Big data and privacy computing and personal digital/data sovereignty are in a mutually symbiotic relationship. The "big data" here refers to the highly private data produced in our current highly automated world, such as:
1. Sensors close to people, such as data generated by the Internet of Things;
2. Data close to human behavior and interaction;
3. "Big data" is highly private data - such as medical care, health, and behavior; you may not mind your "overseas shopping" and "chopping hands" data being collected and monetized by Amazon and Taobao (not distributed to you), but you You will definitely mind and even fear that your medical and health data will be harvested.
This kind of highly private data in "big data" is also called "privacy-heavy" data. One of the characteristics of "privacy-heavy" data is that it has a high degree of coupling with our biological characteristics, which also comes from technological progress—because in the past, due to technical limitations, the coupling degree of private data and biological characteristics was relatively low— So privacy protection in that era was not as urgent as it is today. This kind of privacy data with a relatively low degree of coupling with our biometrics can be called "light privacy" data, such as your "overseas shopping" records on Amazon, your Internet browsing records, etc.
A typical example of "privacy-heavy" data that is highly coupled with biometrics is our "face"
image description
（Twitter：Humans have more facial muscles than any other animal on earth - 22 on each side of the face.）
The human face has more muscles than any other animal on Earth: 22 "expression" muscles on each half of the face.
The extraordinary "social intelligence" that we humans have evolved makes our "face" a social tool—just like our other social tool—voice/language, the face becomes a carrier of "expression".
image description
for example:
（HBO - Hall of Faces - Trump)
In the Internet age, our "social" expression is not directly "communicated", but through "data", so if we can process data with technology, the expressive power of our "face" will have a leap. The rapid development of artificial intelligence technology in recent years has made a great leap in our "face" expressive power:
image description
The ZAO face-changing APP, which has become popular in China recently, is an open source software "Deepfake" derived from this technology.
image description
（Uncanny Valley graph - wiki)
Masahiro Mori's hypothesis pointed out that because robots and humans are similar in appearance and actions, humans will also have positive emotions towards robots; until a certain level, their reactions will suddenly become extremely negative. Even if there is only a little difference between a robot and a human, it will be very conspicuous and dazzling. The whole robot looks very stiff and terrifying, making people feel like they are facing zombies. However, when the similarity between robots and humans continues to rise, which is equivalent to the similarity between ordinary people, humans' emotional reactions to them will return to the positive again, resulting in empathy between humans and humans.
image description
(Uncanny Valley example - Left - Angelina Jolie in Beowulf)
In "Uncanny Valley example - Right - Avatar"
The 2009 blockbuster "Avatar" was the first film to cross the "uncanny valley" because more sampling points were used in motion capture.
The current artificial intelligence technology "Deepfake" is an algorithm obtained by "training" a huge amount of data, and it can easily break through the "uncanny valley". This also shows that artificial intelligence has broken through the cognitive field that only humans could touch before. The era we live in is really very different.
A few days before the viral spread, the ZAO face-swapping app became a popular app for swiping the screen, because it is so shocking and fun. You can replace the face of any star in any film and television scene with your own face. In the past, only celebrities had the pleasure of "watching movies". But ZAO has been hot for a few days, and many people have recovered, and suddenly feel "extremely terrified" - can help me change my "star face", can't I use my face to deceive the bank's face recognition verification, and then roll take my money? !
what to do?
what to do?
At present, the common answer to the privacy and security issues brought about by AI "face-changing" technology like ZAO is (mostly from banks and other institutions that use "face-swiping" authentication): this technology is currently not enough to deceive banks and other places. Facial recognition software. Indeed, shaking the color laser printing paper with your photo in front of the bank's face recognition may not work at present, and for example, the face unlocking of Apple mobile phones will not only collect your face image data, but also collect your face image data. Depth data and capillary data in the superficial layer of the skin are not so easy to break through (it is said that generally twins can't unlock it).
That’s it?
This answer obviously does not reassure us. Do you know how the algorithm of face-changing software like "Deepfake" is obtained? Deepfake is based on the artificial neural network intelligence algorithm currently called "deep learning", especially a type of algorithm called "Generation Adversarial Network" (GAN — Generative Adversarial Network), which was proposed by this Stanford University student: Ian Good Fellow
This is a brain-opening algorithm: create an evolutionary virtual environment, let the machine generate two opposing algorithms for games, one for simulation and one for discrimination:
Do you think this image is real or am I making a fake?
You're so frustrated, you don't look like that at all! you lose!
what about this?
Not like! again
what about this?
Well, kind of like it. . .
Haha, you are wrong! I'm still a fake, you lose!
In this way, the two sides of the game evolve in the confrontation until they create an image that is so "real" that you can't believe it.
Obviously, if we replace one side of the above-mentioned evolutionary game with humans, will the artificial intelligence program on the other side surpass humans when human cognition reaches its limit? Very likely.
How long can human beings stick to this game of "one foot high, magic one foot, and artificial intelligence one foot five"? Soon, humans will find that the best algorithm is the one that excludes human cognition, and let artificial intelligence deal with artificial intelligence. Then we will find that this is an artificial intelligence arms race, and whoever has the resources to develop a stronger artificial intelligence algorithm can temporarily win. In our resource-bound economy and society, in the end only the monopoly with the most resources can have the ability to play this artificial intelligence arms race game, and those without resources, that is, the majority, will become losers.
But if there ever comes a day when humans who are no longer in the race may be deemed redundant by "artificial intelligence," that day will be a terrible one.
Therefore, our current resource-bound economy is simply not suitable for the era in which artificial intelligence has arrived.
Back to the issue of the challenge of "face changing" to privacy protection, what we need to do is to change our thinking.
Back to the other side of the question: If we don't want to win the evolutionary game against artificial intelligence, maybe we can try to do it the same way we evolved our own economic society?
One of the characteristics of "privacy-heavy" giant data is its deep coupling with our biometrics. We can try to go in the direction of "decoupling".
For example, one direction is to "decouple" our traditional single "identity" and biological characteristics. We should reflect on the role of single identities in our societies and economies. The information revolution started by the Internet has brought us into an infinitely expandable virtual reality. Our real world has already been greatly expanded. Our future is a world containing human thinking space, in which countless virtual spaces can be created. In these spaces, human beings will truly control their own destiny and coexist with artificial intelligence. Artificial intelligence will become an equal member of thinking human beings in this infinite world. No one can enslave anyone else, because this world is big enough. In every space we create, we can have countless "avatars", we no longer need a single "identity", and there is no need for a "single identity" that is deeply coupled with "biological characteristics".
(full text)
(full text)
