Amazon Alexa is so smart it's stupid

Amazon's intelligent personal assistant has amassed more than 15,000 skills. If only we could find them

Amazon Alexa is so smart, it's stupid

The Amazon Alexa platform is on a roll, now crammed with more than 15,000 “skills,” up from 10,000 in February, as reported by Voicebot. While this sounds amazing, the reality is that the vast majority of Amazon Echo customers don’t know 99.999 percent of those skills exist. Worse, there seems to be no viable way for anyone to discover what all those skills are.

So while Amazon keeps fanning the flames with developer outreach like the upcoming Alexa Dev Days, Amazon’s far larger problem is Alexa skill discovery, not skill development.

We’ve grown fond of pillorying the typical mobile app for failing to sustain its user’s interest. You know the woeful stats: The average app loses 77 percent of its users within three days of installation, and a whopping 95 percent of its users within 90 days.

Guess what? For voice apps, aka skills, the numbers are even worse.

If an Alexa skill is developed in a forest…

Earlier this year VoiceLabs released a report showing that 97 percent of voice apps are completely forgotten within a week of their first use. From personal experience, this sounds about right. My kids loved asking “the magic 8-ball” to tell their fortunes with Alexa... for about a day. Over the past year, the only Alexa skills that any of us seem to remember are “Alexa, set a timer for…”, “Alexa, play Hamilton the musical,” and “Alexa, play ______ book from Audible.”

Don’t get me wrong, these skills alone are enough for me to feel like the purchase was worthwhile, but that leaves 14,997 Alexa skills untapped and forever unknown.

Not content to rest on its 15,000 laurels, Amazon has now launched Alexa Dev Days, with the goal to help developers “Engage in hands-on sessions where you will learn to create voice experiences and hear the latest on UI design and industry trends.” And that is only the latest step Amazon has taken to stoke developer interest in Alexa, following upon AWS credits, free Echo devices, and cash prizes.

What Alexa Dev Days doesn’t do, however, is make it any easier to actually hear about any of those skills. Let’s say Alexa skills jump from 15,000 to 150,000. Will anyone notice?

Discovering Alexa

My InfoWorld colleague Andrew Oliver just wrote about 12 hot technologies that haven’t lived up to their hype, and given the massive gap between Alexa skill development and use, Alexa might be cause for adding a 13th to his list. In fact, Oliver’s commentary on the sorry state of chatbots suggests a future for Alexa, too: “[C]hatbots will only be useful as an interface to a search engine—as the thing that asks follow-up questions to refine your search to find exactly what you’re looking for.”

Today, Alexa skills are somewhat like obscure command line directives: “Alexa, ask the Magic 8-Ball if I’ll ever remember any of these skills.” Amazon has built intelligence into Alexa that makes it easy for me to use Amazon services (e.g., buy replacement air filters, play Audible books, etc.) but has left much of the skills territory to third-party developers. This would be awesome if, as mentioned, it were easier to uncover these skills.

But wait, you say, there’s a website devoted to helping you find new and exciting Alexa skills. That’s correct. Not only to discover but then enable a new skill—Alexa skills nearly always require enablement and then a special set of voice commands to trigger them—you have to visit a website. It’s a voice interface that requires you to type into a desktop web interface. Kinda silly, don’t you think?

True, Amazon also has given us the Alexa Skill Finder for uncovering Alexa skills (once you’ve enabled the skill via the website) by voice, but it’s clunky at best. Even those who laud the idea of a Skill Finder wonder why the functionality isn’t built more natively into Alexa: “I only wish this was built into Alexa directly so I didn’t have to ‘ask skill finder’!” Others are less enthusiastic: “The app only appears to be useful. In fact, it only allows to answer with a Yes or No when asked ‘Would you like to hear more skills?’ DOES NOT allow you to ask questions. It shuts down.”

Even once you have a skill discovered and enabled, some of the most obviously useful skills can be a kluge to operate. Hailing an Uber, for example, is easy: “Alexa, order me an Uber at my home.” The prerequisite to that command, however, is work you first have to do on the web: link accounts, enter a home address. It’s not terribly hard, but it’s also not terribly AI-fantastic. Other skills, like ordering a pizza from Dominos, should be easy but really aren’t: “Alexa, open Domino’s and place my Easy Order.” You have to call the Dominos skill, then you can only ape a set order. Want to go crazy and get extra pepperoni? Forget it.

Hello Google Assistant?

For this reason, Google may actually be able to offer Amazon hefty competition, despite only boasting a mere 378 skills for the Google Assistant today (according to Voicebot data). Google has been training its AI for years with petabytes upon petabytes of data. It understands search intent better than any other company.

For information retrieval, which relies on search, Google is arguably already better than Amazon’s Alexa. Google simply needs to add select, core functionality that is easier to discover and it won’t need 15,000 Voice Apps (the equivalent of Alexa skills). Simply figuring out a natural way to order pizza would be enough for our family to bow reverently before the Google Home device.

Amazon may be killing it on Echo sales, and seemingly has cemented an unassailable lead in voice apps for its platform, but the truth is less clearcut. Developers may be flocking to Alexa but consumers will ultimately go where they don’t have to use the equivalent of a Unix command line. The company that makes it easy to use voice apps will be the one that wins big, and Google is just as well-positioned as Amazon in this area, if not more so.

Copyright © 2017 IDG Communications, Inc.

How to choose a low-code development platform