So, @grimmcyber's conference today @AlyssaM_InfoSec had a talk on #ProjectDeepfake - but wanted to add something to her presentation, on two parts, from AR/AI and cognitive phycology implications. 1/

For the CogSci aspects, particularly with disinformation, it to think of the change of replacing it within individuals mental models is difficult to do "logical predisposition" which has been exacerbated due to social media, and the "bubbles" people circulate within 2/

Individual mental models tend to now align towards the models they have built up in their minds due to input they have received and therefore replaces general logic reasoning, and this stretches further into perception of information 3/

Second, regarding the AR/AI aspects of deep fakes and live or near live generation will be hampered by procedural generation issues that were first encountered in computer graphics in the early 1990s when doing landscapes and backgrounds - shading and lighting. 4/

Say for instance, the issues a former employer of mine had with doing VR/AR storytelling was that typical content creators, writers and directors could not or would not easily be able to generate multiple story paths that were interactive enough to simulate reality 5/

Without being prescriptive or directive along preplanned paths - say "choose your own adventure" but with very limited choices. You'd have to effectively film each path and angle or re-adjust the models based on the viewers' own position and view. A virtualize camera problem 6/

Add that to the computing power required to regenerate a synthetic person/instance, with speech synthesis as well as a decision model (AI), you'd need quite the computing harness to do that for 1:1 interaction let alone multiple. As you add subjects, that experience degrades 7/

So, as much as we will want to throw this as a threat, t's an unreasonable threat at this point and is more novelty - until somebody finds a shortcut or solves the interactivity question. (Just my two cents - but from some experience on this from the non-specialist role) /END