After writing “Here’s What Worries Me About AI,” I realized AI presents a serious threat for documentary, news and other media creators who present programs that seek to truthfully explain “the real world.”
NOTE: I make a notable exception for “reality programming” which, though highly dramatic, is neither real nor truthful.
News presents events in the real world to help viewers understand what it going on. Documentaries deconstruct real people and real events to explain and illuminate them.
An essential element of both news and documentary media is the ability to convince the audience that what they are seeing is accurate and “real.” In the past, the line between “real” and “imaginary” was fairly clean-cut.
(Image courtesy Alex Knight.)
However, today, we can create realistic, moving, talking-head avatars that look like real people. As well, we can have those avatars speak mostly-natural-sounding sentences with natural expressions and lip movements reciting words that were first typed as text.
For example, it will be easy in a few years to create a realistic avatar of Abraham Lincoln, walk up to a podium amidst the destruction of recent battle and deliver the Gettysburg Address; all while covering it from multiple camera angles. Is it important to tell the audience that all of this – from the environment to his gestures, even to how he spoke – is just a guess? There is no known audio or visual record of any part of his speech.
Or, perhaps even more important when you are covering something controversial, how do you convince viewers that your images and sounds are not fake?
4 Responses to An AI “Truthfulness” Challenge for Media Producers
For several years already, many images have been fake, particularly from photographers who have been changing their photos, especially the sky, or the color of the ocean waters, to look like something akin to those fake photos in travel brochures.
Then we have been hearing their excuses, such as “this is the scene as the photographer saw it before” or something … Bull. That is the fake scene which nobody will ever see. Fake, not reality at that location.
George:
I agree this “image enhancement” has been going on a long time. But it took a certain skill level and tool set which limited access.
My concern is that the ability to recreate reality is now in the hands of anyone interested, which shifts the burden of proof from “most images are real” to “most images are fake.”
That shift will have major repercussions.
Larry
The genie is out of the bottle; fakers are active now, and they count on and cater to an audience that is unwilling to do any of their own due diligence as far as verification. Probably the next thing to come out in the industry will be one or more “official” verification service companies that will analyze the footage for a fee with forensic tools that are , ironically, A.I. based… and give it an accuracy rating, much like the rumor-checking web site Snopes dot com does now for the junk claims Uncle Joe passes around your family dinner table.
But then we get recursive, as in; who will verify that the verifiers are legit, and not bought off? For example, does anybody really believe in any awards or ratings from J.D. Power?
In the near term, only things that are demonstrably live and in real time might be believed fully, and with time, we’ll be able to fake those as well.
Joe Haldeman nailed today’s deepfake crisis way back in 1974. I remember a scene in his classic SF novel: “The Forever War”. A space-traveling Earth soldier arrives back home and is interviewed by a chat show. Our hero is a man out of time; thanks to relativistic time dilation, he’s been away from Earth a hundred years while aging and experiencing only about one, and everything around him is strange and different, culture itself has left him far behind. During the chat show interview recording session, the producers ask him a few questions to “calibrate their cameras”, then proceed with the show. He gives his interview truthfully, expresses uncomfortable opinions and facts about the war and how it has been conducted and what he thinks of Earth now versus his own time… and everything he says is uncomfortable, even threatening, for the government of that time. When he gets to his hotel room and sees the finished show on the air, every word of his interview has been, not just edited, but completely re-built into pro-government and pro-culture propaganda, image and words and voice, perfectly synthesized from the “calibration samples”. The only thing they left that was true, was his name.
Mark:
Sigh…. frightening. But sadly, not out of the realm of possibility. (Especially the “verification” part.)
Larry