Why isn't it better after five years of Alexa from Amazon?

Five years ago today, Amazon surprised the technical press by revealing Alexa and the Echo, the first smart speaker. "Well, this one came from nowhere," was the response from Verge reporter Chris Welch.


Five years later, Alexa is a household word – literally. If you have children and an Alexa device, they probably know how to ask the robot to play their favorite songs at breakfast. You may have Alexa in your kitchen, on your bedside table, or even in your car.

But half a decade after the launch of Alexa there is another question: has the Amazon assistant actually become more useful?

Years later I still use my Echo for the same basic tasks.
Photo by Chris Welch / The Verge

For me the answer is a clear "no". I started using an Echo speaker not long after it became available in the UK, and after using some silly skills in an app store's Amazon haunted house, I quickly decided on some important use cases: playing music and radio , set timers and occasionally ask about weather and trivia at Wikipedia level. Years later I still have an ultrasound and I still use it for the exactly the same tasks. Nothing has changed. Don't get me wrong, I think the Echo performs these functions pretty well. But like many users, I have learned that asking trickier questions requires problems.

No one would disagree that Alexa has been a huge success for Amazon. The company has added its voice assistant to an abundance of gadgets, from smart rings to microwaves. It has sold more than 100 million Echo devices, Alexa integrated with 85,000 smart home products and the app store now has more than 100,000 skills. On a cultural level, Alexa even managed to give the murderous and merciless Amazon a little personality.

The company has also deployed enormous resources to make Alexa more useful. It has provided new features such as whisper mode, Alexa Guard and Alexa routines, while the underlying promise is being taken away that Alexa is slowly improving in conversations and chit chat. According to Amazon, this problem can be tackled more complex tasks (such as ordering an Uber or informing you of a delayed flight) in a more natural and graceful way. But Amazon has been promising this type of functionality for a long time and does not seem to get much closer to its goal.

Alexa is now available in gadgets of all shapes and sizes.

A few factors play a role. First, it is difficult for computers to understand the language, a challenge that I saw firsthand while winning the $ 3.5 million Amazon prize from Amazon. Thanks to new machine learning techniques, we have become reasonably good at converting speech audio into text from which we can extract basic commands. But any more complex interaction than that is still beyond borders. Language is just inherently nuanced and human. It's our best civilization tool, and although we can use verbal tricks to give computers a sense of understanding, these shortcuts don't last long.


Second is the absence of Alexa on mobile devices. It is present everywhere in households, but I don't know anyone who uses it instead of Siri or Google Assistant on their phone. This limits the possibilities to function as a daily helper. You live on your phone; every assistant must also be there.

The third problem is the most immediate: to make Alexa work as advertised requires too much effort and trust from users. You can use Alexa to set up a morning routine in your smart home, but that takes time. It also means that you have to trust Alexa, not only in terms of privacy (where Amazon has made repeated mistakes), but also in the confidence that the assistant will do what you have asked. For example, I don't use Alexa to shop on Amazon because I don't trust that I get all the information I need about what I buy. It is simply not reliable in a way that encourages repeated use.

All the while, Amazon's competitors have worked hard to catch up. Google in particular has quickly improved its Assistant. I am not a regular user myself, but colleagues who say that the ability to answer questions has an advantage over Alexa. That kind of reliability helps the use of Assistant to a reflex, such as Googling a question. However, I do not believe that even machine learning from Google is sufficient to quickly resolve the underlying challenges of computer conversations.

Ultimately, Amazon has been extremely – even scary – successful in spreading Alexa across the world of internet-compatible gadgets, but the functionality still feels thin. It is a technology that is miles wide, but centimeters deep, always in a pinch but never as good as you would like. And in many cases, it adds exactly zero intelligence to gadgets, instead of talking only about voice control. Amazon does not help with this. Even during the most recent Echo event, the company surprised us with product after product but no sense of clear direction.

I am happy to use Alexa for the same tasks that I used years ago, but I would also exchange my Echo for a generic voice-controlled speaker that can perform the same functions. Maybe I'm just that. Maybe I just don't feel comfortable doing more with Alexa. But I don't think there is anything unique about this technology or, after years of development, that it is almost keeping its promises.

If Amazon doesn't improve the technology in the next five years, Google, Apple and many others have shown that they like to step into the shoes of Alexa.