| at the moment, there are telltale signs of deepfakes if you know what to look for. various video editing issues, image glitches (visible in the obama/ahmadinejad photo in the article), lighting inconsistencies, sound oddities, etc. frighteningly, these glitches can be hard enough to detect even when the fakes are made with an amateur's level of attention to detail. this implies that there is already a very widespread use of the technology by people with the capability to edit out these glitches. with the technology of today, i think that state actors or potentially corporate actors could very easily generate nearly undetectable deepfakes if they cared to do so. i expect that they already have, in fact. but probably not where there would be a massive public eye to scrutinize an improbable or a ridiculous forgery. that would be too risky. instead, the state of the art of deepfakes is likely invested in making the realm of the plausible "into reality" in a targeted way for the sake of shifting opinion of small populations rather than 100% forming new opinions or causing an about face of prior opinions. consider, for instance, two allied militant groups in some warzone that the US has an interest in. the default condition for these populations is the fog of war, wherein information is often hard to come by and rumor runs rampant, especially among the combatants themselves. the US doesn't want the alliance of the groups to persist for whatever reason. now, introduce a very convincing deepfake video of the commander from one militant group plotting to do something that the other group -- and perhaps some of the original group's members -- would disagree with. there might be immediate chaos between the groups, but probably not. the most likely result will be that some people from each group see the deepfake and get convinced, then perpetuate the false information to others who haven't seen it. over time, the relations between these two groups will become strained, and their alliance may fray or fall apart altogether. the incentive to be able to cause these kinds of changes is undeniable. especially when populations are compromised in their ability to fact-check information by hard problems like insufficient internet access, deepfakes will be especially effective. the article worries about the total capitulation of reality at home to that of the fabricated image. no, that is not the risk at present. we have too many alternative channels to validate information. in fact, we are likely to evaluate information without even intending to by discussing information learned via one channel -- perhaps a deepfake -- with other people who may have learned about it via a different and contradictory channel. skepticism is becoming built-in. for people who aren't in the deep-information environment, however, deepfakes are already powerful enough to tilt reality to its preference... |