Helpful sounds from the abyss.
OK, Google, listen for a moment…
“A man in red and black was riding a surf board”, I heard someone say.
“Suddenly, he disappeared, and I could hear a faint shriek behind the walls”, added another, quietly.
An Indian voice spoke, calmly, as if he hadn’t just heard the scream or noticed the missing man: “A firefighter stands on top of the firetruck.”
“Look closely to the left”, came a voice from the growing crowd. “And I will tell you how to find the hidden treasure”, she said. “It’s a lot of gold.”
“Nobody knows where it’s from”, a fellow noted casually.
A German spoke next, off topic yet undeniably correct: “Never advertise razors by shaving a monkey, it’s cruel and results in lawsuits.” The inflections of his voice were tinged amusement and perplexity, like even he was a tad surprised by his words.
“Listen, don’t believe what he says!”, someone implored. We trust they weren’t, in fact, a fan of monkey-shaving.
“What have you caused all of this trouble for?”, an American woman said, perhaps rebuking the German for even bringing up the inhumane subject.
Next a deep voice with a slow, thick southern drawl added helpfully (as you’ll soon realize): “Mouuuunt Eeeverrrrest is foouuund on the interrr-naytional boarr-der between Jina and Neeeepal.”
“Enemy spotted.” announced another.
“An army is coming!”
The above is a loosely coupled and highly concocted narrative, yet based on actual events. Sort of.
The text in italic is just low-grade literary adhesive to connect the dots between what was indeed said (though not actually in that particular order). The speakers in this tiny tale weren’t really participating in any sort of fiction, but rather are part of a far larger and more important story.
These people — volunteers, all of them — were recording their voices reading odd little snippets which contain an expansive array of sounds and words. Other volunteers then listen from across the country or the planet and ensure the audio recording matches the sentences used as prompts.
It looks like this:
When you put all this together, what do you get?
An ambitious project to protect the openness of the web (I’ll explain).
The project is building a massive database of recordings from people of all demographics, and is being used to develop open-source AI which can understand and produce human speech. Capabilities such as these are essential for “smart speakers” like Google Home, and Amazon Alexa.
If the project succeeds, you might see a proliferation of open source smart devices capable of conversing in human languages. (The first may already be here: a startup called Mycroft is producing their Mark 1 based on open source software and hardware.)
These devices could work even in places where the existing cloud-based services haven’t yet gone, whether that’s a desert island or the confines of your own home (without sending your voice across the NSA warrantless wiretapping tool we call the Internet, en route to rooms in datacenters which hopefully aren’t anything like AT&T’s infamous 641A).
And if the project fails?
We risk the Internet applications of the future — the way we communicate with others or interact with the world electronically — becoming dominated by a small handful of powerful companies like Amazon and Google. By controlling the terms of use or levying charges for every utterance decoded, these companies could pose a powerful threat to a truly open and free web.
In some ways, taking on the likes of Amazon’s Alexa is similar to the early browser wars, when the web nearly became a landscape dominated by Microsoft’s Internet Explorer.
Luckily, the hoard of volunteers training open source software is being organized by none other than Mozilla, the nonprofit best known for producing the Firefox browser which successfully did just that when it knocked IE of its throne.
If you have a few moments, join the battle for a free and open internet at https://voice.mozilla.org/.
“OK, Google, can you hear the voices multiplying?”
They sound like this:
👏 Thanks for reading! Please share and add your voice. 👏