PART 2: THE FUTURE OF ALEXA
“ALEXA, THE FUTURE OF AMAZON” is a series of two articles with the purpose of understanding the potential of voice-enabled devices. The first article will look closely at how the AI personal assistant functions and how Amazon leverages the technology. In the second article, Alexa’s challenges will be addressed and predictions will be made regarding the future of the virtual assistant.
In recent years, privacy around data has become a widely controversial topic, especially around the GAFA. Amazon came under scrutiny in 2019 when major newspapers such as Bloomberg and Time reported that thousands of Amazon employees listen to voice recordings of Echo users. These employees work from around the world (India, Romania, Costa Rica…) as a mix between internal staff and outside contractors. When asked further questions regarding the exact use of our data, the teams responsible for working 9 hours shifts and reviewing around 1,000 audio clips per person refused to talk, stating a strict nondisclosure agreement. Furthermore, Bloomberg reports, thanks to a screenshot of the audio reviewing interface, that employees in charge of reviewing recordings have access to the user’s account number, first name, and device serial number.
Amazon often uses the fact that the user has the option to manually opt-out of the “voice recording analysis” as a counterargument to critics. The reality is that users only opt-out of voice recording analysis for the development of new features; Amazon may still analyze recordings for the regular review process.
This calls into question the transparency of the company and nourishes users’ trust issues around data collection. In the future, the growing fear around personal data misuse and manipulation could become a threat to the development of Amazon Alexa. Furthermore, the growing awareness around data collection increasingly arouses interest from regulators. The European GDPR law on data protection and privacy passed in 2016 is only the beginning. Governmental entities around the world are starting to pay attention, and the lack of knowledge on the subject paired with the current lobbying will not be able to hold regulators much longer.
With data collection and analysis at the essence of the accuracy and overall performance of the voice assistant, data needs to be widely available for Alexa to succeed. Without proper consideration, data and privacy concerns could become a main obstacle for the company.
Amazon is not exactly known to be inclusive, despite its effort to implement a diverse company culture in recent years. Alexa has often been the target of gender inequalities critics and her AI algorithms, like many in the field, can be biased.
A report published by UNESCO in 2019 called “I’d blush if I could” states that Alexa’s feminine name, female voice, and flirty responses reinforce problematic gender stereotypes. The particular example leveraged to support the claim is that when users say “Alexa, you’re hot”, her response will be “That’s nice of you to say!”. Saniye Gülser Corat, Director of Gender Equality at UNESCO shared in a statement that “obedient and obliging machines that pretend to be women are entering our homes, cars and offices”, feeding into gender stereotypes and changing the way we think of the female voice. This issue is complementary to the fact that in recent years, there have been growing inequalities around women’s digital education. In the AI industry, women are “extremely under-represented” as they only make up “12 percent of AI researchers, six percent of software developers, and are 13 times less likely to file ICT (information and communication technology) patents”.
Bias in data sets is another challenge that Alexa has to tackle. Natalie Schluter (Professor in NLP and Data Science at the IT University of Copenhagen) talks about the importance of training data for it to match all stakeholders. Behind algorithms lies data that can be collected from anywhere and everywhere. Data is a representation of humans and humans are biased; for that reason, a very large amount of data has to be collected for it to be representative of a larger group. However, prejudices or intolerances can still emerge despite multiple verification checks causing misrepresentation. Researchers have found that there is a racial divide in the machine learning models powering speech-recognition systems due to bias data sets causing technologies like Alexa to “misidentified 35 percent of words from people who [are] black” while understanding “white people fared much better”.
With increasing concerns, AI-powered algorithms are now being used on filtering data that will become inputs for another AI model. For Alexa to be widely successful, she will need to understand everyone and not only a narrow group of people like, let’s say, for example, white, young men that live in Silicon Valley.
A very large amount of information today proves that AI can reflect some of the worst human values. Models can be sexist and discriminatory due to biased data sets. As this challenge is upon all Tech companies through the devices they create, Alexa needs to set itself apart to become a benchmark in terms of equality and social representation.
