One of the most vexing challenges for artificial intelligence is language interpretation and translation, not only because there are so many languages but because machines still have a devil of a time figuring out all the nuance and context of human speech and the written word.
My dad, Tom Uhler, is trying to learn Spanish. His sister, my aunt, lives in Spain, and he figured it’d be easier to get around when he visits if he had more control of the language. He’s signed up for a course, but it’s slow going. So when he needs to say something in Spanish, he’ll often turn to Google Translate on his phone.
When he asked, “Where’s the bathroom?” Google, matter of factly replied: “¿Dónde está el baño?”
But when he asked for a translation of “Give me a ring when you are ready,” something English speakers say all the time, Google had a little more trouble. The phrase is too idiomatic for a machine. The phone replied: “Da me un anillo cuando estes listo.” Google thought my dad was talking about the kind of ring you put on your finger.
Translation has come a long way in the last few years, but there’s still a lot of work to do. And what scientists really want to develop is something like that ugly creature from “The Hitchhiker’s Guide to the Galaxy,” the Babel fish.
“That is the holy grail,” said Andrew Ochoa, CEO of Waverly Labs. His company plans to launch a product in November called Ambassador. It’s an over-the-ear translation device that will require Wi-Fi, a smartphone and Bluetooth.
“As the information is being captured via your earphones and earbuds, that data signal is being sent through your smartphone to the cloud where it’s processed and then relays back through the smartphone to the cloud,” Ochoa said.
The Ambassador is going to sell for $149.
Bigger tech firms like Google and Microsoft are investing billions in artificial intelligence, including language translation.
Google recently revealed it’s working on something called Translatotron, which is supposed to convert what you say into another language in your own voice simultaneously.
Dave Kemp runs the website FuturEar. He thinks Google might find translation’s holy grail before anybody else.
“Google is a search company, but they’re also a data company,” he said. “And what they’re so good at is taking their data and basically learning from their data.”
Google didn’t want to talk but said in an email that it’s not trying to profit directly from translation. Instead, it sees mastery of world languages as a way to extend its global reach.
Kemp thinks translation will likely end up a moneymaker for Google and explained one way that might work.
“You know if you say, ‘Navigate me to the best Parisian cafe’ if I’m in Paris and somehow that’s tied to not only the location settings but the translation,” he said.
If Google perfected that piece of translation, it could sell more ads.
Lorien Pratt, chief scientist at an AI firm Quantellia, noted that what Google is perfecting is translation for everyday life, casual conversation.
Translation tech has to advance much further for conducting business where mistakes could cost a lot of money.
“And then there’s jargon-specific translation, so like if I’m a medical practitioner or a lawyer or something like that, obviously that needs extra intelligence,” Pratt said.
That’s the stuff businesses can’t get wrong because the consequences could be far greater than using the wrong noun when you’re looking for the bathroom.
Pratt said the current technology is advancing quickly, but the true Babel fish is still just science fiction.