The technology is uncanny — and incredibly frightening.
As many of us who spend time on social media are painfully aware, it doesn’t take much subterfuge for an outlet to pass off fake news as the real deal. Suspect web domains and sensationally biased headlines regularly go unnoticed by those who want to believe something is true no matter how outlandish or groundless.
While steps are being taken by Facebook and Google to fight the proliferation of these stories that profoundly damage the media’s credibility and the integrity of their subjects, a recent innovation will make the fight even more difficult.
A research team at the University of Washington has harnessed the power of machine learning to study facial movements of a speaking subject then re-create the video so that it appears they’re saying most anything a person would like them to.
The innovation may sound curious or even exciting in the abstract, but when viewed in action, one could easily see how this could be harnessed to spread very convincing and very fake evidence to bolster a narrative.
Fortunately, the technology is far from perfect, and humans, due to their constant observation of others speaking, maintain a very sharp ability to discern any inconsistencies in recreations such as these. However, the perfection of this technology, like so many others, is just a matter of time. Once the technology becomes more reliable, it will then become cheaper, allowing most anyone with the inclination and dedication to target public or private figures as de facto sock puppets, speaking whatever words their programmer wishes them to.
One more silver lining is the ability to the technology to not only create fake speech but also detect it. No doubt that will be a very desirable utility as the innovation is levered by unscrupulous people wishing to drive public sentiment wherever they please.