Re:Stopping deepfake news with an AI algorithm that can tell...
by eie7eu3hrruur on Monday August 10, 2020 @05:28AM
(#24107234)
This would be more concerning if the public was actually well informed currently. At present, most media companies have wildly lowered their content standards, normalized clickbait titles that sometimes outright lie, all the while ramping up "it's an op-ed so it's not a lie it's just an opinion!" pieces. At the same time we've never been more interested in what government agencies have to say including what they have to say about the government itself. This world of a semi-informed populace doesn't exist, the public is already choosing information from sources that appeal to their emotions rather than logic. What person on the street actually has a nuanced understanding of an industry they're not participating in? How many have any real idea what government contractor work actually looks like in those same industries or understands what the warzones their government operates in are actually like? Most don't even understand other cultures that vaguely resemble their own, just look at how few westerners have the vaguest understanding of sex culture across asia. The only thing people know anymore is memes and rhetorical factoids. You can't poison a toxic well.
Re:Stopping deepfake news with an AI algorithm that can tell...
by JoshuaDavid on Monday August 10, 2020 @03:39AM
(#24106602)
While video is currently unlikely to be entirely fabricated at this point, you might be surprised at what can be accomplished with camera angles, choosing the start and end times of your clip, and cuts and reordering of clips: for a silly example of this, see "Obama sings Call Me Maybe"[1]. Videos can also be staged with varying levels of convincingness.
If you are digging for evidence, the source of your information must be taken into account. Deepfakes will probably be detectable in retrospect, even if they are not detectable at the time of publication, so if you stick with sources with a long history of only publishing real information, you're probably still fine even in the age of deepfakes.
The arguable lack of journalistic sources with a demonstrated history of integrity is its own problem, but not one that is made significantly worse by the presence of deepfakes.
As it becomes possible to fake more and more stuff, we're going to lose that. Different silos are going to have different, meticulously supported versions of the facts with an equivalent degree of direct-evidential and reputational support. And it's not clear to me as an individual one ever disentangles this.