A leftover from last week in case you missed it, to cleanse the palate. There are nerdy explanations online for how this was done (plus another in the clip below, which gets better as it goes along), but the gist is that Samsung’s AI people taught their computers how to combine video of Person A with still images of Person B to produce video of Person B, matching “landmarks” on Person B’s face to the footage of Person A. The more still images of Person B that the computer can access, the truer the likeness and the more realistic the synthesized “video” footage will be. But it can be done already, albeit crudely, with as little as one still image, as you’re about to see. Paintings can be brought to — not life, exactly, but something closely resembling it.
Given that many millions of people have hundreds of photos of themselves posted publicly on their Facebook or Instagram accounts, the technology is *already* at a point where it can produce a pretty convincing yet utterly fabricated video likeness of, well, practically anyone. Your children and every generation that follows will grow up understanding from their earliest moments that they can’t trust anything they see on a screen, ever.
My libertarian side is whispering to me that that’s not the worst instinct to cultivate in a population. My less libertarian side is whispering ehhhhhhhhhh.
The thought I keep having about this technology is why any mainstream private entity would think to develop it. I understand very well why governments would want it. I understand why pornographers might want it. To what virtuous end is Samsung hoping to apply this, though? Is it purely a sh*ts-and-giggles thing for entertainment, so that Marilyn Monroe’s likeness can be cast in new movies? That benefit seems so colossally dwarfed by the technology’s potential for mischief, with such terrible consequences potentially for social cohesion, that I’m straining to believe there must be some nobler purpose that I’m missing. What is it? What important function will this serve?