Voice activated
“IMAGINE this for a second,” says Mark Zuckerberg on an Instagram video clip, his lips moving strangely. I won’t hold you in suspense. It’s not Mark Zuckerberg but a ‘deep fake’ created by some artists trying to wake people up about privacy infringement.
“One man, with total control of billions of people’s stolen data, all their secrets, their lives, their futures,” Zuckerberg rattles on, his lips bouncing in a Botox-like mumble.
It doesn’t matter — the synch is good enough to deceive anyone with a bad Wi-Fi connection, conditioned to buffering video or lagging frames.
We’ve talked about the breakneck arrival of fake voices before — how scammers are likely to use these to imitate your boss, the bank manager or even a member of your family.
Plenty of start-ups have launched into this booming space, all competing to bring the most accurate versions of your voice to life for commercial purposes.
Meet the CEO of Modulate, Mike Pappas, an avid proponent of the technology, talking up the utility of ‘ voice skins’ scraped from your voice print by machine learning algorithms.
“When someone speaks into our system, our algorithm listens to the sounds they are making and immediately begins synthesising those same sounds as expressed by a different speaker,” Pappas says.
“The result is exactly what you said, with exactly the emotion and cadence of how you said it, but spoken using a totally different set of vocal cords. This is the voice skin.”
I can imagine actors would love having a digital version of themselves they can rent out.
Pixar could create entire movies using algorithms of actors rather than the actors themselves.
Chris Hemsworth could sun himself on a beach while his fake face and voice star in a CGI superhero movie.
Now merge the voices of Nicole Kidman and Keith Urban together and get a totally new, cool sound to sell a perfume.
Watch their lawyers do nothing about it because the ‘Nicolurban’ mash-up is exactly that — a creative composite plucked from a multitude of bit-sized sources that can’t be pinned down.
In fact, no one owns their own voice. Not Nicole, not Keith, not you. There’s no copyright to it. And there shouldn’t be, because any attempt to copyright your voice would infringe on freedom of speech.
So, it is no surprise then that Google and others have been able to roll out voice assistants that capture and process your vocal chords with almost zero legal resistance. But hang on a minute. How can you defend yourself against someone ripping off your voice or your face for that matter?
Why should it be open season on your identity, taking every little bit of you and creating a digital version that can be shoved into any commercial marketing campaign, even a movie, without your permission?
Celebrities have successfully taken trademarks out on their names for exactly that reason, to thwart the exploitation of their ‘brand’.
Donald and Melania Trump have protected their names.
So have Victoria Beckham and Taylor Swift.
I’m sure they’ll do the same for their voices if there’s a buck in it.
But you and me? We’re unlikely to enjoy the same protections against such cut-and-paste piracy.
Getting a trademark over the line is a very expensive, timeconsuming business, and there’s no chance that the system can support everyone trademarking themselves.
Just the lucky few get that privilege.
But there are rumblings in the musty corridors of power.
The US Congress is hauling in experts to help it create a Deep Fake Strategy because the pollies are getting nervous about the 2020 election and the possibility the electorate might have trouble discerning what they say from absolute rubbish.
One proposal would make it an offence to “maliciously create and distribute deep fakes”.
Another would force creators of deep fakes to add a watermark.
I can’t see that stopping the juggernaut, can you?
Not a chance. Peter Judd is newsroom operations manager for News Corp and a former editor of the Geelong Advertiser.