I did not wrote this Article but i have the full Permission to publish it from the People who wrote it and they dont want to be mentioned by theyr Names so they asked me to publish it,i just converted it for publish :)-greetings to the pervs,Have Fun
Deepnude
Not ready yet? On June 23rd, 2019, the childhood dream of many finally came true: x-ray goggles, or at least it's software pendant DeepNude[1] was launched. Just as the advertisements on the last comic book pages had promised us decades ago, the software DeepNude fulfilled this promise to make clothes disappear.
Thanks to AI, or to be more specific, through generative adversarial networks (GANs), DeepNude transforms pictures of dressed women into pictures of undressed women.
This report gives a short summary of the app’s functionality, and the events around DeepNude during the last year.
[1] www.deepnude.com; www.twitter.com/deepnudeapp
As mentioned above the app DeepNude makes teenage dreams come true and has the functionality to transform a picture of a dressed woman into a picture of her being naked within seconds - what seems to hit a nerve. Released on June 23rd the app turned out to be a big success. Within three days of launching the app got worldwide attention followed by a rush of over 500,000 downloads.[2] Beside professional interests, this example caught our attention for two reasons. First, the extraordinary step of the developers to take the app offline despite its success, and second, the looming question of how society is affected by being increasingly confronted with fake pictures and videos. A question that needs to be posed when becoming aware of the dimensions. Deeptrace made the development visible: In 2017 only a couple of videos existed - now there are more than 17.000.[3]
In order not to be dependent on second sources, we wanted to get an impression of the software and its functionality. Therefore we obtained a copy of the software and let it run inside a sandbox, for security reasons. Using the app is simple and it worked just fine, in fact there is only a single button to select an image, no other options. Uploading the picture (of a woman) - thirty seconds of waiting – ready!
The problem that occurred was that the watermark used to indicate the fake, was easy to remove and therefore problematic. Even though the nude pictures generated could have been identified as fake by taking a closer look. But using a blurred picture or not taking a closer look makes the pictures appear pretty real. Therefore - in wrong hands - these pictures could cause a lot of damage.
[2] Antje Barthold für dasding.de (2019): Nacktscanner-App war zu echt - DeepNude ist offline.
https://www.dasding.de/lifestyle/netztrends/Nacktscanner-App-war-zu-echt-DeepNude-ist-offline,nacktscanner-100.html
[3] Patrick Dax für Futurezone.at (2020): Deepfakes: "Dem Gesicht kann man nicht mehr vertrauen"
https://futurezone.at/digital-life/deepfakes-dem-gesicht-kann-man-nicht-mehr-vertrauen/400750383
Katelyn Bowden, the founder and CEO of revenge porn activism organization Badass, raised some important issues regarding DeepNude: “This is absolutely terrifying. Now anyone could find themselves a victim of revenge porn, without ever having taken a nude photo. This tech should not be available to the public.” Furthermore she stated: “Yes, it isn’t your actual vagina, but... others think that they are seeing you naked. As a deep fake victim said to me—it felt like thousands saw her naked, she felt her body wasn’t her own anymore.” Sexual harassment is a problem in our society, as we all see in the media every day. Since DeepNude makes some sort of sexual harassment easily available with one click, it has to be at least discussed if it supports sexual harassment.
On the other hand, popular programs like DeepNude also have the potential to draw the attention of a broader public to a more general problem: the trustworthiness of images and videos. Since it becomes easier and easier to fake images and videos, how can one believe what he or she sees on the internet or in other media. The potential misuse goes far beyond sexual harassment and could, in the worst case, even trigger conflicts on a large scale.
How it works
“The coolest idea in machine learning in the last twenty years.” - Yann LeCun[4]
DeepNude is based on an AI technique called GAN, which has its origin in Ian Goodfellow’s doctoral thesis[5] from 2014. One can imagine the underlying principle as follows: two AI’s (e.g. deep neural networks) are playing ‘truth or lie’ against each other, the first Generator-AI generates a random fake image, while the second Discriminator-AI is trained to discriminate between real and fake images. For obvious reasons the two AI's are often referred to as the 'artist' and the 'critic'. The latter automatically filters out inferior fakes, the result will be an artificial but certainly realistic image. The following figure schemtically illustrates the underlying concept of GANs. In the case of DeepNude, the GAN was trained with over 10,000 nude photos of women and then tried to improve against itself. The developer justified why the application only works with women with the fact that pictures of nude women are easier to find on the internet, but he is hoping to create a male version too .[6]
[4] LeCun: Silver Prof. at NYU, Chief AI Scientist at Facebook, one of the “Godfathers of AI”
[5] Goodfellow, Ian: Deep learning of representations and its application to computer vision, PhD thesis, University of Montreal, Canada, 2014.
[6] https://www.vice.com/en_us/article/kzm59x/deepnude-app-creates-fake-nudes-of-any-woman
Not ready
As mentioned above, the developers had initially the idea to avoid misuse by placing a “Fake”-stamp in the top-left corner of each output image. However, this did not solve the problem, since it can be easily cropped or removed using image editing software or GAN-based methods. Note the irony.
A few days after the release, when the developers became aware of the huge global impact of the app and its harmful potential, they came up with another radical action. The developers’ solution to the problem of possible misuse of the app or to be more precise of the pictures was to take the app offline. They have posted that they won’t provide the app anymore. The explanation offered by them was that they weren’t expecting such high numbers of downloads. They expected just a few sales a month intended as entertainment only.[7]
"The world is not yet ready" - @deepnudeapp
[7] Antje Barthold für dasding.de (2019): Nacktscanner-App war zu echt - DeepNude ist offline.
https://www.dasding.de/lifestyle/netztrends/Nacktscanner-App-war-zu-echt-DeepNude-ist-offline,nacktscanner-100.html
The old question arises how much scientists and developers have to take responsibility for the technological applications that they make possible. As Andrew Ng, one of the most respected figures in machine learning twittered: “I’m glad DeepNude is dead. As a person and as a father, I thought this was one of the most disgusting applications of AI. To the AI Community: You have superpowers, and what you build matters. Please use your powers on worthy projects that move the world forward.”[8]
Neither the watermark nor taking the app offline is a satisfying solution to the problem aroused by the launch of the app. The watermark does not work. The app is not directly, but still available. Copies of the app or its code are buzzing around the world wide web. Making it still possible to use this technologie without any possibility to control its traffic or use. Questionable is also the developers ethical concerns as reason for the step to take the app offline. They argued that they don’t want to make money this way but took it first, which leaves space for speculations.
[8] https://twitter.com/AndrewYNg/status/1144668413140144128
Conclusion
Ready or not, deep fakes are and will be a big thing. Eric Goldman, law professor and director of the High Tech Law Institute at Santa Clara University, has assessed the situation as follows: “We have to prepare for a world where we are routinely exposed to a mix of truthful and fake photos and videos.” We agree with Prof. Goldman’s assessment of the situation. The potentialities of AI are at hand, but we are beginning to observe also the downside of this medal. Without extensive and infeasible regulations, it will be very hard to prevent misuse, e.g. deep fakes. As a consequence of large-scale AI misusage, information value will erode and human perception of the truth in media will relativize consequently. And therefore may add to what is called the post-truth[9] era.
[9] https://www.welt.de/kultur/article159560304/Danke-Merkel-fuer-das-Wort-postfaktisch.html
DeepNude has impressively shown us the socio-technological possibilities of AI. In fact, an old SciFi dictum says that the possibilities of new technologies are most easily recognized by the way they are used by criminals and artists. Which of the two categories the developers of DeepNude belong to is up to the reader.
Navied Mahdavian, New Yorker, March 5, 2018