Amazon needs to understand that to reach social acceptance and to appeal to even more customers, Alexa needs to be built with social inclusion in mind, through values such as equality and diversity.
Alexa is considered the most popular virtual assistant, thanks to the Amazon Echo segment, leader in smart speakers with 53%of the market share in 2019 in the United States. She is however far from being the only virtual assistant. Other companies such as Microsoft with Cortana have recognized the potential of voice. More specifically, in terms of performance, Alexa is greatly challenged by Apple with Siri, and Google with its Google Assistant.
To compare the three leading AI personal assistants, Loup Venture decided to test them in its “Annual Digital Assistant IQ Test”. The research found that the Google Assistant, once again, was the most accurate and answered correctly to 92,9% of a sample of 800 questions. Its success is mainly due to the quantitative data sets that the Google Assistant’s ML algorithm gets to train with. As the world leader in the web, Google has managed to create colossal data sets of structured data. The company also has a more advanced speech recognition software allowing the Google Assistant to understand every question in the sample. Coming in second place, there is Siri, with 83,1% of answers correct and she understood 99,8% of the same question sample. When launched in October 2011, Apple promised a revolution in digital assistant but could not yet deliver on that promise. With time, Siri has managed to keep its place at the top as Apple tries to export the voice assistant outside of the IOS environment with the latest Mini Pod, a smart speaker directly in competition with the Amazon Echo. Finally, Alexa comes third after answering correctly to 79,8% of the sample while understanding 99,9% of the queries. On top of being less performant than other personal assistants in the research, Alexa is not backed into the OS of a phone. This becomes a disadvantage as the Amazon virtual assistant only lives inside third-party apps and does not have access to commands such as phone calls, messages, or emails. Another difference is that Alexa entirely relies on voice, unlike its competitors which can offer a regular touch interface in case of frustration.
Alexa Skills Kit helps to grow the massive developer community while AVS is disrupting many industries. Alexa is one of the most powerful AI personal assistants, but that does not mean that Amazon is the only company well-positioned in the fast-growing market.
The adoption of voice assistants is growing worldwide, and predictions are encouraging. TechCrunch reports, according to Juniper Research, “the use of voice assistants is set to triple over the next few years” going from 2.5 billion to 8 billion digital voice assistants in use by 2023. More specifically, the growth will be tremendous in segments where Alexa is already well-positioned. This sentiment is exemplified by the smart TV voice assistants that are projected to grow 121% in the next five years.
The research reinforced what Amazon already knew, voice-only interaction will exponentially increase in the future. As a tech giant, Amazon has lots of influence in the field and is heavily investing in voice-related technology. In 2019, the company’s total research and development expenses reached 35.9 billion dollars, a year-to-year increase of 25%. With such limitless resources to pour into innovation, the tech giant is exploring countless opportunities to grow Alexa.
The Alexa team has filed a patent for technology that would scrap the “wake word” from the user command formula. It would allow users to simply address their request to the device in the following manner: “play me some music, Alexa”. Not only it would make the information more fluid, but it would also transform the interaction into something a lot more human-like.
On the same track, Amazon is developing Alexa’s Conversation Skill, allowing her to ask questions and to make the conversation two-sided. Currently, in beta mode, Alexa Conversation is a deep learning-based approach with the purpose of creating a better conversation flow between devices and humans. To feed the deep learning conversation model with enough data, the Skill will expand Alexa’s memory allowing her to collect more information to sustain longer, two-way conversations with a user. The feature will incorporate asking more than one question at once, letting the user make corrections, and allowing the customer to make a correlation to past queries.
In the future, Alexa will revolutionize the way we interact with technology. By mutating interaction between humans and machines, Amazon’s AI personal assistant will disrupt industries by allowing seamless automated conversations between businesses and their clients. Lastly, Alexa is aiming to become your virtual companion, putting her capabilities to use through harmonious voice.
