Monday, July 01, 2019 :::
Every silver lining's got a touch of grey.
There's been some consternation in the past couple of weeks about recent improvements in the ability of computer algorithms to produce fake videos that look real. I'm more — well, let's call it "optimistic", but you can choose your own labels.
The Grateful Dead
Certainly there's a long and accelerating history of technology improving the ability of people to provide evidence, to fake evidence, or to expose faked evidence, such that crimes and hoaxes alike are uncovered years after the fact. In that, this isn't new. People's faith in evidence, I think, tends to be a bit too credulous over all, but also to lag behind new technology — we trust what was good evidence a generation ago, and we're more skeptical of what was dubious then, but maybe not as much as we should have been. This particular edition is new, but the issue isn't. People will learn, eventually, to be skeptical of video evidence. This is undoubtedly a good thing.
We have seen numerous cases in the past couple years of videos that go viral, and shortly thereafter a new video providing more context essentially refutes the apparent significance of the initial video. This, too, is a new manifestation of an old problem; videos have really only been going viral on social media for a decade, but Jon Stewart in particular was notorious for taking videos and editing them selectively, and twenty years ago I received the advice that it was better to give interviews to newspapers than TV precisely because people expect the newspaper interview to be edited, and will take video as complete truth. It's been more than 25 years since NBC was exposed for construing videos of trucks exploding as showing that the trucks were unsafe, rather than as showing that NBC new how to put a car bomb on a truck.
I'm somewhat sad that security camera videos will be less reliable than they used to be; those tend not to require a whole lot of context that would be in dispute. If a security camera shows that a person was in a particular place at a particular time, that's usually all that's needed; if video shows that someone is breaking into a car, that person can try to explain why, but the factual disputes that security cameras resolve tend not to rely on context.
In most contexts, though, people put too much stock in deracinated videos, and the skepticism people learn to give them may be worth making them even less reliable than they are now.
::: posted by dWj at 12:23 PM