Amazon’s Alexa was announced on November 6th, 2014. A passion project for its founder, Jeff Bezos, Amazon’s digital voice assistant was inspired by and aspired to be Star Trek’s “Computer” — an omniscient, omnipresent, and proactive artificial intelligence controlled by your voice. “It has been a dream since the early days of science fiction to have a computer that you can talk to in a natural way and actually ask it to have a conversation with you and ask it to do things for you,” Bezos said shortly after Alexa’s launch. “And that is coming true.”
At the time, that future felt within reach. In the months following Alexa’s launch, it wowed early buyers with its capabilities. Playing music, getting the weather, and setting a timer were suddenly hands-free experiences. Packaged inside a Pringles can-shaped speaker called an Echo, Alexa moved into 5 million homes in just two years, including my own.
Amazon Alexa at 10
The Verge explores how far the voice assistant has come in a decade: its successes, failures, and potential future.
Fast-forward to today, and there are over 40 million Echo smart speakers in US households, with Alexa processing billions of commands a week globally. But despite this proliferation of products and popularity, the “superhuman assistant who is there when you need it, disappears when you don’t, and is always working in the background on your behalf” that Amazon promised just isn’t here.
Alexa is still mainly doing what it’s always done: playing music, reporting the weather, and setting timers. Its capabilities have expanded — Alexa can now do useful things like control your lights, call your mom, and remind you to take out the trash. But despite a significant investment of time, money, and resources over the last decade, the voice assistant hasn’t become noticeably more intelligent. As one former Amazon employee said, “We worried we’ve hired 10,000 people and we’ve built a smart timer.”
Alexa is still mainly doing what it’s always done: playing music, reporting the weather, and setting timers
It’s disappointing. Alexa holds so much promise. While its capabilities are undoubtedly impressive — not to mention indispensable for many people (particularly in areas like accessibility and elder care) — it’s still basically a remote control in our homes. I’m holding out hope for the dream of a highly capable ambient computer — an artificial intelligence that will help manage our lives and homes as seamlessly as Captain Picard ran the Starship Enterprise. (Only preferably with fewer red alerts).
Today, I may have an Alexa smart speaker in every room of my house, but it hasn’t made it more useful. Alexa has gained thousands of abilities over the last few years, but I still won’t rely on it to do anything more complicated than execute a command on a schedule, add milk to my shopping list, and maybe tell me if grapes are poisonous to chickens. (They’re not, but Alexa says I should check with my vet to be sure.) If anything, on the eve of the voice assistant’s 10th birthday, Alexa’s original dream feels further away.
Photo by Jennifer Pattison Tuohy
It’s easy to forget how groundbreaking Alexa was when it first appeared. Instead of being trapped in a phone like Apple’s Siri or a computer like Microsoft’s Cortana, Alexa came inside the Echo, the world’s first voice-activated speaker. Its far-field speech recognition, powered by a seven-microphone array, was seriously impressive — using it felt almost magical. You could shout at an Echo from anywhere in a room, and that glowing blue ring would (almost) always turn on, signaling Alexa was ready to tell you a joke or set that egg timer.
It was Amazon’s pivot into smart home control that provided the first hints of the promised Star Trek-like future. Silly fart jokes and encyclopedic-type knowledge aside, the release of an Alexa smart home API in 2016, followed by the Echo Plus packing a Zigbee radio in 2017, allowed the assistant to connect to and control devices in our homes. Saying “Alexa, Tea. Earl Grey. Hot,” and having a toasty cuppa in your hands a few moments later felt closer than ever.
This was genuinely exciting. While tea from a replicator wasn’t here yet, asking Alexa to turn your lights off while sitting on the couch or to turn up your thermostat without getting out from under the covers felt like living in the future. We finally had something that resembled Star Trek’s “Computer” in our homes — Amazon even let us call it “Computer.”
In retrospect, Alexa brought with it the beginnings of the modern smart home. Simple voice control made the Internet of Things accessible; it brought technology into the home without being locked behind a complicated computer interface or inside a personal device. Plus, Amazon’s open approach to the connected home — in a time of proprietary smart home ecosystems — helped spur a wave of new consumer-level connected devices. Nest, August, Philips Hue, Ecobee, Lutron, and LiFX all partly owe their success to Alexa’s ease of operation.
But the ecosystem that sprang up around Alexa grew too quickly. Anyone could develop capabilities (which Amazon calls skills) for the assistant with little oversight. While some skills were simple and fun, many were buggy and unreliable, and specific wording was needed to evoke each one. It all added up to an inconsistent and often frustrating experience.
Asking Alexa to turn up your thermostat without getting out from under the covers felt like living in the future
Then Alexa hit a wall. There’s an assumption with technology that it will just keep improving. But instead of developing the core technology, Amazon relied on third-party developers to make Alexa do more, focusing its resources on putting the voice assistant in more devices and making it capable of controlling more things.
The more devices that worked with Alexa and the more capabilities Amazon added to the platform, the harder it became to manage, control, and access them all. Voice control is great for simple commands, but without easier ways to talk to Alexa, these new features were lost on most users.
