Beth Holmes, lead engineer at Amazon Alexa, talks about the voice assistant that powers over 100 million devices worldwide and her life as a technologist with autism. Text and portrait by Nick Smith.
Voice activation has become ubiquitous in recent years, with Amazon Alexa cloud-based voice service becoming a major player on over 100 million devices worldwide. Billions of people interact with these devices every week, according to Beth Holmes, Principal Knowledge Engineer at Amazon Alexa.
Holmes has been working on the product since November 2014, when “a small multidisciplinary team launched the Amazon Echo, hoping to use artificial intelligence to revolutionize everyday convenience.” Prior to that, she was part of the technology team at startup Evi Technologies, whose acquisition by Amazon meant that its natural language artificial intelligence program has become a key component at the heart of Alexa.
The statistics are incredible. Holmes provided a bunch of numbers. More than 200 million smart home devices are connected to Alexa, from lamps and power outlets to robotic vacuum cleaners and grills. According to Holmes, “customers use Alexa and their smart home devices every few milliseconds to streamline their morning routines, remove chores from their to-do lists, and keep their homes safe.” There are over 140,000 products that are compatible with Alexa. With predictive and proactive features like Alexa Guard, Routines, and Hunches, Alexa initiates a quarter of all smart home interactions without the customer saying anything. There are over 130,000 Alexa skills (the equivalent of apps) and the number of people using the skills is growing at 40% annually, with significant growth in music, audio and gaming.
Just as apps are the lifeblood of a smartphone, “skills” are the catalyst for what Holmes calls “a paradigm shift in human-computer interaction” in voice computing. Prior to the introduction of the Echo — Amazon’s line of devices offering the Alexa voice platform — “customers were used to searching on desktops and phones, and they were entirely responsible for sifting through blue links to find answers to questions or connect to services. Online app stores offer the “have an app” convenience, but the cognitive load on customers continues to rise. Alexa-based Echo disrupts these human-computer interaction paradigms by shifting the cognitive load from the client to AI, resulting in a tectonic shift in the way we interact countless services, searching for information on the Internet, managing smart devices and connecting with other people.”
In other words, skills are little apps that help you do things using just your voice. As Holmes explains, “Customers can use Alexa to do a lot of things: play music or video, answer questions, set timers and reminders, notify you about package delivery or sports updates, control your smart home devices, send messages to your friends and family. calls.”
Holmes confirms that he uses his product in everyday life: “I like using Alexa because typing on the computer all day is not healthy. It’s nice to chat to get some tasks done. It’s especially important to talk, for example, older people with hand arthritis or other disabilities mobility, visually impaired or children who cannot yet read.”
Holmes, originally from Cambridge, became the lead knowledge engineer for Alexa in early 2020, where she provides technical oversight to the knowledge engineering team. She describes knowledge engineering as a branch of ontology, or the study of how concepts exist and relate to each other. She explained that a simple ontology (or model) of a film can contain several concepts such as cast, filming schedule, release date, and more. “For example, Prime Video needs to model whether an Italian dub of the director’s cut is available in Canada.
“I am currently working on projects to ensure consistency between the mutually open ontologies of the Alexa team. I also lead the Ontology Health movement: when a large number of employees contribute to ontologies under time pressure, hacks and technical debt begin to pile up. so we have cleanup projects.”
These ontologies can cross departmental boundaries, meaning that Holmes will communicate with ontologies from other departments, such as the Amazon retail site. “I’m working with them on a cross-ontology community initiative. For the last few years, I have been hosting our annual KnowledgeCon and participating in their employee satisfaction analysis.”
In response to a general question about her past, a 45-year-old woman sent me an article titled “Lesbian Motherhood and the IVF Scandal of 1978″ in which she explained that “My mother was a lesbian when people were absolutely shocked by our family structure. part of my upbringing was raised by women who chose their own path, which probably allowed me to have less rigid expectations in the first place.”
At school, Holmes was fond of academic subjects, but not sports. During this time, she developed a parallel interest in art, wanting to be a taxi driver, an Olympic gymnast, and even “a great writer like Enid Brydon. I don’t think I had any kind of support system. that I will graduate from knowledge engineering. My mother thought that I would be a teacher because she was a teacher. My school knew that I would get good grades, but they didn’t seem to have a problem with how I should do it. ”
Holmes was particularly fond of mathematics because “it was easy and I really liked it”, which paved the way for her to earn a bachelor’s degree in mathematics from what is now Aberystwyth University in Wales and then a bachelor’s degree from the University of Birmingham. D. (also mathematics), her dissertation in computational group theory was entitled “Monster Computations” (“Monster” is the largest of the sporadic simple groups “too large for standard computational methods”).
