The failure of Amazon's Alexa shows Microsoft was right to kill Cortana | Hacker News Reader

The failure of Amazon's Alexa shows Microsoft was right to kill Cortana

windowscentral.com

12 pointsfranczesko12 comments

When will Siri be able to properly respond to "Give me directions to the nearest Post Office?"

If I'm driving, and ask verbally for navigation to "nearest X", it instead gives me a list of "X" on my phone for me to swipe through.

But at least they're roughly sorted by distance!

I do feel like the 'intelligence' of these AI voice assistants (perceived by the end user) has waned quite significantly in recent years. I work in tech, so maybe my perception is skewed, but when I see non-tech savvy family and friends using these things, they only really interact with the thing thru maybe 3-4 known commands. I don't think people trust these units to do much beyond "play music" or "call mom" or "set an alarm".

I remember looking at places to stay on booking.com spit. They didn't have anything directly there, but many places at the next closest city, 26km away. Except, it's 26km over a bay, and there are closer places/bigger cities when measured by driving distance.

I wonder how Apple Maps measured the distances of your search results.

In some sense the failure of Alexa is a surprise, in some sense it wasn't. I gave a talk titled "Chatbots in 2017" in December 2016 that was not all prescient

https://www.slideshare.net/paulahoule/chatbots-in-2017-ithac...

There's an expectation that technology gets better over time and one would have thought that Alexa would have gained the ability to do new things but it never did. It's a cautionary tale for the "New A.I".

There's a problem with expectations.

A voice bot is a terrible CLI with no documentation. And it's intentionally that way. I suspect by avoiding a full documentation model, they hoped to curb user expectations. "It didn't understand when I asked it XYZ, perhaps it just couldn't" doesn't fly if XYZ is on a widely documented support list. Conversely, marketers probably figured people would find it "magical" if they were encouraged to babble at the box and some things actually worked.

This neglects the users. They'll try it out early on, and develop their own set of expectations. "This is what actually worked when I first tried the box in 2016, so I won't bother asking it anything outside that bound." In order to upset that, you need dramatic changes to visibility.

In that regard, I always thought the right form factor was something like an Echo Show. It had more opportunity to directly push command visibility, with an idle screen is a good place for "Now available: try command DEF". It also has opportunities to reduce the worst pain-points of voice UIs by showing lists of options or even just the current message, rather than waiting for the slow cadence of speech to get to the point you care about.

Part of the problem may also be conflicting incentives. I suspect there's a huge tug of war in these platforms between "how do we maximize customer experience and utility" and "how do we use these as portals to encourage vendor lock-in and take up of additional services." This might also conflict with the visibility directive: if they want to promote anything, it's going to be "Alexa, do blah-blah with Prime services" even if the user base would see more value in awareness of a new local-only feature or third-party integration.

If the platforms were more open-- to the level of ordinary users being able to evaluate the rules and set up custom macros and overrides-- that helps to neuter efforts that sacrifice the customer experience for growth hacking.

That's what was so neat about Alexa as a demo was that it offered a few services such as playing music that were both possible to implement and possible for the user to feel like they understood what it was capable of doing.

They never moved from that place, however, to anywhere better, either in terms of users perceiving it as a lot or better or being something that really makes money for AMZN.

I didn’t see this backlash on voice recognition/“assistants” coming. But classic case of I never use them but sure seemed like people must have used them from what ever indications.

Paradoxically, it’s Google so can’t be surprised when they axe anything.

I think I have about five Echo Dots (bought one, got the rest for free for one reason or another) but none of them are plugged in because my wife and son are paranoid about an "always on" microphone. I might have set one up at the office but I'd have to spoof MAC addresses to get one attached to the network at work and I don't need trouble with Central IT.

I liked the way the ring would light up when I got Amazon packages, I have a very long driveway that the delivery drivers are terrified of and that was a nice feature. I also liked using it to play music while I did the dishes.

Alexa was a lot of fun in my opinion but not particularly useful. I don't think there is a "backlash" from end users, but Amazon is tired of paying for something that doesn't make a profit when they sell it, doesn't charge you for using it, doesn't really make you more inclined to subscribe to Prime, doesn't really make you more likely to spend more at AMZN, etc.

I wonder how emblematic your family is of people not wanting a hot mic around them. If that’s a widespread phenomenon I would honestly feel much better about the future.

| Yet, the division has been named as one of Amazon's biggest failures in recent times, which reminded me that, once upon a time, Microsoft also sought a piece of that surprisingly unlucrative pie.

Is a company detouring from perhaps internal empire building and excessive hiring, particularly in uncertain economic times, the equivalent of failure?

I wonder what the future of Siri and Google assistant looks like.

Siri is mostly following a different model of being an interface to installed applications or devices recognized by HomeKit that offer specific functionality.

Alexa, Google, and Cortana are interfaces to cloud computing that relays commands out to another service.

This different focus changes the "what is it good for" and "what does it cost to run" and "how is it being monetized?"

If you are using the cloud and trying to monetize compute in the cloud... that can get expensive. You have something that you want to be a profit center that is not making a profit at all.

On the other hand, if you are trying to provide an easier interface to already monetized it is a cost center that you may like to make smaller (e.g. on device speech to text) but its seen as a value added to the existing devices and ecosystem.

And so my prediction is "Siri is just fine but will only get 'smarter' as local devices get more powerful." I don't use enough google services to have a feel for how the voice on their system is used (I have a google home device that has been turned off for several years since all it would do its read back to be the top result of a search and... well... that's boring).

Reply on news.ycombinator.com