Editorial - A lurking danger

Again and again it happens that through the progressive development of new technologies certain milestones are reached, which change humanity forever. For example, the development of integrated circuits thanks to semiconductor technology, or the invention of the Internet. Each of us has several devices with different microprocessors, which communicate with each other and also connect people one with the other. Hundreds of millions of people have a job, their source of information and an important communication channel due to this. A sense of life, to exaggerate it. The world has changed forever as a result of these developments and would be unimaginable without them.
This is exactly the kind of fundamental change we are facing in the near future. I'm just afraid the consequences are far more negative. But let me explain the topic, or rather let me explain the title for now. This post is the English adaption of my recent German post: LINK

What is the Uncanny Valley?

Uncanny Valley is the term used to describe the acceptance gap for a paradoxical effect on viewers regarding the acceptance of artificial characters. Originally this effect was first described by Masahiro Mori as a "phenomenon of the uncanny valley". [1] The acceptance of artificial figures does not increase linearly with anthropomorphism (similarity to humans), but in a certain range there is a very big slump in this respect. The following graphic illustrates this very clearly:

Illustration of the Uncanny Valley phenomenon - Source:LINK The vertical axis defines the likeability for the audience, the horizontal axis defines the humanness score. So the theory of the "Uncanny Valley" says that at a certain level of humanity, acceptance is totally diminished. Only when the similarity is so huge that the artificial figure can no longer be distinguished from humans anymore, the acceptance increases again.

Explanations for the Uncanny Valley

There are several explanations for this phenomenon, none of which serves as a sufficient scientific explanation. [2]
1. Media-psychological explanation: Figures that appear clearly machine-like at first glance are credited with additional human behavior or human characteristics. However, if a figure obviously tries to appear human, any kind of mechanical behavior is noted as a mistake or error.
2. Psychological explanation: Robots that try to present themselves humanly are subconsciously measured by the observer with the same standards as humans, so unnatural characteristics are immediately taken as being negative.
3. Neuroscientific explanation: Investigations by magnetic resonance imaging have shown that the deviations of the internally formed prediction could possibly cause or enhance the phenomenon.
This is also one of the reasons why animation studios like Disney or Pixar go much more in this direction of making characters look realistic, but not too human. They are more accepted, both consciously and subconsciously.

We've already crossed Uncanny Valley!

Many experts such as CGI artists, engine programmers or GPU manufacturers already regard the Uncanny Valley as being crossed. The combination of increasingly capable engines and animation software products combined with the annual increase in compute power, especially in the graphics card sector, has ensured that photorealistic CGI is considered doable. As long as you have enough resources to invest. A good video which explains this circumstance can be found here: https://vimeo.com/237568588

Personally, I don't think we've crossed the Uncanny Valley completely yet, but that's a subjective opinion. However, I do think that we are approaching the end at great speed. Accordingly, we should finally think about the consequences. An open discourse is needed.

AI meets CGI - Where is the danger?

It's about the technology that's on everyone's lips right now: artificial intelligence. In terms of images/videos it was initially only pattern recognition, but now we are much further along thanks to neuronal networks. With the help of artificial intelligence, software products like Photoshop can automatically crop human motifs[2]. A product from Google can give strongly pixelated portraits a recognizable face.[3]

It becomes dangerous from the moment artificial intelligence is combined with the power of modern engines and the pure deception of the viewer is declared as a goal. Something is created that even experts can no longer recognize as an artificial product. This is exactly the trend we are facing in the near future.

Deep Fakes - Everything becomes a fake

In Deep Fakes, people' faces in videos are replaced by other faces. The term itself is a combination of "deep learning" (deeper learning through neural networks) and "fake". All that is required for deep fakes are a few individual images of the person you want to retouch into the original material. As already so often in the technical world, this technique found its first large audience through pornography. Users used images of celebrities to fake porn. Reddit users have worked together to improve the results and created a huge number of deep fakes. I'll link you here to the appropriate adult-free subreddit which contains examples of deep fakes: GifFakes
As unbearably horribly fake porn for the persons portrayed is, deep fakes can have much worse and greater effects. I will briefly show here an example created with the 44th President of the United States, Barack Obama: https://www.youtube.com/watch?v=AmUC4m6w1wo

If you add speech synthesis[4] to the mix of AI and CGI, you can not only optically fake the person, but also let them say whatever you want. Of course with the right tone color too. This combination can reach terrible proportions. At the moment the calculation or the creation takes quiet long or is not yet executable by any amateur:

For example, Jordan Peele's production company Monkeypaw Productions put the Obama video together using FakeApp and Adobe's popular After Effects program for graphics editing. BuzzFeed reports it took 56 hours of automatic processing, overseen by a video effects profesional, to get the video right.[5

