The works we saw in the exhibition are actually two projects that are related, they are Low AI Box and Organic Element โ . For an unknown reason, among 20 projects of Da Zheng, these two are introduced only in Chinese. So the following blog is processed by my translation.
What it is about
- Low AI Box:
It is both a so-called “flesh packet” and AI machine. It integrates collected life data from the artist and the environment, then uses it as samples to train the machine, which serves as “the algorithm source of dynamic installation art and video art”.
In a way, it generates an “AI flesh of mixed data” of the artist. The artist can quantify himself through sensors, transcend the limits of physical perception, evolve, and live forever in the form of an AI corporeal body.
From another point of view, the machine acquires the ability to almost simulate the human body, while the human becomes the mere data provider of the machine. From this perspective, the machine acquires “autonomy”. In short, the machine is the ultimate creator and the human is only the data provider.
Here is a sentence that I did not fully understand:
“The way the low ai box works is the productivity of the artist, and the low ai box becomes the meta-language of the artist’s creation.”
- Organic Element โ
This work is called “Organic Element I”, which is equivalent to a vessel of data. The “vessel” is filled with organic liquid. The electronic components connected to the intelligent data and AI data were soaked into the pool which forms an artificial cybernetic space. This space is equivalent to an electronic ecological landscape with its own ecosystem.
In essence, both works are trying to express the reduction of the difference. The difference between human-generated data and machine-generated data, or the difference between the real world and the artificial virtual world. The reduction of the difference from the data perspective is the process of getting to zero.
How it relates to ML & AI
It uses supervised learning (like us) to train the model, which greatly increases machine involvement and accuracy. Therefore, with more training, the machine will also become more accurate and have a greater ability to anticipate, which means more “autonomy”.
It uses the transformer model, which is actually “used primarily in the fields of natural language processing (NLP)[1] and computer vision”. Therefore, I don’t fully understand why Da Zheng chose this particular model. My guess is that this model has the mechanism of self-attention, which examines the relationship between different parts of input data and focuses more on the important parts, perhaps more similar to the way the human body works?
A brief introduction of Da Zheng’s work
Da Zheng’s work conveys many consistent ideas. It’s not the focus, but it’s worth recounting.
One of them is about transparency. He established the Low Tech Lab because he has always tried to avoid the mystery and alienation that comes with high tech. He also hopes that by making the process transparent and visible, he can increase the audience’s understanding and intimacy with the work, which is why both projects are extremely transparent that we can clearly see how the internal parts of the machine work.
His creative journey has also been somewhat consistent. The first “Generate Leads” is a program that simulates his habit of using the mouse, which uses the machine to simulate human behavior. The subsequent “The Physiological Reaction“, “Luster of Unknown Universe” and “Skin of the Machine” are machines simulating nature, but the machine itself has a certain autonomy. It is re-expressing (relaying) the phenomenon with the logic of the machine itself. Low AI Box, on the other hand, is completely transforming data into an AI system, resembling and replacing the original data to a greater extent.
How I experienced it
I think these two works are still very basic (especially when compared with his other works), and still in the development stage (kind of routine of him, for example, he made three versions of The Physiological Reaction), so they do not fully express what he wants to show (maybe it’s a problem of ideas or technology).
Apart from these two works themselves, what actually inspires me more is the development of Da Zheng’s works. Along with this process, we can clearly see the difference in the positioning of machines. From the initial imitation of people and environment to showing some autonomy, to having more autonomy. And the increase of autonomy is supported by machine learning.
This also inspired me to think more about the application of machine learning, or the level that machine learning can make machines reach. I think I probably started out at the first level, thinking about how to apply machine learning to make machines more human-like, or to replace people to do a certain job. But I was more inspired by his second layer (and more attracted by his second layer of works) that maybe machine learning can also let machines gain their own vitality, instead of just imitating to reproduce humans.
A fragmentary record that has nothing to do with machine learning
Kapitulieren from EMMA6: (The ending part is brilliant)
In that gray, colorless, empty room that echoed with the chirping of machines and birds, I seemed to be out of focus.
Standing in front of the big screens, I never thought that summer would break into my world after pressing that button. Therefore, I didn’t even react for a moment to what I was seeing when the colors rushed to me.
A crossroads that seemed close at hand. A girl, wearing a halter dress, walked into the screen and started waiting to cross the street. That world was peaceful, tranquil, colorful, far away, a reality that belonged to another world. Suddenly I realized that it was a real-time camera. It was at this particular moment that there was a girl, and I was watching her, waiting unaware and innocently in some unknown place of the earth.
At that moment, it seemed that everything I was torn, sad and painful no longer mattered. I surrendered. I exiled myself. I escaped from reality, from sorrow, from my life. Was that a fall, or an ascent?