News in general, but television news in particular, has been on the giving and receiving end of “fake news” accusations at least since the 2016 campaign season.
Even if the media or politicians were partaking in this nefarious manipulation of public opinion, those who subscribe to the notion that the truth ultimately wins out in a free market of ideas could take some solace in the belief that reality is eventually revealed through seemingly endless rounds of reporting, analysis and debate.
That’s all well and good in a Fake News 1.0 world where the battle of ideas isn’t skewed by undetectable gaming of the process.
However, I fear that A.I. technology is about to usher in Fake News 2.0, an era in which it will be impossible to believe one’s own ears and eyes. Look no further than LyreBird, an A.I. app that in the words of the company “allows you to create a digital voice that sounds like you [or anyone else, for that matter] with only one minute of audio.” If you are unfamiliar with LyreBird, you might want to check out https://youtu.be/YfU_sWHT8mo.
Now sprinkle in a couple of other A.I. developments. The first is a paper presented at SIGGRAPH 2017 entitled “Audio-Driven Facial Animation by Joint End-to-End Learning of Pose and Emotion,” which describes the use of machine learning to drive a real-time 3D facial animation from an audio input with low latency.
The second is a paper by Berkeley A.I. Research (BAIR) researchers entitled “Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks.” The paper describes “an approach for learning to translate an image from a source domain X to a target domain Y in the absence of paired examples.”
Jun-Yan Zhu, the lead author of the paper, has posted a short YouTube video of this approach being used to transform a horse into a zebra.
The obvious danger is that some ambitious reporter one day will use perfected versions of these technologies to create the A.I.-based equivalent of “Jimmy’s World.”
However, it seems to me that the less obvious, but even more pernicious effect will be the use of these technologies to mislead reporters and editors by government entities, industry and others with an ax to grind or reason to deceive.
As one who has spent years reporting on emerging technologies that TV broadcasters could use to better inform the public –Ku-band SNG, IP newsgathering, dual polarization weather radar and mobile tech to name a few– I am truly saddened that the next wave in technology will make it extremely easy to lie to the public.
How long will television news endure when viewers no longer can believe their own eyes and ears?