First this is incredibly cool and a fantastic technical achievement -- seriously that's amazing.
This might be the future of state level propaganda but it probably isn't the future of fake news. Fake news sites right now are poorly written, poorly edited, cite no sources, and provide no verifiable evidence to their claims and yet they still spread like wildfire on social media. Not only is this level of sophistication not needed but it the low quality works in favor of these sites to cultivate a rabid audience because they select for precisely the people who would believe them. If they were higher quality they might actually be less effective.
Here's a metrics insight to help illustrate this- even for a legitimate news site, there are tons of articles that get far fewer pageviews than facebook shares, sometimes by a thousandfold. In other words, a catchy headline, provocative photo and a brief lede is all you need to be effectively disruptive.
Edit, just to be clear: people often share things without reading them.
Yeah, this is something which - if sufficiently well done - you use to destroy a politician through leaks of the faked horror video to their colleagues and the mainstream press.
For good old "fake news" it just needs a certain demographic to want it to be right; it doesn't really matter if any video ever existed or not or if the claims are trivially falsifiable or something already well established to not be true. We live in a world where despite the results of the US election being extraordinarily well publicised, a full 49% of surveyed Trump voters say they believe he won the most votes from the general population and nearly a quarter of Democrats insist Hillary won the Electoral College (http://www.politico.com/story/2017/07/26/trump-clinton-popul...)
Agreed, that is an awesome demo. I would agree with your point about state level propaganda, I recently read a novel "Version Control by Dexter Palmer" in which the United States has an entire division of people working around the clock to act as the president and tailor ads and other things specifically to individuals. I thought it was an interesting idea but if this technology gets a little more advanced I could see it becoming a reality.
It's an interesting demo, but I think it's still a bit complex for the future of 'fake news'. Remember, most people's standards for proof are not 'video required or get out'. They're quite willing to believe simple text and image based stories if it backs their preconceptions and the source looks semi plausible.
Those who are prone to believing fake news (an audience which likely consists of about 80% of the population, many intelligent people and journalists included) will believe it without fancy video evidence.
Those who aren't prone to believing it will likely realise it's on a website with questionable credibility and not give it the time of day based on that.
A government may use this for propaganda, but it seems too expensive and too complicated for your typical fake news site to use.
>Remember, most people's standards for proof are not 'video required or get out'. They're quite willing to believe simple text and image based stories if it backs their preconceptions and the source looks semi plausible.
The shift to video won't be over a high standard of truth, it'll be chasing the higher ad revenue available on TV/YouTube. It's just the more popular medium
If their goal is to convince people of something false that would be enough. Their goal is usually ratings, and this kind of stuff makes for a better show.
> Those who are prone to believing fake news (an audience which likely consists of about 80% of the population, many intelligent people and journalists included)
Citation needed. I am certain you made that number up on the spot.
The mouth looks like a slightly more sophisticated version of the Annoying Orange. I remember an academic fake talking head demo a while ago that was much more impressive.
Definitely. It looked like there were a few instances where they left the teeth in as a small white line between the lips when the mouth should have otherwise been closed.
When I see technology like this and imagine the implications of AI being able to leverage it, it only makes me more pessimistic for the future of our species.
The old fashion method requires resources ($$$) and considerable skill (i.e. scarce). I believe the point of this exercise is to show the potential of creating realistic fakes on a mass, and distributed, scale. I wonder if official videos from the white house of the future will include a signature of some sort to prove authenticity.
They may have gotten the voice right, but they could have worked on the video better, maybe use snapchat filters that take your facial expressions and impose them on Obama's face.
This might be the future of state level propaganda but it probably isn't the future of fake news. Fake news sites right now are poorly written, poorly edited, cite no sources, and provide no verifiable evidence to their claims and yet they still spread like wildfire on social media. Not only is this level of sophistication not needed but it the low quality works in favor of these sites to cultivate a rabid audience because they select for precisely the people who would believe them. If they were higher quality they might actually be less effective.