L0V3 SP3LL

 
#Process

#Meaning

#Direction

29 OCT 2023

︎ Tag: Sandbox, Implementation
︎ Duration: 2 days
︎ Method: Visual Design, Markov Chain
︎ Tools: ChatGPT3, Figma, VSCODE, TouchDesigner
︎ Place: Seattle

A satirical piece on pop love. 

For my project proposal, I made what Laura called a “handmade” neural network. I explored my memories, reflections, and dreams, then pasted them together into an organized (!) stream of consciousness. My focus is on exploring the alternative texture of love and intimacy - one that decenters Romance, Patriarchy, Heteronormativity, and Commercialism.

In my first exploration of this stream of consciousness, I tapped into my previous project where I translated what was at that time Billboard Top 1 song: STAY by Justin Bieber.







Before going into the execution phase, I clarified on my...

Preliminary

• Data set
• Medium
• Concept
• Method

1.  Data set

I want to appropriate content from popular culture in order to create a disruptive narrative and cultural critique piece. What I ended up using was lyrics from popular love songs, curated by yours truly and ChatGPT3.

2. Medium

I decided on web media. Taking in my low-code abilities, this is the most accessible medium for me to create interactive content within a short timespan so I could test and explore this project’s possibilities.

3. Concept

My preliminary concept was to appropriate popular love lyrics. This is a way to challenge the commercialization of love and question our notion on love. The expansive concept that’s realized by executing this project is added as I go.

4. Method

1. Find and clean up data
2. Code: Apply Markov Chain
3. Web design on Figma then export to Cargo 
4. Test

I first probed ChatGPT3 to give me 50 highlighted lines from popular love songs and put them into a spreadsheet. It could give me up to 200 lines at once before telling me to do it myself because it did not want to violate copyright laws. To which I said, “please?” After some more probing, we have created a dataset which contains 471 lines of popular love songs.

Next, I integrated them into VS Code, using JavaScript code I either borrowed from ChatGPT or found on GitHub forums (I'm not a pro coder yet but I do somewhat understand the syntax) to implement the Markov Chain method. What this does is generating new sentences from the existing love lyrics.

The generated lines using Markov Chain were like things a chicken would say using human syntax. They are funny but not that sensical. My favorite example is:







So I thought a bigger dataset would help Markov Chain do a better job in creating human-like sentences. I turned to GitHub again. I did find a bigger dataset (with 1.7 million songs to be exact!), but with it brought more problems:

     (A) the dataset is actually too big and my laptop cannot process it
     (B) the data is not about love exclusively.

Hiro Watari helped me read the (pickle) files, using Pandas in Jupyter Notebook. But the dataset was still too big for us to do anything on our laptops, so we turned to a dataset of 50K songs, although they were not about love. It took a few hours to realize that these data might not be helpful at all since it was generating more syntaxically incorrect sentences. Notably, although the dataset was not about love, it tend to generate sentiments about love anyway:








After getting the dataset down, we learned that Cargo has too many bugs, so we started developing our example locally instead. The new generated lines were of much higher human-sytax quality:






















My search for data with ChatGPT3 affirms that our public discourse on love is anglo-centric and patriarchal. It defaulted “love” to romance, and the meanings of lyrics are quick to grasp within a sentence, alluding to what I call “a quickie culture”.

︎ Anglo-Centric

︎ Male-Dominant

︎ Default Love to Romance

︎ Quickie Culture

Songs are like spells, they are repetitive and have the capacity to haunt, hypnotize, and shape one’s mind. As cultural consumer and producer, it is crucial that we pay close attention to what thought-patterns we are circulating, what spells we are chanting personally and collectively.










I want to associate this project with the contemporary discourse on WITCH HUNT, wherein women who spoke out are once more subjugated and denounced, their projects being labeled "WITCH HUNT." Thus, the design intentionally incorporates the Witch and Dark Academia aesthetics. The background for the start page depicts the entering of a vagina.

I also had created a circle using TouchDesigner that could transform into a believable vagina or a womb (forgot to take photos). I want some fixes before incorporating them but I lack the technological skills to do so for now.

The logo in the start page incorporates Japanese, which wrote ‘OCCULTIST EXPO’. The text is in Japanese because of my collaboration with Hiro, who is Japanese, but also (and mainly) to give the project a disorienting sense that this is not from a White culture, yet the generated lyrics and content are very anglo-centric, displaying the dominant voice of white entertainment/culture.

For the generated lyrics, I replace all letter “e” with number “3”. This is an association with the text code “<3”, which implies “love” or a heart, and sometimes 3s are used as the replacement for letter “e”.  Hopefully, this association gives rise to the feeling of abruptness and shortness that reflects our pop / text culture (the quickies).

Because of the short timespan of the first implementation, many intended associations were not fully established. I want to continue strengthened these messages visually as the project develops.

#1 Improve code & dataset to get human syntax. Or not?

#2 Deepen project concept with comparative research on spells and pop songs. Find existing artworks / concepts to strengthen the message.

#3 Try to answer: How to create an interactive & immersive experience?

continue translating some popular “love” songs to display their violence, then use projection mapping and TouchDesigner to create the

hide and seek effect ︎︎︎

or something like this

treasure trove ︎︎︎

perhaps: make “candies” for people to go home with: use TouchDesigner and Arduino to track whenever people bow, then print out the generated love spell using

generative printer ︎︎︎

Would this project be more sustainable (a longer lifespan) if I were to make it live on the web only (instead of turning it into an IRL installation)? What would give the audience a deeper shock?