Last year, we saw former US president Barrack Obama in a very convincing YouTube clip, only to find out it was the result of an advanced lip syncing tech. Today, researchers from China’s biggest search engine announced a program that can clone any voice and reuse it as it sees fit #softwaremagic
Baidu researchers have developed an AI program that needs only to hear seconds of your voice to replicate it and alter it. Basically, it can put a different accent on your voice or remodel it to sound female or male. That sounds deceptively easy, mind you. Until their breakthrough, Adobe VoCo was able to mimic a voice after 20 minutes, not seconds, of input!
According to New Scientist, the program created one voice capable of fooling voice recognition software with 95% accuracy! That’s huge! Even humans were deluded into thinking the cloned voice was real. If this technology falls into the wrong hands, the consequences would be dire, of course. It would be easy to use AI to manufacture speech and create fake news.
Keeping that in mind, we should also consider the benefits of such tech. Parents could use it with their kids when they’re away, ensuring they have a soothing voice to listen to. Even better, people who have lost the ability to talk would be able to hear themselves talking again and facilitate interactions with their dear ones. Digital assistants would talk in different tones, instead of using the same universal voice.
In the end, like all technology breakthroughs that blur the line between humans and computers, there are ethical issues to consider. Let us just hope that the right decisions will be taken and AI will be the best not the worst thing to happen to humanity, as Hawking said.