases creep in when the training data is not properly collected. 

He gave the example of taking data from customers in cities to build a model and then giving out recommendations for people in villages. “Those recommendations would obviously be prejudiced.” But, he says, there are ways of mitigating the biases. 

“If I built a tool for recruiting...and there is a bias that it tends to prefer people from IITs and males...just remove those features. So that the model then has to look at other characteristics and make a decision,” said Rastogi.