However, given the enormous development speed in the fields of computing power and neural networks, it is only a matter of time before it works at the push of a button.
Speech synthesis has been well advanced for some time. Adobe has been integrating such a function into its audio software product for quite some time now. It converts a text input it into the appropriate voice: https://youtu.be/I3l4XLZ59iw?t=1m28s

In Austria, for example, the announcements at the train stations are realized exactly by such a system. For this purpose, an algorithm was previously fed with a large number of voice samples of the respective person. [6]

Infinitely dangerous - Possibly changing the world

The respective examples are of course endless. Just think that hackers would short-circuit a TV signal from a station and broadcast a presidential declaration of war to a neighboring country. Or one could completely expose political opponents in the shortest time. This would be fake news at a level we can't even imagine at the moment.

“The prospect of any Internet rando being able to swap anyone’s face into porn is incredibly creepy. But my first thought is that we have not even scratched the surface of how bad ‘fake news’ is going to get.” - Julian Sanchez[7]

The technology is there, and is growing incredibly fast. As a society, we must therefore consider countermeasures.

Is there anything we can do?

Even normal fake news reports are a problem at the moment. You only have to look at all the coverage of the last presidential election campaign in the US. However, this is usually news that can still be relatively easily exposed as a fake, at least by experts. Here media ethics and early education about the use/consumption of media can strongly counteract this problem. In addition, it is often easy, even for normal people to expose the fake news circulated by linking to actual facts/evidence/multiple (correct) reports.

In a piece for The Outline, journalist Jon Christian puts out a worst case scenario, that this technology "could down the road be used maliciously to hoax governments and populations, or cause international conflict".[8]

With perfect self-created fakes, of course, it's completely impossible. Even experts cannot do this. If something is packed into an ordinary video recording it is also completely unclear whether a recording was created by a human team or just a neural network is behind it. Everything becomes fakeable.

Does the Blockchain offer a solution?

This is a question that currently has no solution. For single images there are already suitable algorithms, which look at the image structure and/or the metadata of images and say accordingly whether such an image was manipulated or not.

Single images can be easily exposed - source: LINK With videos it looks completely different, especially because of the amount of data and the necessary computing power. Even further, the existence of such software products is worthless if they are not used. These would have to be required by law to be integrated into all platforms such as Facebook, Youtube and Co. Of course, any software check can be passed one or the other way. With the principle that something in the blockchain cannot be deleted, the blockchain offers an interesting feature in this respect. Thus it would always be theoretically possible to determine the first upload/version of a video and thus to expose possible fakes or to declare the correct version later. It is questionable how this can be integrated into everyday life or mainstream media. In any case, I find it very odd that in none of the many reports I have researched on this topic the word Blockchain was used even once. That's why it's your turn now.

I would be interested in your opinion/suggestions

Although I am very familiar with video technology and the related topics, I am still a beginner with the blockchain and do not feel able to give ultimate statements about the possibilities here. Luckily there are countless experts on this subject here on Steemit, so I would be incredibly happy to receive qualitative input from you. For example, I could imagine a digital signature that links video files to a camera during recording and clearly declares them as real filmed material. The storage of this digital signature would take place in the blockchain. However, I don't know if I can think this possibility to the end.

Summary

The Uncanny Valley will soon be completely behind us, whether in the world of robotics or in the field of of computer-generated images and videos. With a new technological chapter of this magnitude it will also be necessary to define new rules and processes in order not to lose control. In robotics and in the general field of artificial intelligence, there is a slowly growing debate on this, even though I still think it is far underdeveloped. However, this does not yet exist with regard to the imminent revolution of the media that is going to happen. But time does not stand still in this area, as it does nowhere else. Please let us not need a bad consequence first before we define rules and regulations here. It's not like it's not predictable.
I hope you enjoyed this report and perhaps it gave you some cause for reflection. As usual, I would be very much looking forward to a discourse on the subject. Maybe my master thesis will go strongly in this direction, so I am very happy about any input.

Sources:
[1] https://www.wired.co.uk/article/uncanny-valley-tested
[2] https://en.wikipedia.org/wiki/Uncanny_valley
[3] https://arstechnica.com/information-technology/2017/02/google-brain-super-resolution-zoom-enhance/
[4] https://en.wikipedia.org/wiki/Speech_synthesis
[5] https://www.tomsguide.com/us/deepfake-faq-fake-video-fakeapp,news-27003.html
[6] https://www.aristech.de/en/solutions/
[7] https://www.lawfareblog.com/deep-fakes-looming-crisis-national-security-democracy-and-privacy
[8] http://www.bbc.com/news/technology-42912529

My other channels: @tobypics: LINKYouTube: LINKInstagram: LINK

What is waiting behind the Uncanny Valley? Perfect fakes!