“OK Google,” What Are the Problems with Speech Recognition Technology?

How Amazon and Google’s big bets on the smart speaker market affect their customers’ privacy

Benjamin Powers
6 min readJan 18, 2018
Google Home units on display during CES 2018. Photo by Alex Wong/Getty Images

Speech recognition technology and the voice user interfaces (VUIs) we use to engage with it have gotten so good that they now make errors only about 5.5 percent of the time. That’s about the same error rate as a human.

And this speech recognition technology has advanced rapidly — particularly in the past 10 to 15 years —and is becoming commonplace in the bedroom, kitchen, and the rest of your house. It is light-years beyond its first iterations in the 1990s, when one of the earliest commercial products was Dragon Dictate, a typing software released in 1990 that was error prone and cost $9,000. (Google alone has reduced its speech-recognition error rate by 30 percent since 2012.) Both Google and Amazon have jumped into the market with gusto, with Google Home and Amazon Echo now accounting for the two largest market shares in the industry. Microsoft also recently released its own smart speaker featuring an AI assistant named Cortana, and unsurprisingly, Apple supposedly has a similar device in the works. All in all, the value of the virtual assistant market, and the speech recognition it necessitates, is expected to exceed $3 billion by 2020. What…

--

--

Benjamin Powers

Benjamin’s writing has appeared in Rolling Stone, New Republic, and Pacific Standard, among others. You can find all of his work at benjaminopowers.com