“Its minimum representation is a set of ~200,000 matrices. My PhD. There are unresolved questions about monster structure.” move for family reasons, she decided to leave the academy.
With the help of a friend, a highly skilled mathematician found a job “in knowledge engineering at a start-up company. I had no previous knowledge or experience in this area because it was a niche area. They just want new people to come and hopefully learn. Startup Evi Technologies works like this: it takes a question as natural language, translates it into a machine-readable query, runs a query to get data results, which are then converted back into natural language answers.” Holmes explains how this works: “The client enters: “When is the next public holiday?” Evy translates it like this: “Find the client’s country/region, find a list of public holidays and their dates. , select the earliest”. Evie’s query engine returns: England Good Friday 14 April 2022. Evie then turns this into a sentence: “The next public holiday in England is Good Friday 2022 April 14.”
This kind of natural language processing is perfect for Alexa, says Holmes, because its developers “already have speech technology that goes from audio to text and then from text to audio. for both stages. Before the startup I worked for was acquired by Amazon, my role changed from individual contributor to manager. The acquisition resulted in reasonable perks and the ability to hire more people. As a result, as a manager, I was able to a great extent. Building my team online helped me move forward in this role.” A bigger team means more leadership and more opportunities to meet and collaborate with other people on the Alexa track. But managing was not the right path for Holmes: “I didn’t think it would work for me, so at the beginning of 2020 I became an individual again.”
Around this time, Holmes was diagnosed with autism. While her official diagnosis at age 44 “didn’t make much of an impact, one of the things it allowed me to do was to be more open about it.” She said she was certain she had autism only in the years leading up to her diagnosis. “I’ve always felt like I’m very different from other people and recognizing why that is, how I’m different and where it’s easier for me to understand why I find things harder or easier. Everything is fine with me. I can start with these things instead of trying to be good at what someone else is good at, now I’m more likely to succeed.”
Holmes says that autism is “disabled”. But they were there whether they diagnosed me or not. But the diagnosis offers some explanation.
She said she used to feel alienated, “especially the ‘work woman’ message.” It seemed like there were a lot of beliefs that I couldn’t share.” She also can’t understand some of the negativity: “We’re told it’s hard for women to speak at conferences, but I’m being very direct.” can “get rid of the sleazy, sleazy, socially awkward image of the nerd programmer” to attract more women. In fact, she says, these are “common features of Asperger’s syndrome that I associate myself with.” So I feel like the female-focused effort is about recruiting “real women” and not like me. ”
Holmes sees potential opportunities for people with autism in STEM fields. “Many people with autism excel in software development. From what I’ve heard on the community forums, it’s like a lot of jobs in other scientific fields, like lab assistant. Indeed, there are companies that require good people skills from you. too many abilities, especially in large companies. But many people with autism (particularly women) have learned these skills, and for those who haven’t, there are still many employers who only need to complete these tasks.” various backgrounds.
Worried about how she might influence others, Holmes vehemently denies being a role model, preferring to leave the label behind “a person who has achieved greatness like Greta Thunberg”. She went on to say that while many people with autism often “suffer from depression and life’s difficulties,” any problems she had to deal with are a thing of the past. “I feel good in most aspects of my life these days. The abilities that have brought me here in my career are not outstanding. Rather, they are common features of autism that many other people have.”
Those traits include “learning topics in depth, retaining information about them, and forming independent opinions without the influence of groupthink,” she said. She concluded by noting that it was important for her to take advantage of opportunities such as media interviews so that she could “show autistic people with similar traits that ‘it’s okay’ and show neuronormal people that autism can be dangerous.” how”.
© 2022 Institute of Engineering and Technology. The Institution of Engineering and Technology is registered as a Charity in England & Wales (no 211014) and Scotland (no SC038698). The Institution of Engineering and Technology is registered as a Charity in England & Wales (no 211014) and Scotland (no SC038698). The Institute of Engineering and Technology is registered as a charity in England and Wales (number 211014) and Scotland (number SC038698). The College of Engineering and Technology is registered as a charity in England and Wales (number 211014) and Scotland (number SC038698).