A turning point in Personal Assistants

The dream of the AI “personal assistant has had many false starts over the years.  I clippyworked on the ill-fated Microsoft Office Assistant (e.g. Clippy) years ago, but the dream hasn’t died.

Apple kicked the market into gear with Siri in 2011, followed by Google’s Google Now in 2012, and Microsoft Cortana in 2014.  It’s been fascinating to watch the different approaches companies have taken, and they feel like real progress.   The availability of cloud processing, access to vast data from users and search, and ubiquitous mobile devices made a big leap forward possible.   But they still struggle to be truly natural, conversational, and useful UIs.

In 2014, seemingly out of nowhere, Amazon initiated a new phase with the introduction of the Amazon Echo and Alexa assistant.   I was surprised to see Amazon enter with such a competent offering – I’d assumed they lacked necessary assets.  But they’d been quietly acquiring and building them in-house.

The Amazon Echo & Alexa

echoThe Echo and Alexa were surprisingly robust.   Echo’s far field voice technology with simple and cheap hardware (BOM of about $60) were better than I’d have thought possible.   Their intent recognition technology was very rudimentary – Alexa is essentially a voice driven command line – but it proved good enough to enable a wide variety of scenarios and tasks reliably.   While Echo sales are unknown (estimated around 4M), their product has been intriguing to millions of customers.

Their go to market strategy has roughly followed that laid out by Charlie Kindel (ex Microsoft, now on the Amazon Alexa team) in his blog:   The best way to build a platform is to not build one.   Enter with a compelling product to attract users.   Echo is marketed as a smart music player, along with clever assistant tasks like timers and unit conversions.  Later Amazon added home automation control and a voice-app platform.   Now they’ve begun to attract developers and opened the assistant to 3rd party devices – although we have yet to see a truly compelling example of either.

The tech world is in  love with Echo/Alexa, and this excitement has obscured clear analysis of its real strengths as weaknesses:

Strengths

  • 1st rate far field voice recognition
  • Very low response latency
  • Excellent performance on limited set of tasks
  • Great music support for Spotify, radio with tunein, and music library with Amazon Prime Music.
  • Rudimentary shopping scenarios.
 Weaknesses

  • Voice performance degrades rapidly with multiple speakers and background noise.
  • Single user only – no multiuser scenarios.
  • Doesn’t work with iTunes or Google ecosystems.
  • Very limited services (e.g., tasks, calendar, email)
  • No support for long tail tasks (e.g., search)
  • Assistant not available across more endpoints.

As someone who uses all of these products, I’d share some observations.

  • Speech recognition – while amazingly good, it’s still not good enough for non techies. It frustratingly fails to work for my un-SR-trained family.  Speech recognition requires a very quiet kitchen, so often commands must wait for everyone in the room to shut up to work.
  • The core value proposition of the Echo is still music and radio. A variety of other tasks supplement this value, but surely wouldn’t sell the device on their own.
  • My house is fully loaded with Smartthings, Hue, etc. I’ve configured voice control with Alexa, but this functionality has failed to “stick” with anyone in the house.
  • Amazon’s lack of a full services suite really limits Alexa. It can’t give me schedule information from my calendar, send or read emails, or give me traffic alerts (some are possible via complex IFTTT procedures).   They’ve added some links to Google Calendar using public APIs.
  • Shopping – the core reason Amazon built Echo – hasn’t really come to life yet. Most users I’ve surveyed haven’t actually bought much from Amazon using Alexa.
  • Lack of multiuser and user recognition.   To fix this I’ve merged my whole family’s music libraries into a master account, but that was a lot of work.   This limits the utility of many scenarios.

Nonetheless, the Echo is amazing, but has a lot of room for improvement.   Or competition….

NEXT:  Why Amazon is doing Alexa and Echo, and why Google had to respond.