Page 17 - Fortune-November 01, 2018
P. 17

ar tifi c i al  intelli gence  + t h e b a t t l e  f o r v o i c e





         2013—about the same so-so level of accuracy   THE FUTURE OF VOICE
         human listeners achieve. One of the great
         recent triumphs in the field has been teaching
         the engines to filter out nonspoken background
         noise, a distraction that can frustrate the keen-  ASK ALEXA something easy,  Ostendorf, a UW professor who
         est human ear. These systems reach this level,  like, “Play ‘In My Feelings’ by  advised the team, “is open-ended
         however, only when the question is simple, like,  Drake,” and she gets it right about  conversations where someone
                                                       every time. Try to engage in an in-  suddenly asks why or how some-
         “What time is Mission: Impossible playing?”   depth conversation on the mid-  one did something. The machine
         Ask the Google Assistant or Alexa for an opin-  term elections, and you’re likely  has too many choices.”
         ion or try to have an extended back-and-forth  to find her tongue-tied. A.I. sci-  Even so, voice is getting
         conversation, and the machine is likely to give  entists say the holy grail of voice  smarter. Najim Dehak, a com-
         either a jokey preprogrammed answer or to     recognition is to make computers  puter and electrical engineering
         simply demur: “Hmm, I don’t know that one.”   so smart that they can have a  professor at Johns Hopkins, uses
                                                       long, nuanced conversation with  voice recognition to identify irate
                                                       humans. The better computers  customers. He has fed his voice
                     O CONSUMERS, voice-driven         get at this, the more useful they  software recordings of 2,000
                     gadgets are helpful and some-     will be as smart assistants.  negative and 2,000 positive
             T       times entertaining “assistants.”  $1 million reward for any univer-  customer conversations gleaned
                                                         Each year, Amazon offers a
                                                                                  from a company call center. By
                     For the tech giants that make
                     them—and keep them con-           sity team of programmers who  focusing on keywords—pro-
                                                       design a piece of software that  fanities are a dead giveaway for
         nected to the computers in their data cen-    can keep a conversation with a  negativity—the machine can
         ters—they’re tiny but extremely efficient data
                                                       human going for 20 minutes with-  identify which calls are from
         collectors. About 60% of Amazon Echo and      out getting confused or straying  teed-off customers. “The idea,”
         Google Home users have at least one house-    off topic. No one has yet to claim  says Dehak, “is that managers
         hold accessory, such as a thermostat, security  the full prize, but a team from the  can listen to those negative calls
         system, or appliance, connected to them,      University of Washington came  to know if the service agent did
         according to Consumer Intelligence Research   close last year and walked off with  something wrong or needs more
                                                       $500,000 for keeping the chatter  training.” The software is in test
         Partners. A voice-powered home accessory      going for 10 minutes on average.  phase, but Dehak says it should
         can record endless facts about a user’s daily  “The biggest challenge,” says Mari  be available within a year or so.
         life. And the more data Amazon, Google,
         and Apple can accumulate, the better they
         can serve those consumers, whether through
         additional devices, subscription services, or
         advertising on behalf of other merchants.     company will pull in $4.61 billion from digital advertising in
           The commercial opportunities are straight-  2018—a spokesperson says it does not currently use Alexa data
         forward. A consumer who connects an Echo to   to sell ads. Google, counterintuitively, considering its giant ad
         his thermostat might be receptive to an offer  business, also isn’t positioning voice as an ad opportunity—yet.
         to buy a smart lighting system. Creepy though  Apple, which loudly plays up the virtue of its unwillingness to
         it may sound to privacy advocates, the tech   exploit customer data for commercial gain, claims to be ap-
         giants are sitting on top of a treasure trove of  proaching voice merely as a way to improve the experience of
         personal data, the better with which to market  its users and to sell more of its expensive HomePods.
         more efficiently to consumers.
           As with their overall strategies, the tech gi-          ESPITE ONE OF AMAZON’S early selling points, what
         ants have different approaches to the data they           people aren’t asking their devices to do is help
         collect. Amazon says it uses data from Alexa to  D        them shop. Amazon won’t comment on how
         make the software smarter and more useful to              many Echo users shop with the device, but a
         its customers. The better Alexa becomes, the              recent survey of book buyers by consulting firm
         company claims, the more customers will see   the Codex Group suggests that it’s still early days. It found that
         the value of its products and services, includ-  only 8% used the Echo to buy a book, while 13% used it to lis-
         ing its Prime membership program. Although    ten to audiobooks. “People are creatures of habit,” says Vincent
         Amazon is making a big push into advertis-    Thielke, an analyst with research firm Canalys, which focuses
         ing—the research firm eMarketer projects the   on tech. “When you’re looking to buy a coffee cup, it’s hard to





         118
         FO R T U N E. CO M //  N O V. 1 . 1 8
   12   13   14   15   16   17   18   19   20   21   22