“Machine learning is ubiquitous at Amazon today,” said Rajeev Rastogi, Vice President, Machine Learning at Amazon India, in an interview with Gadgets 360. “At retail, we are widely using machine learning to recommend products to customers. , predict the future demand for products, and improve the quality of a product catalog, both classifying products and eliminating duplicate products.”
One of the most basic examples of how Amazon uses machine learning (ML) is when you misspell a search query in the search bar. The ecommerce site, Rastogi noted, looks at the phonetic distance between the typed misspelled query and the correct query rather than looking at their textual distance to give accurate results — regardless of whether you misspelled something.
For example, if you type “geezer” on Amazon to search for the available geyser options, the market will automatically correct the spelling and show you relevant results. Amazon also uses ML models to translate the content on its site into the Indian languages it now supports.
Of course, this kind of use of computers is commonplace now, and not something most of us think of when we look at the terms artificial intelligence (AI) or machine learning.
Rastogi revealed that his team is currently working on a seed initiative that aims to deliver a conversational shopping experience. It is aimed at first-time online shoppers who are more comfortable interacting with offline merchants than placing an order through an e-commerce site.
Conversational commerce, through chatbots, through smart assistants like Amazon’s own Alexa, is one of those ideas that comes back every few years as technology improves, and Rastogi talks about how it will start with text, in English, but grow into other languages. and to vote.
“A machine can read a document and then answer any question about the document, which is difficult. Nowadays AI can’t generate a review for a movie, for example… Even summarizing a series of documents is a challenging problem. It is not solved by AI in any way,” Rastogi emphasized.
AI has been used to analyze text and speech at different levels. But computer engineers and data scientists have not yet been able to find a relevant mix for using AI and machine learning to generate accurate ratings, such as movie or product reviews. A research paper published by researchers Gerit Wagner, Roman Lukyanenko and Guy Paré of the Department of Information Technology, HEC Montréal, on how AI can be used in the literature research process, notes that even “technically perfect tools (like researchers)” sometimes struggle to evaluate information from sources that use ambiguous, confusing language and presentation.
McKinsey Global Institute (MGI) partners Michael Chui, James Manyika and Mehdi Miremadi also pointed out in an article that AI models “have difficulty transferring their experiences from one set of conditions to another” and require companies to even when the use cases are very similar. This adds additional resource requirements.
Shreyas Sekar, an assistant professor of Operations Management in the Department of Management, University of Toronto Scarborough and Rotman School of Management, said the effectiveness of an AI-based bot that communicates with people and gives them appropriate results, especially in markets such as India , not sure. . Sekar has conducted extensive research into how ecommerce platforms use machine learning at both consumers and warehouses to improve their operations.
“If you ask these chatbots simple questions, like no, will it rain tomorrow? Or can you play the song from this movie for me? They do a great job. But if you get more and more complex questions, like hey, can you help me find a good shoe for my trek? I think it is very difficult for the chatbot or even Alexa to clearly break this question down into what your intent is? What do you do as a person and how do you distinguish yourself from other people? And which products suit you?” he said.
Dealing with prejudices and mistakes
One of the biggest challenges in using AI and ML today is limiting bias and errors. Companies from Google and Facebook to Microsoft regularly deal with these blunders. Amazon is not infallible in this area either.
Sekar of the University of Toronto Scarborough and Rotman School of Management noted that Amazon’s AI implementations contain many biases that the company is already aware of and apparently working to resolve, but it’s not clear how successfully it is achieving desired results. has achieved.
“Maybe in the past users have clicked on a certain brand of earphones, and then what happens is that in the future I keep amplifying that exact brand again and again. So this is usually referred to as a kind of popularity bias, where I see products that are already popular. I’m trying to put the spotlight on, and basically I’m helping the rich get richer in the system,” he said.
However, Rastogi strongly disagreed, saying Amazon’s goal is to help human workers, not replace them entirely.
Who does this help?
Using AI and ML helps Amazon deliver what you need by understanding your buying behavior and purchase history. However, this sometimes leads to impulse purchases and simply convinces you to buy something you don’t really need. Experts believe it would continue to grow with a more casual shopping experience.
“I think AI and ML can definitely expand the idea of turning shop window buyers into regular buyers,” Sekar says. “And this is definitely something that I think is a good way to see Amazon as a very compelling seller.”
Consumers can overcome this behavior themselves by understanding how algorithms can influence their choices.
“Even though we are the ones who click on a product at the end to buy it, we are guided through the algorithm by the algorithm in several places, be it the recommendation or the reviews,” Sekar said.
Ankur Bisen, Senior Partner and Head of Consumer, Food and Retail Divisions at management consultancy Technopak, said the nature of how Amazon uses its algorithms to entice consumers to buy more is exactly like what advertising, marketing and even discounts at a shop did.
“Amazon does it with a lot of precision because it’s defined,” he said. “Conversational AI isn’t just near Amazon’s monopoly domain. Yes, they are very good at it because of Alexa. But you’ll see conversational AI popping up in different forms offered by other technology platforms.”