Smart speakers
There have been some significant changes in the smart speaker world, and not just on the outside
It’s remarkable how quickly smart speakers have become part of our lives. And there have been some big changes to the tech — and not just on the outside.
When you think about it, it’s quite amazing how quickly smart speakers have become so completely normalized. Speech recognition has been around in some limited form since the 1960s when IBM built the Shoebox, capable of recognizing 16 words and numbers. In the mid–1990s, Dragon Dictate introduced reasonable (though sometimes comically imprecise) speech recognition on the desktop as long as you were willing to invest around twice the price of a computer to pick up V1 of its software. You could argue that it was Apple that brought voice tech to the masses with Siri, though Siri actually started development in 1993; Siri Inc was snaffled up by the Cupertino giant in 2010, and hit the iPhone in 2011.
But to look at any company other than Amazon as the cause of the spectacular growth of smart speakers would be to look in completely the wrong direction. Alexa, first introduced with the Amazon Echo exclusively for Amazon Prime members in 2014, beat Microsoft’s Cortana to the market and came two years before the Google Assistant. That first Echo caught a huge amount of press and public attention: here was something brand new in the tech field. A self–contained AI brain, squashed into a speaker that wasn’t too bad at pushing out a tune. Amazon expanded its capabilities with Alexa Skills, it released a friendly API with terms of use acceptable enough that all and sundry began building compatible smart devices. It made the voice assistant a thing.
FLAT–BOTTOMED SPHERES
That’s not to discount the contribution that Amazon’s competitors made. Were Alexa an island, Amazon might have rested on its laurels. But Alexa showed the world what was possible, it kick–started the second wave of the smart home, and it made those competitors possible. Some fell on their faces — even Microsoft is backing down on Cortana a little and the Samsung–only Bixby is of questionable usefulness. But we’ve been left with a big–three of AIs and a big–three of speakers: Alexa in Amazon’s Echo line, Google Assistant in the recently rebranded Nest range, and Siri in the HomePod. All three are headed towards the next stage of smart speakers.
Look at the new fourth–generation Amazon Echo and you will note that Amazon has given the outside of the Echo and Echo Dot (right) a radical redesign. The cylinder and puck shapes of previous generations are out, replaced by flat–bottomed
spheres. Both now have speakers which point in a single direction rather than spewing their sound around the room, which the larger of the pair can use to its advantage; it gains the Echo Studio’s ability to calibrate itself to a room, employing its big woofer and twin tweeters for a more complex sound. The $49.99 Echo Dot is stuck with a single small speaker, and doesn’t get any sound calibration, but hey, it’s cheap. A redesigned shell and some new audio tricks are interesting, but it’s what’s going on inside the Echo that’s really exciting at this point…
First, Amazon has killed the Echo Plus entirely, and instead stashed its Zigbee smart hub into the mainline Echo, making that feature available on a far cheaper device. It has also integrated support for Amazon’s own Sidewalk smart device networking system, which uses Bluetooth low–energy connections to help smart devices talk. At some point it’ll be able to do things like track your pets, and while we don’t expect it to replace existing smart home protocols, it opens up options which might well relate to the inter–company smart home discussions that’ve been going on.
Going deeper into the internals, we find the most exciting change. Amazon has equipped both new devices with a new chipset, developed in conjunction with MediaTek, which includes what Amazon calls the AZ1 Neural Edge processor. This chip is powerful enough to pull off speech recognition tasks onboard the device itself — though strangely the Dot does not have enough RAM to do so, only the bigger Echo. That means for many tasks there’s no need for a round trip to the cloud to get your voice data processed. The difference in reaction time at this point is fairly negligible: Alexa hasn’t exactly been
slow without it but it means that the Echo can now respond faster to your basic requests.
GUTS AND GLORY
Amazon’s little big hardware change is most significant for a few reasons. First of all, it will no doubt inspire competition and improvement elsewhere. The rebranded-fromGoogle Nest Audio ($99.99), the latest device in the main line Google Assistant family, doesn’t have any such processing, but we’d be very surprised if future devices from the Nest camp didn’t take the same path, perhaps doing it even better. Apple is in charge of its own hardware destiny; a piece of Apple silicon dedicated to Siri would make future iterations of the HomePod, or any other Apple device for that matter, so much slicker.
Secondly, the Neural Edge processor brings forward the potential for the Echo to go totally offline in the future. It doesn’t do that yet, but it could. An Echo which doesn’t connect to the internet at all could be a very handy device both in a portable and home setting. This is all theoretical, but imagine a smart speaker with enough storage inside to pack a core set of knowledge, able to answer your questions without reaching out to the internet at all. It might not be the most reliable resource, but consider (for example) the size of Wikipedia: its article database, just the text, is just under 20GB, which is easily small enough to be stored on something as small as a microSD card. Combine this with locally communicating smart home devices, perhaps taking advantage of Amazon’s Sidewalk protocol, and you’ve got an offline smart home set–up with the potential to do just about everything Alexa can do now. Pack it into a tiny
shell like that of the Echo Auto and you can take Alexa wherever you go with no need to worry about a cellular connection.
CHANGING TIMES
So, er, why bother, when the availability of an internet connection is all but ubiquitous? Privacy. Many already worry about the amount of data Amazon takes in and keeps, and they’re concerned by their voice recordings ending up online. The concept that an Echo speaker (and, indeed, any other smart speaker) is always listening is a worry. By default, a smart speaker only listens for its wake word, and to do more would lead to that speaker being hit with a raft of legal notices and sour headlines — but many people are understandably suspicious. A completely isolated speaker would alleviate their concerns. To be clear, we’d be very surprised if Amazon ever actually does dumb down the Echo in such a way, because your voice data is valuable to the company, and it’s what has made Alexa so much more clever over the past six years. But the concept is exciting.
And so we move on to Apple, which has finally done more with the HomePod, squishing it down and releasing the HomePod mini ($99) to, if not an unsuspecting public, a public that has had an inkling something might be up for about two years now. It has upped the HomePod’s smarts even further than before, and while it might not be able to answer your questions without heading to the internet to find an answer, the mini excels in a host of other departments. Apple’s ever–interoperable ecosystem gives it the ability to precisely detect when an iPhone is in the vicinity, for example, which is handy. And it now includes a feature called Intercom, which is pretty much equivalent to the Echo’s drop–in functionality, allowing you to call between devices in your home. Not just HomePods, either, but iPads, iPhones and even CarPlay can be called with Intercom.
There’s also computational audio tuning inside, which apparently analyzes the output 180 times per second for a perfect sound. Given that Apple is also employing a long list of audio trickery here — with passive radiators and a conical baffle sending sound around the room, for example — the HomePod mini does more than its single–speaker construction really should. Apple has even had to build in a suspension system to stop the HomePod mini shaking itself off the shelf. Then there’s the proximity controls, which show you what’s playing (and let you take control) if you bring your iPhone near, and give you the same handoff controls of the HomePod in a much smaller device.
Frankly, the latest generation of smart speakers are all making huge quality–of–life improvements — and we mean that quite literally.