The Next Crappy Technology That Will Take Over The World

According to the Times, voice recognition is the next frontier in technology. As a five-year veteran of voice rec software, I'm extremely skeptical.

Steve Lohr and John Markoff write that voice recognition technology — already in use everywhere from cars to phone-based customer service systems — could one day replace receptionists or even medical assistants. Despite their claim that the technology is about to revolutionize society — they write, "the prospect, according to scientists and economists, is not only that artificial intelligence will transform the way humans and machines communicate and collaborate, but will also eliminate millions of jobs, create many others and change the nature of work and daily routines" — their article reads like a compendium of silly voice rec gaffes. There's the misogynist medical transcription software: says one doctor, "It's unbelievably better than it was five years ago. But it struggles with ‘she' and ‘he,' for some reason. When I say ‘she,' it writes ‘he.' The technology is sexist." There's the automated Arabic-to-English translator used in Iraq:

When a soldier asked a civilian, "What are you transporting in your truck?" the Arabic reply was that the truck was "carrying tomatoes." But the English translation became "pregnant tomatoes."

Then there's an MTV producer's experience with a "virtual personal assistant" called Siri:

Recently, he asked Siri for the location of a sushi restaurant he knew. Siri replied with directions to an Asian escort service. "I swear that's not what I was looking for," he said.

I've been there. I've used voice recognition software (currently Dragon NaturallySpeaking 9, if you care) for five years due to wrist problems. The he-she problem is constant, but probably Dragon's most egregious screwup was in graduate school, when I tried to dictate the sentence, "That was a great story!" in a letter to a classmate. Dragon has never liked exclamation points — it rendered the sentence as, "that was a great story Asian porn."

I pretty much have to correct every single sentence by hand after I dictate it. Here's what it looks like if I don't: pretty Fox up. Have I mentioned that dragon is extremely prudish and doesn't understand curse word? it also has a really hard time with plural, capitalization, and all were, and proper in an (it especially hate Morocco, more than one rendering his name as Osama). On the plus side, it never makes typos, but as you can see, because they can make a sentence impossible to understand.

[Translation: pretty fucked up. Have I mentioned that Dragon is extremely prudish and doesn't understand curse words? It also has a really hard time with plurals, capitalization, small words, and proper nouns (it especially hates Barack Obama, more than once rendering his name as Osama). On the plus side, it never makes typos, but as you can see, its mistakes can make a sentence impossible to understand.]

In theory, you can go back and make corrections by voice as well, but I never do this anymore because it's just too slow. That's another problem with voice recognition — it takes up a huge amount of RAM and processing power, and seems to eat up more computer resources with every new version. I think my computers aren't keeping pace with the program's insane requirements, so in the five years I've been using voice rec, it hasn't gotten any more accurate — and I think it's gotten slower. It's also capable of doing really awful things — sometimes it thinks I said, "close window," and I lose whatever work I haven't saved. If I want to answer the phone or run some water or do anything else that causes ambient noise, I have to make sure to turn the microphone off, or the program will freak the fuck out, sometimes causing my whole system to hang for minutes on end.

I still use voice rec because it has resolved my wrist issues (which come back the minute I think I can get along without it), and because when it's not misbehaving, it is kind of fun — for instance, I have a wireless mic, so I'm dictating this part of the post from my kitchen. Whoo! However, given its many problems, I'd be unlikely to trust voice recognition to, say, diagnose a child's illness — one potential application apparently in the works. Then again, if the iPhone is any indication, maybe this is just how technology advances now: a new product comes out, it totally sucks, but we just keep using it until we're totally dependent and know no other way of life. Pretty Foxing awesome.

Image via mathom/Shutterstock.com..

Computers Learn To Listen, And Some Talk Back [NYT]

Earlier: Do You Talk To Yourself?