Alexa Routines emerged as a solution to corralling all the different gadgets and functions you could use Alexa for, but this relied on you spending time programming in an app, alongside constantly troubleshooting devices and their connectivity.
Hearing “‘Lamp’ isn’t responding. Please check its network connection and power supply” after issuing a command is beyond frustrating. And spending hours a month configuring and troubleshooting your smart home wasn’t part of the promise. This is what a smart computer should be able to do for you.
We’ve had to learn how to speak to Alexa rather than Alexa learning to speak to us
Amazon masked Alexa’s failure to get smarter with an ever-ballooning line of Echo hardware. New smart speakers arrived annually, Alexa moved into a clock and a microwave, and new form factors attempted to push customers to take Alexa outside of the house in their ears (Echo Buds), on their fingers (Echo Loop), on their faces (Echo Glasses), and in their cars (Echo Auto).
Many of these devices were forgettable, did little to advance Alexa’s capabilities, and mostly served to lose Amazon money. The Wall Street Journal reported earlier this year that Amazon has lost tens of billions of dollars on its broader devices unit.
Even with this “throw everything at the wall and see what sticks” approach, Amazon never cracked that second must-have form factor. In 2017, it invented the smart display — an Echo with a touchscreen that added benefits like video calling, watching security cameras, and showing information rather than just telling you. But sluggish processors, finicky touchscreens, and too many ads meant the smart display never really furthered Alexa’s core benefit.
Today, many users complain that Alexa has become measurably worse at its core capabilities. People seem to buy Echo devices primarily because they’re cheaper than the competition, and they can use them to do basically what Alexa could do in 2014: set timers, check the weather, and listen to music. There’s no expectation for something better from a device that costs as little as $18.
Photo by David Pierce / The Verge
After all these years, just talking to Alexa remains the biggest hurdle. We’ve had to learn how to speak to Alexa rather than Alexa learning to speak to us. Case in point, my connected kitchen faucet still requires me to say, “Alexa, ask Moen to dispense 2 cups of hot water.” As my husband points out, if Alexa really was “smart,” wouldn’t it just know that I’m standing in front of the kitchen sink and doing what I ask without the need for hard-to-remember phrases?
The good news, at least on that front, is that technology is catching up. Large language models and generative AI could bring us an Alexa we can talk to more naturally. Last year, Amazon announced that it’s working on a “new” LLM-powered Alexa that is more proactive and conversational and less pedestrian and transactional.
This alone would be a big leap forward. But while generative AI could make voice assistants smarter, it’s not a silver bullet. LLMs solve the basic “make sense of language” problem, but they don’t — yet — have the ability to act on that language, not to mention the concerns about a powerful AI hallucinating in your home.
What Alexa really needs to become “Computer” is context
What Alexa really needs to become a “Computer” is context. To be effective, an omniscient voice assistant needs to know everything about you, your home, and the people and devices in it. This is a hard task. And while Echo speakers with ultrasound tech and smart home sensors can provide some context, there is one crucial area where Amazon is way behind the competition: you.
Unlike Google and Apple — which have access to data about you through your smartphone, calendar, email, or internet searches — Amazon has largely been locked out of your personal life beyond what you buy on its store or select data you give it access to. And its privacy missteps have kept people from trusting it.
But Google and Apple haven’t cracked the smart home yet, and while they are making serious moves in the space, Alexa still has a sizable head start. According to Amazon, the “New Alexa” can complete multistep routines you can create just by listing tasks. Add in context about who lives in your home, where they are at any point, and what they should be doing, and it’s feasible that the assistant could handle a task like this with just one command:
Alexa, tell my son not to forget his science project; set the alarm when he leaves. Disarm the alarm and unlock the back door for the plumber at 4PM, then lock it again at 5PM. Preheat the oven to 375 degrees at 6PM, but if I’m running late, adjust the time.
This type of capability would bring a whole new level of utility to Alexa, maybe enough to justify charging for it, as the company has said it plans to.
It’s time for Alexa to boldly go where no voice assistant has gone before
However, despite last year’s splashy launch of this LLM-powered assistant, we’ve heard nothing more. Amazon even skipped its big hardware event this year, where it traditionally announces dozens of new Alexa and Alexa-compatible devices and services. This is likely because, based on reports, Amazon is far from achieving its promised “New Alexa.”
But it needs to pull off its promised reinvention of Alexa, or Apple and Google will overtake it.
In 2014, Amazon set the stage for voice control in the home and, over the last decade, laid the groundwork for a smarter home. Today, Alexa is the most popular voice assistant inside a smart speaker — with over two-thirds of the US market. Outside of the home, Google’s Assistant and Apple’s Siri dominate. As those companies invest more in the smart home and eventually bring Apple Intelligence and Gemini smarts to their home products, Alexa’s days of dominance may be numbered.
The path to a generative AI-powered, context-aware smart home is fraught with pitfalls, but with all its history here, Amazon feels best poised to pull it off — if it can get out of its own way. The home is the final frontier, and it’s time for Alexa to boldly go where no voice assistant has gone before and become truly intelligent.