Simplistically, Alexa is Amazon’s secret weapon in terms of purchasing behavior, opening new doors to customers’ personal lives. The company’s focus will always be on selling. With Alexa, the e-commerce leader manages to build on its search and recommendations engine through new types of data. The new data feeding the powerful ML models is not the only opening for Alexa in e-commerce; the personal assistant is an opening to voice-based shopping.
With voice dramatically simplifying processes compared to a regular touch interface, the quicker new purchasing channel is predicted to seduce 21.6 million US consumers by the end of 2020. Smart device makers have not yet gained user’s trust in terms of privacy and payment security causing a slower adoption than expected. There is also a substantial difference to our regular buying channel, which is that with voice, the consumers cannot see what they are buying.
Despite the hesitation from many smart speaker owners, Amazon is betting on voice purchase in the future, as social acceptance around the technology is only climbing.
The next frontier for Alexa to fully help users is her transition to becoming a physical AI assistant. Rohit Prasad, the VP and head scientist at Alexa Artificial Intelligence shared that “AI programs need to see and explore the world if they’re ever going to attain real understanding” because “the only way to make smart assistants really smart is to give it eyes and let it explore the world”. It is clear that by listening to executives, Amazon has in mind, not only for its operational aspect but for better algorithm and AI understanding, a body for Alexa.
With Robo-pets already on the market such as Anki Vector, Sony Aibo, or Zoetic Kiki, Alexa built-in was used by a third party, Anki, to build “Vector”, a tiny home robot with limited capabilities. With exceptional R&D power, it would not be difficult for Amazon to leverage its AI personal assistant and create a body for Alexa in the coming years.
Alongside Alexa’s objective of creating a virtual companion, another goal of Amazon is to spread its voice technology everywhere. With data collection at the center of voice, Amazon’s strategy is two-sided. First, the company has created a vast range of Alexa enabled devices for users to control all types of smart devices with voice. Since its start back in 2014, Amazon has released hundreds of Alexa-enabled devices, and the assistant can now follow you everywhere with Echo Auto or the Echo Frames. The second part of Amazon’s strategy is the company’s aggressive approach to pricing. Apple wants you to have one Pod Mini for $99 as Amazon wants you to have the $20 generation 3 Echo Dot in every room of your house.
With the growth of the Internet of Things (IoT), Amazon is riding the wave, and connecting every device that it can to Alexa. With the technology getting smarter every day, Senior Vice President of Devices and Services, David Limp, explained that “there is no reason not to put them everywhere in your house”.
With now existing devices such as smart microwaves and connected rings, we are only entering Amazon’s vision for a world ruled by Alexa and controlled by your voice.
 “Thousands of Amazon Workers Listen to Alexa Users’ Conversations”, TIME by MATT DAY, GILES TURNER AND NATALIA DROZDIAK, April 11th, 2019.
 “Are robots sexist? UN report shows gender bias in talking digital tech” UN News on May 17th, 2019.
 “There Is a Racial Divide in Speech-Recognition Systems, Researchers Say” The New York Times, by Cade Metz, March 23rd, 2020.
 “Amazon Smart Speaker Market Share Falls to 53% in 2019 with Google The Biggest Beneficiary Rising to 31%, Sonos Also Moves Up” Voicebot.ai, BRET KINSELLA on April 28th, 2020.
 “Annual Digital Assistant IQ Test” Loup Venture, Gene Munster, Will Thompson August 15th, 2019.
 “Report: Voice assistants in use to triple to 8 billion by 2023”, TechCrunch, by Sarah Perez on February 12th, 2019.
 “Amazon research and development expenses” Statistic, by Abhijeet Pratap on October 13th, 2020.
 “Shopping via smart speakers is not taking off”, Techcrunch, by Sarah Perez on February 4th, 2020.
 “Alexa needs a robot body to escape the confines of today’s AI” MIT Technology Review by Will Knight on March 26th, 2019.
 “Amazon Wants Alexa to Move (With You) Far Beyond the Living Room”, The New York Times, by Karen Weise on September 25th, 2019.