So this is basically a regression to a 19th-century level in terms of being able to trust and understand reporting on the world beyond our own front door. People managed before photographic and video evidence was a thing; you could use eyewitness reports from trusted friends and news on the official telegraphs, to the extent that those were trustworthy. But it's certainly still a big step backward from the 20th century, that brief window of time where it was much easier to record physical evidence of an event than to fake it.
Photographic evidence has been subject to manipulation before computers were even a thing, more so after Photoshop became widely available. There has always been forensics for that, which will continue to evolve.
I think the issue with trust is rooted elsewhere - in social relations, politics, and not in AI generated content.
It has, but it used to take a lot more skill to manipulate a photo than to take a photo, and convincing video manipulation was even harder. I'm also skeptical that forensics will be able to keep up, because of the basic principle of antagonistic training -- any technique forensics can use can be applied back into improving the pipeline that generates the image, defeating the forensic tool. That certainly wasn't the case in the 20th century.
