More selected projects


Recursive Prophet:

How Machine Learning Intervene in Social Strata and Change Human's Identity

Author: JINGYI CHEN
May 2020


Abstract and contextual review


Due to the pandemic crisis, the enforcements adopted by governments has triggered a discussion about social surveillance. It seems can't be forgettable intentionally. With the technology explosion, it’s unreasonable for the ruling class and capitalists to tie the weapons up and place them on the top shelf. On the one hand, people in the East Asian countries, like China, have got used to the cameras everywhere in the schools, streets, and other public areas, as well as censorship onto the cultural products and on the Internet. People have accepted them as the guarantee for their safety against potential physical felony. And on the other hand, some politicians in the Western countries, being opposed to adopting tracing the trails of contracted patients by mobiles to protect their privacy. The potential reason for that probably is due to neoliberalism, which is prevailing by the capitalists to diminish the barriers of international capital flows. As Wendy Hui Kyong Chun(2016) asserts, "Neoliberalism, to repeat a cliché, destroy the public by fostering the private. It leads to the rampant privatization of all public services..." Therefore currently, data privacy become a critical issue of society.
Machine learning as an essential part of artificial intelligence and social surveillance, therefore, is the main topic I want to explore. We can find they are used in fields like human face recognition, which can recognize each person in public places, as well as recognizing other things, catching every license plate on cars, and check out if the driver is speeding, also the recommendation of online contents. I'd like to find the progress and principles of machine learning through the supervised and unsupervised algorithm to find the implicit social norms and arbitrary presumption of the designers, by which the elites to continuously shape the social strata and lead to the dichotomy contradiction. And I'd like to found the disobedient possibility inside it.
The main reason for me to explore disobedience in machine learning is due to it is regarded as automating people by the algorithm, and repressing them. However, I suppose it also has the opposite side, which can help us overcome the humanitarian and moral crisis of the new technological revolution. As Garnet Hertz(2016) mentions in Disobedient Electronics, the reason to be a disobedient creator is not meant to revolt for revolt, to reserve due to hatred and rivalry, to shaping and manipulate public opinion but to "appeal to emotion and personal belief". 
Supervised learning, according to Wikipedia, it is the machine learning task which uses a function to map labeled inputs to outputs based on example input-output pairs. However, in unsupervised learning, there are no pre-existing labels and only with a minimum of human supervision. Because of this characteristic of unsupervised algorithms, artificial intelligence can always find something unexpected.
The most obvious side which can reflect the implicit social norm is to make things tagged. In algorithm one of the main algorithms is classification, and in classification, we need to distinguish samples by ourselves which involves our inherent views of the world. For instance, when it refers to gender, and in the situation when most programmers want to sort out people's photos, in general, it is divided into two kinds, male and female. Even the computer takes some mistake at first, programmers will pick out women's photos from men's folder manually. When an individual meets such a detection system, not only embarrassment, if this causes some inevitable huddle in their lives, like schooling or working, they have to adapt the system. As Shoshana Zuboff(2019)argues, "As systems are designed to intervene in the state of play and actually modify behavior, shaping it toward desired commercial outcomes". In her context, the main purpose is to make profits from people so that capitalists get their behavioral surplus and change their thoughts, like in politics and aesthetics. Overall, the biases not only formed by consumption drives also by inherent instincts and traditional values, which are needed to be review in any historical phase to fit the new situation.
Zuboff also came up with a concept “the division of learning in society” which is brought by the surveillance capitalists, in society capitalists have the right to know any data, to transfer any events and objects to become data and decide who to know them. By recommendation algorithm, which partly uses the K-nearest neighborhood algorithm which is one of the classification algorithms. By immersing in the harmonious atmosphere created by Google and the media, watching recommended TV shows on Netflix, people are "surrounded by opinion bubbles and never be challenged." The way of getting information nowadays has been expanded than the pre-Internet era, but the circumvention of self-awareness is not easy to break. Moreover, some media have adopted machine learning to automate producing content, which is supposed to be creative and skeptical. Hence machines have produced countless meaningless texts and images which let people feel drown in cultural rubbish. "According to AI's public relations manager James Kotecki, the Wordsmith platform generates millions of articles per week; other partners include Allstate, Comcast, and Yahoo, whose fantasy football reports are automated. Kotecki estimates the company's system can produce 2,000 articles per second if need be." , reported by Ross Miller(2015). The content is not thought twice by any emotional and critical brain also the binary core of computers can not understand the valuables for humans. The purpose of producing cultural trash is simple just for the market and profits. Shoddy contents floor in, dazzle and involve all. 
On the contrary, the unsupervised algorithm certainly has serendipity. In the picture below, which are produced by convolution cores, by the image output, it tells us how the binary machine to dissect and understand images. It extracted the most obvious feature of the images and found all edges, and in fact, it is similar to the process of how humans observe an object.

