The footage below is generated by Deepfakes.
If you already know of Deepfakes, you may also have coupled the search with your favorite celebrity. If you haven’t, probably don’t do this at work.
Deepfakes is an artificial-intelligence based program that allows the creation of fake videos of people that is nearly indistinguishable from real footage.
The current applications of the program are at best … morally gray.
Although much of the footage generated can be differentiated to be fake at the moment, it will get increasingly more difficult as the algorithm “learns” and is trained for longer periods of time. The fake footage can even be combined with A.I. generated audio from programs like Lyrebird to become even more compelling.Media speculation has already proposed nefarious ways A.I. generated sound bites and footage could be used, from generating fake audio of what was never said, to disrupting entire national democratic processes and the spread of false propaganda through generated footage of powerful figureheads.Those who have kept up to date with recent Cambridge Analytica and Facebook political events may argue that nefarious political organizations won’t need A.I. generated footage and audio clips to disrupt political processes, but with these tools, they will certainly become viciously more effective at spreading fake news and propaganda to fit their agenda.See for yourself of what these programs are capable of (safe for work):
Who can take advantage of these programs?
The program is so accessible that any lay person or powerful organization can leverage it for lighthearted humor, debaucherous desires, political influence and slander, as long as the hardware is available to them. Most of the computing power required is available in lower end computer graphics card found in most people’s computers.[1]
If the footage and audio generated by these programs become indistinguishable from live footage, what are the implications?
If Cambridge Analytica was simply able to manipulate an entire population just from well placed ads on a social media network, imagine what fake video and audio propaganda can do to disrupt political status quo and influence people’s beliefs and opinions.
Not only can the video and audio be used as a smear campaign against people in power, but those who have said or done things that may have been career ending or even criminal can have the plausible deniability to defend themselves. With the programs so prevalent, anyone can argue that those statements recorded may have been artificially generated or falsified from what occurred in actuality, or never to have occurred at all.This short piece from NPR on Lyrebird provides a summary of how the generated media can destroy credibility and who we can really trust.[2]
Can You Believe Your Own Ears? With New 'Fake News' Tech, Not Necessarily
Soon, we might not be able to believe our own ears. New technologies for creating faked audio are evolving quickly in…
https://www.npr.org/player/embed/599126774/599149229
“Eighteen months ago when that audio recording of President Trump came out … if that was today, you can guarantee that he would have said ‘it’s fake and he would have had some reasonable credibility in saying that as well’”
Knowing how Trump tweets today, what can he call fake news if the Trump Dossier allegedly containing the infamous Trump pee tape is ever to be verified by the Justice Department? Although there will be other ways for digital forensics experts to verify the authenticity of video and footage, the time and resources required would not be trivial. The need for some way to distinguish between reality and fabricated footage is growing very real.
What can be done?
Videos are compilations of images, and images are compilations pixels and data. With the advent of blockchain and distributed ledger technology (DLT) like The Tangle, data can now be stored and tracked forever on an immutable database, where every action performed to the data is recorded.
New images and videos can be recorded using blockchain or distributed ledger devices or applications that would simply be downloaded from various app stores. Blockchain distributed ledger technologies like IOTA and Ethereum and other emerging crypto-platforms may be the foundation of where these applications are built. Once the images are recorded from the application connected to the distributed ledger, the images and videos can now be assigned cryptographic signatures which match them to the distributed ledger.
The cryptographic signature can include data such as the GPS location of the footage, the time and date which it was recorded, the aperture and frame rate, any related data and even the footage themselves.
If there is ever a questionable video, audio, or footage as to its authenticity when controversial footage is released to the public, it can be compared and matched to this distributed ledger to verify its authenticity. If the footage matches exactly what is on the ledger, there needs to be no extensive forensic analysis or question that the footage or image viewed was once recorded without editing.
Further, editing applications can install distributed ledger plugins which record all edits and changes performed to an image and footage. Digital forensics might require experts today like Youtubers wearing silver face paint to figure out, but tomorrow, it may become an entirely different field, or simply a hobby.
Links and sources mentioned in article:
[1] https://www.deepfakes.club/tutorial/#Requirements
[4] https://blog.iota.org/the-tangle-an-illustrated-introduction-4d5eae6fe8d4
The fakes do a DECENT job, but you can still spot the edges...
These fakes coming out may be more convincing, but then, even if a real video comes out, it will be called a deep fake.
I could see these deep fake videos requiring some serious server time to get the job done too... I don't think too many people are going to use their desktop running for like a month just to build a 30 second clip.
I like your plan to use the block chain technology, that would be a great way to establish a YouTube replacement, and would help protect content creators.
I agree bmanmcfly, currently the fakes are still quite easily discernable, but who knows how much better than can get?
What's amazing is how much is already on the net just from every day people using their computer GPU's, imagine what can be done on an institutional super computer that has significantly greater resources!
Congratulations @xinyuhe! You received a personal award!
You can view your badges on your Steem Board and compare to others on the Steem Ranking
Vote for @Steemitboard as a witness to get one more award and increased upvotes!