How Amazon's Alexa Uses AI and Machine Learning
Thu, April 22, 2021

How Amazon's Alexa Uses AI and Machine Learning

Some of you might have already used these commands with Amazon’s Alexa, an intelligent personal assistant that makes use of machine learning and artificial intelligence. Just like any other company, Amazon is heavily investing in AI / Photo by: seewhatmitchsee via 123RF

 

"Alexa, show me the weekend forecast." "Alexa, show me movie showtimes." "Alexa, help."

Some of you might have already used these commands with Amazon’s Alexa, an intelligent personal assistant that makes use of machine learning and artificial intelligence. Just like any other company, Amazon is heavily investing in AI. One great proof of this is Alexa. Since it was first launched on the Amazon Echo in November 2014, Alexa has attracted more and more users every year. It is designed to be always on and woken with a voice command.

Alexa is capable of so many things, such as creating calendar events, performing web searches, ordering products, modifying to-do lists and notes, reciting emails, reading Twitter posts, and many more. It can offer hundreds of additional capabilities such as requesting an Uber, adding items to your grocery shopping list, and getting recipes. But Amazon designed Alexa in such a way that it could adopt even more abilities. For instance, the tech company added a wide range of new Alexa capabilities in September, including the Alexa Presentation Language for building visual Alexa Skills for use on Alexa-powered devices with screens.

A 2019 report by Consumer Intelligence Research Partners (CIRP) revealed that Alexa devices maintain a 70% market share in the US. According to Marketing Land, a daily publication that covers all aspects of online marketing, including analytics, social media, mobile marketing and more, the US-installed base of smart speaker increased by 76 million units from only 70 million units in the March 2019 quarter and 50 million units in the June 2018 quarter. Other surveys argue there are more than 100 million units in American homes.

Another survey by Microsoft showed that Alexa was the second most-used virtual assistant among consumers, with Siri and Google Assistant at the first spot. The top use cases were searching for a quick fact (68%), asking for directions (65%), searching for a business (47%), researching a product or service (44%), and making a shopping list (39%).

How AI Helps Alexa

The most appealing part of Alexa’s system is that it can work solely with voice commands. According to Amazon, it designed the Alexa Voice Service (AVS) to mimic real conversations, which explains why it can conduct a conversation with the users. “Alexa” is simply the “wake word” that alerts the service to start listening to your voice. The tech company’s AVS can be used to voice-enable any connected device that has a microphone and speaker. “Alexa is always getting smarter with new capabilities through machine learning,” Amazon’s Developer site said.

According to Fast Company, the world's leading progressive business media brand with a unique editorial focus on innovation in technology, leadership, and design, Alexa’s voice sounds more human when it translates text into speech because of its neural networks. This has allowed Amazon to take a different approach to generate speech. Rohit Prasad, who heads Alexa machine learning and artificial intelligence, explained that neural networks are being used to generate whole sentences of text in real-time, creating a vocal sound that’s more fluid and more human-sounding. 

According to Forbes, a global media company focusing on business, investing, technology, entrepreneurship, leadership, and lifestyle, Alexa is considered the “AI” that lives with us. Alexa is also an example of a conversational AI where devices can interact with people. Through a voice user interface (VUI), voice services like Alexa can communicate with people in ways that feel effortless, can solve problems, and can get smarter over time.

Through AI, Amazon expanded Alexa’s hearing to detect more than just human commands in 2018. Unlike before, Alexa now has a sensitive microphone that detects the sound of glass breaking and smoke alarms going off when nobody is at home. Through its “away” mode, the voice assistant can listen for human-related sounds in the home such as the sounds of footsteps, coughing, and doors closing. Alexa would then send an alert to a user if it detects one of these sounds. All of these were made possible through a deep learning model. 

However, Amazon admitted that Alexa is still a work in progress. “The more data we use to train these systems, the better Alexa works, and training Alexa with voice recordings from a diverse range of customers helps ensure Alexa works well for everyone,” the company said. According to The Verge, an American technology news and media network that publishes news items, long-form feature stories, guidebooks, product reviews, and podcasts, an in-depth investigation from Bloomberg revealed that Alexa is being improved by having human beings listen to recordings of users’ voice requests.

The most appealing part of Alexa’s system is that it can work solely with voice commands. According to Amazon, it designed the Alexa Voice Service (AVS) to mimic real conversations, which explains why it can conduct a conversation with the users / Photo by: bee32 via 123RF

 

Alexa’s Privacy Issue

Just like other devices and technologies powered by AI, Alexa also faces its own privacy concerns. In recent years, members of the public have reported that the voice assistant secretly recorded their private conversations. In San Francisco, Shawn Kinnear claimed that his Echo activated itself and said cheerfully: “Every time I close my eyes, all I see is people dying.” A woman from Portland, Oregon also reported that the device had sent recordings of private conversations to one of her husband’s employees. 

Thus, concerns over Alexa and other voice assistants have increased, highlighting the possibility that they are invading users’ privacy. Jennifer King, director of consumer privacy at the Center for Internet and Society at Stanford University, stated that the concern is warranted because this type of advanced technology is not fully understood by consumers. "We are basically testing fodder as they roll out more and more technologies that are based on artificial intelligence. These companies are crowdsourcing these algorithms. As a consumer, you're basically helping them learn as they go,” she said. 

According to The Conversation, a network of not-for-profit media outlets that publish news stories written by academics and researchers, another concern is the potential for law enforcement to make use of the technology, essentially putting their ears in our homes, schools, and workplaces. Thus, experts have stated that tech companies should do a better job at partitioning different types of information with different layers of security in voice assistants like Alexa to prevent private information from being so easily shared.

While Amazon's Alexa is a great help for users at home, we should still be cautious of investing in these devices because tech companies are not doing enough to prove that the data or information gathered from users through these are kept safe.