WHY THIS MATTERS IN BRIEF
Artificial intelligence is gaining new skills all the time, and now it’s creating scarily accurate impressions of celebrities.
The Lyre Bird is a unique bird that lives in the Amazon Rainforest because it’s especially skilled at copying the sound of its surroundings. It’s so successful in fact that in the 1990’s it rose to fame as the sad symbol of the decline of the Amazon when it was first heard perfectly copying the noises of the chainsaws and car alarms of the loggers who were busily destroying its habitat. And now it’s probably mimicking the sound of WiFi routers… but that’s a different story.
Now, a Canadian Artificial Intelligence (AI) and Machine Learning company, LyreBird, has received ringing sound bytes and endorsements from Donald Trump, Hillary Clinton, and Barack Obama discussing its technology. But it’s not actually them, it’s an AI powered imitation that’s scarily accurate – even if it does sound jerky, but over time, as we recently saw with DeepMind’s own WaveNet Project that’ll vanish.
LyreBird uses machine learning to produce realistic sounding voice audio and it does it by listening to just one minute – a whole minute – of audio. I know, why so long you ask?
By analysing the voices of all three celebrities, and others, the new system was able to re-produce their voices with amazing accuracy. LyreBird works by taking a sample of any voice and by analysing their waveforms and the cues that “make the voice unique,” it then takes these reliable deviations from the platonic ideal of an English voice and tells its voice synthesis component to make those same adjustments to its audio “wave-forms” as those sound curves are generated. By doing this with enough specificity, the voice generated can have not just the general sound or accent of a person, but the quirks and minor tics, as well. And that’s what makes it so accurate.
However, in the era of fake news, which Facebook are throwing more and more resource at – some of which will undoubtedly be fooled by the new system – some people are questioning the new system’s potential impact on society
It was only a few months ago when I reported on Voco, a new Adobe tool that let’s you do with voice what Photoshop allows you to do with images – basically photoshop voices on demand – and I can’t help thinking that if you combine Voco with LyreBird you’re going to get some awesome audio – and some huge biometric and security problems. Putin declaring war on the world? You could probably create that audio today and we’re only at the beginning of this tech. You get the idea… but just as it could be used for “evil” it could also, like most technologies, be used for great good, for example, it could be used to give people with voice boxes, after they’ve lost their voice boxes to cancer, their real voices back.
As we say in the futurists game – technology is unbiased, it’s humans who use it for good, or evil… but there is one person who needs to be very, very careful and that’s Alec Baldwin – his Trump impression is awful and he might just find himself getting replaced. Oh the joy, er travesty.