Fig1&2:Pictures created by convolutional core during the training.


The process of researching how the neural network training data is inspiring. So unsupervised machine learning always can produce what humans can't imagine and conclude new rules. Artist Holly Herndon (2020) says, "The AI can extract the logic of something outside its operator’s own logic and then re-create it". And she thought the capability of machine learning should not merely be used to get similar work based on which is made before but to help the artist to get out the aesthetic cul-de-sac. Her work "Eternal" is made by Herndon's face and other musicians' faces, the video of the performance is realigned by the algorithm that obvious beyond the human's ability and arouses the sense of connection as a human being.
One of the obedient characteristics of machine learning is it is open source. Which means no one can make unreasonable profits from it. In addition, it is a free tool can be accessed by proletariats to create unexcepted things, by which new technology can reverse old economical empires, it has the power to change old stiff game rules. With more and more people learning AI as an essential skill today, new ideas have always changed our world. 
 

Artifact
Misplaced portrayal



My artwork is an installation that is like a mirror. When the audience stands far from the camera, their image will morph and like an ape, but it looks very vague because of the distance. If they move towards the installation, their image will become normal. But if they continue to close the camera, their figure will more and more like a cyborg.

Fig. 3 &4 & 5: My photo mixed with the pictures of an ape and a cyborg. By adjusting the weights during the training I can decide the degree of fusion. I want to control the weights by the distance of the audience and the camera.  Created by iOS APP FaceFilm.


The suppose of the artwork is rethink what means as a human being and the position of the human in the crossroad of physicality and virtuality. As the pandemic crisis happens, the vulnerability of the human body has been exposed. Even with countless techniques that have been developed, human beings still can't overcome the harsh nature. I link people's figures with an ape, is to implict human are arrogant now and the momery of suviving in nature has been ignored gradually. And the pandemic is also propmt the digitalization of the society, with the develping of the hologram and augmented virtuality and other technologies, I feel my physical body is redundant today, in order to reinforce the gradually withered body, many wearable devices have sprung up. And the data of limbs will be collected and analysed. We are going to an era when our bodies have been changed. What if the corporations offer critical components are out of business? And we are going to easier to manipulated by technological corporations and capitals. In fact, mobiles have been an indivisible part of our body and social connections. We are more tight-knit in the mesh but each node is more isolated today.
Firstly, I want to use Python and its libraries to get pictures of cyborgs and apes from Google image. When I get enough(about 50000 for training and 1000 for the test) pictures, and preprocess their sizes. And after that, I want to use CycleGAN to extract the features of cyborgs and apes and they can be better to mix with audiences' faces in real-time. Alternatively, I can also choose to use APIs of exsiting face fusion algorithm on face processing platform like face++ or Tencent cloud. By receiving face data through cameras and processing it on the platform and sending back, I can processing faces of the audience in real time.
The picture below is the simulation of the reality. I would put the main part of microcontrollers in the wall and the cameras beside the big LCD display.

Fig. 6: Created by C4D.

Annotated bibliography

a1.

Chun, Wendy Hui Kyong. Updating to Remain the Same: Habitual New Media, MIT Press, 2016. ProQuest Ebook Central, https://ebookcentral.proquest.com/lib/goldsmiths/detail.action?docID=4537749.

In this book, the author came up with an equation "Habit + Crisis = Update" and discussed what is habits today, how the Internet has changed and how people's views about the Internet has shifted from "cyberspace" in the 1990s to "surveillance" about mid to late 2010s, as well as the flux of public and private space. As the change of meaning of habits, people 's behavior "from voluntary to involuntary, the conscious to the automatic". Besides the author also insulted the propaganda of neoliberalism, which made me realize the origin of the deputes in recent years about privacy and surveillance.

 a2.

