A recent episode of “The Book of Boba Fett” showed just how far Deep Fake technology has come, resurrecting the character of Luke Skywalker as he was 30 years ago.
Deep Fake imagery (digitally superimposing someone else’s face onto another person) has been with us for a while and there are even phone apps which can achieve a simple version of it. What was remarkable about this rendition, for me, was the voice. Luke’s voice was, as it was, 30 years ago. Initially I had assumed they has used the same actor, Mark Hamill, and somehow de-aged his voice. In fact the voice was completely artificial. Geek Tyrant describes the process which involved feeding an AI with many, many samples of Mark Hamill’s voice from that time so the AI could recite the dialogue pitch perfect.
For now, Deep Faking is process heavy and time consuming. However, processing speeds continue to improve. The situation reminds me of Autotune.
How Autotune Changed Music
Autotune was invented in 1997 and is used to correct the pitch of singing. Using a mathematical method called autocorrelation, it does this in practically real time. Music engineers considered it impractical because of the massive computational processing required. The inventor, Andy Hildebrand, invented the technology for seismic wave analysis to pinpoint drilling sites and was thus unhindered by such preconceived notions. Using a “mathematical trick”, he greatly simplified the calculations required, thus reducing the necessary processing.
Today, autotune is used on practically every popular song as part of the sound engineering to make the song “play ready”.
Real Time Deep Faking
Let us assume similar “mathematical tricks” can be discovered for Deep processes. Obviously this allows for real time broadcasting of anyone pretending to be anyone else and potentially kills voice recognition security. I can see an era of politicians using stand-ins for speeches both live and recorded. Actors will ‘de-age’ their voices in the same way they ‘de-age’ their appearances with surgery. Audiences will be given a choice for who they want to see playing the parts in movies or presenting live events. Someone that looks more like them perhaps, or someone who they trust.
Mixed Deep Faked Reality
Now consider the combination of another nascent technology: mixed reality. We are seeing smaller and less invasive versions of Hololens and Oculus-like devices and we already have glasses with bone conducting headphones. Add in noise cancelling technology and it is not hard to imagine a world where we can auto-adjust reality’s sights and sounds to whatever we want. In this world we can Deep Nude everyone we meet; we can make everyone sound like anyone we choose. Perhaps we love the sound of our own voice and want to hear it from everyone else’s mouth or there is a figure we feel comfortable and relaxed with; the ultimate ice breaker at parties.
The is where the title of this article comes from. If you are unfamiliar with the artist Aphex Twin, the video to his song, Windowlicker involves a gentleman, and some associates in white bikinis, dancing and enjoying each other’s company in a limousine. The part of the video which is slightly disturbing is they all look exactly like him.
This notion of adjusting the appearance of those around us to our preference, depending on the circumstance, I believe, will become a natural part of our lives. For people we know, we may allow them to define their appearance. For strangers, the technology will adjust them to our preference, changing their appearance to highlight “high value” individuals, based on clothing brands, known importance from online presence etc. and I will leave it to the reader to consider what this technology means for realising fantasies in the bedroom.
For those otherwise reluctant to engage in social settings, I see benefit in lowering the social barriers; it has never been easier to picture a room of people without their clothes on, but I also see it making our world even more superficial. Are we really capable of determining a person’s true worth from their shoes and LinkedIn profile? What do we lose in removing someone’s physical presence to suit our own aesthetics? The technology will encourage us not to accept people for who they are, but for who we want them to be or who they are online which, for many of us, is a very rough approximation.
I am both excited and concerned at what this technology will bring and what it will take away but I also see it as inevitable.
That CRM Blog 