Deepfakes what are they and how long has that phenomenon been around?

Filed in NEWS ANALYZES by on 26 June 2019 Comment

source: medium.com

I have frequently discussed the techniques with which deepfake characters can be created. For the new readers, I would like to repeat that here in a little more detail in a special article dedicated to this topic. Because if you follow the news daily, it is extremely important that you become acquainted with this subject, because you will see what techniques are available for simply playing the people. Very easy.

Deepfakes are made through GAN (Generative Adversarial Networks)) software techniques. This is artificial intelligent software that, based on multiple AI systems in a network, creates characters out of nothing. AI is English for Artificial Intelligence; what stands for artificial intelligence. Another AI network then tests the images created by the first network and rejects or approves them. By doing this in a cycle, the characters become more realistic with each step, so that you can ultimately generate completely fictional people who look like ordinary everyday people (whom you might just meet on the street). If you want to know exactly how this works, first watch the video below from NVIDIA (the well-known graphics card manufacturer for PCs).

It is not only useful to know that this deepfake technique exists, but also how a deepfake character can be used in, for example, videos or to generate social media profiles (including an entire history; including photos and videos and likes from others) deepfake social media profiles). For example, social media discussions can easily be monitored online by "home workers" or employees of a telemarketing agency, for example, where the characters with whom you are discussing may hide behind such a deepfake profile (whose friends network is also filled with deepfake profiles). They can attack people on their timeline in discussions in order to steer sentiment among the people in a certain direction.

Let us look at all application possibilities, but before we start, it is useful to know that the game and film industry, but also TV producers, have had such techniques for a long time. However, the work is now simplified to such an extent that you can do it yourself on a home-garden-and-kitchen PC.

When Paul Walker died in the middle of the Fast and Furious 7 recording, the Weta Digital company was called in to complete the film version of Paul Walker. Based on a combination of methods such as old images, body scans of Paul's brothers and digitization of Paul's head, Weta Digital brought Paul Walker back to life. The video below provides a summary of how this worked.

The 3D motion capture technique has existed for years in which actors wear suits to record their movements and then superimpose characters created digitally via CGI. That is comparable to the technique used for Paul Walker, only with live actors wearing a motion capture suit. That technique is now also available for people with a low budget (see video below), but a good example of a film in which this technique has already been used is the movie Avatar from 2009 (see here).

NVIDIA has already caught up with the use of these suits and CGI technology, because it uses neural networks to train the software. In fact, this is the same technique behind the deepfake faces. NVIDIA is now not only able to generate non-existent faces, but can drive through a city with a camera and turn it into a winter landscape (in real time). Such techniques can for example be used to train the AI ​​software of self-driving cars in changing weather conditions, but they can also be used to make a motion capture suit unnecessary. A simple GoPro camera or webcam is enough. Take a look from 1: 03 min. In the video below to see how that works.

Now you may think that the possibility to do this in real time does not exist. Think again. We have already seen above that it is possible to create non-existent persons via Generative Adversarial Networks. We now know that both an urban environment and a character can be generated via neural networks. The question is whether that is also possible in real time. That's where the technology of real-time facial reenactment comes in. This has been around for a simple home PC since the year 2015 (see video below).

All in all, we can therefore state that it has been possible for years to generate deep-fake videos. However, the technology has now become so simplified with the emergence of Generative Adversarial Networks, neural networks and real-time facial reenactment, that you can actually create an entire history of a non-existent person in a matter of minutes, a live interview of that non-existent person can create any environment from any cameras perspective and any weather condition.

What are the implications of this? To start with, you can say that you have not been able to trust 100% for years. View here how long CGI techniques have been used in the film industry. However, at the moment it is so simple that anyone with a budget of a few thousand euros can already do this. If we assume that the media is fair, then we can assume that they have not been using such techniques for years. However, if we take into account the possibility that governments use psychological operations to psychologically bring the people into the acceptance mode of new and stricter legislation, then we must realize that for years technically there has been nothing standing in the way of producing fake news. In that context it is very interesting to know that the country's largest news agency (the Algemeen Nederlands Persbureau; abbreviated ANP) is in the hands of a TV producer (who is also a billionaire). How big must we be to be sure that these techniques have not been used for years?

It seems that the media are eagerly looking to close the leak that Martin Vrijland has hit in the bottom of the big mainstream media ship. For several years now, I have been explaining how media can manipulate images. Jort Kelder and Alexander Klöpping were therefore allowed in the Kelder & Klöpping TV program show what deepfakes are. Also the radio program Image determiners BNR Nieuwsradio (perception managers) recently mentioned what I have been writing about for so long. It is clear that the panic is always noticeable and that the program makers must try to keep the viewer and listener on board. You must continue to trust in the media and in democracy, because there is nothing worse than the mob revolting (to speak in the words of Jort basement).

Of course the "solution" for all this is that governments and tech companies are going to try to add a kind of watermark to movies, so that they can be checked for authenticity. The only question is that if governments themselves have been using fake news for years to push through legislation and play on the people whether or not that watermark is so reliable. Is a butcher going to reject his own meat? No of course not. All the news from John de Mol, NOS, De Telegraaf and so on has always been completely reliable and honest! Cough. Do you really think that John de Mol will appear on TV today or tomorrow to say: “Sorry ladies and gentlemen, I have made fake news with all the TV studios and software that I have at my disposal. I have presented you with fake news and played with psychological operations at the cost of the tax pot and filled my bags"? No of course not. And of course you have to keep faith in the media and in the government, because who else do you have to trust? read here...

Possible deepfake applications:

  1. deepfake social media profiles
  2. photos and videos from the past including family and friends
  3. live interview with a non-existent person
  4. images from security cameras
  5. video as evidence in the news (fake news productions)
  6. and so on

Source link listings: bnr.nl, wikipedia.org

74 Shares

Tags: , , , , , , , , , , , , , , , , , ,

About the Author ()

Comments (1)

trackback URL | Comments RSS Feed

  1. guppy wrote:

    The leaders of this planet still use the same techniques. In terms of technology, they do keep up with the times. In the past you could also drive whole tribes crazy with a non-existent figure. They have always been one step ahead of us by controlling history with the help of book rolls and such.

    In the past there were also Martin's who pointed out to people that they are being fooled.

    The bad thing is that people today think they are smarter than our predecessors. I don't think much has changed, we are still slaves working themselves hard for a bite and a drink.

    They used to say "you don't have to believe everything they say"

    Today we say "you should not believe everything you see"

Leave a Reply

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to 'allow cookies' to give you the best browsing experience possible.If you continue to use this website without changing your cookie settings or you click on "Accept" below then you agree with these institutions.

close