Naughton, John. "The goal is to automate us': welcome to the age of surveillance capitalism", The Guardian, 20 Jan. 2019.  https://www.theguardian.com/technology/2019/jan/20/shoshana-zuboff-age-of-surveillance-capitalism-google-facebook. Accessed 23 April 2020.

In the Article, Naughton and Zuboff introduced the new book of Zuboff and they discussed Zuboff's main views of "Surveillance capitalism. They also criticized the state quo of the society and the Internet, as well as the power of Facebook, Google. The main idea of behavioral surplus probably affected by Karl Marx in the end. Moreover, Zuboff also concluded several phases of capitalism, mass production led by Ford Motor and the logic of managerial capitalism discovered by General Motors. The article is the main one that inspired me to figure out the connection between machine learning and surveillance.

a3.

McDermott, Emily."Holly Herndon on Her AI Baby, Reanimating Tupac And Extracting Voice", Art in America, 7 Jan. 2020.https://www.artnews.com/art-in-america/interviews/holly-herndon-emily-mcdermott-spawn-ai-1202674301/ Accessed on 26 April 2020.

In the interview, the artist Holly Herndon spoke the reason and the experience of creating artworks by machine learning. She and her partner created an "AI baby" by the neural network, which is a synthetic multivoiced singer named Spawn. They used it as a chorus. In addition, she gave her ideas about the way creators using machine learning in their creation and her views about machine learning, surveillance, platforms, and social opinions. This triggered my idea about use machine learning as a spur to reflect and cherish the emotion and connection as humans.

Bibliography

  1. Chun, Wendy Hui Kyong. Updating to Remain the Same: Habitual New Media, MIT Press, 2016. ProQuest Ebook Central, https://ebookcentral.proquest.com/lib/goldsmiths/detail.action?docID=4537749.
  2. Hertz, Garnet. Disobedient Electronics, 2017.  http://www.disobedientelectronics.com
  3. "Supervised Learning", Wikipedia: The Free Encyclopedia.Wikimedia Foundation, Inc. 22 July 2004. Web. en.wikipedia.org/wiki/Supervised_learning. Accessed on 1 May. 2020. 
  4. "Unsupervised Learning", Wikipedia: The Free Encyclopedia.Wikimedia Foundation, Inc. 22 July 2004. Web. en.wikipedia.org/wiki/Unsupervised_learning. Accessed on 1 May. 2020. 
  5. Naughton, John. "The goal is to automate us': welcome to the age of surveillance capitalism",The Guardian, 20 Jan. 2019.  https://www.theguardian.com/technology/2019/jan/20/shoshana-zuboff-age-of-surveillance-capitalism-google-facebook. Accessed 10 April 2020. 
  6. Miller, Ross. "AP's 'robot journalists' are writing their own stories now", The Verge, 29 Jan. 2015.https://www.theverge.com/2015/1/29/7939067/ap-journalism-automation-robots-financial-reporting. Accessed on 27 April 2020. 
  7. McDermott, Emily."Holly Herndon on Her AI Baby, Reanimating Tupac, And Extracting Voice", Art in America, 7 Jan. 2020.https://www.artnews.com/art-in-america/interviews/holly-herndon-emily-mcdermott-spawn-ai-1202674301/ Accessed on 26 April 2020. 
  8. Rushe, Dominic. Let me into your home: artist Lauren McCarthy on becoming Alexa for a day", The Guardian, 14 May 2019.https://www.theguardian.com/artanddesign/2019/may/14/artist-lauren-mccarthy-becoming-alexa-for-a-day-ai-more-than-human. Accessed in 15 April 2020. 
  9. Mizota, Sharon."Alexa, meet Lauren: L.A. artist turns her apartment into an experiment in artificial intelligence", Los Angeles Times, 28 Nov. 2017.https://www.latimes.com/entertainment/arts/la-et-cm-lauren-mccarthy-review-20171127-story.html. Accessed at 15 April 2020.      

 

 

Other Materials:

        Fig1&2: https://blog.csdn.net/weixin_41417982/article/details/81412076