The most recent concept that I’ve been working on is creating a narrative based in a speculative future in which machines with different specializations are working on a research project. Their project is about figuring out and writing the history of earth, more specifically of what happened in the early 21st century. But all of their source material for this research study is a set of memes which are making no sense to the machines. So what ends up is a comical set of dialogs which makes no sense from an analytical point of view but are hilarious to read.
My idea behind this project is to create a humorous view on how machines interpret everything in a scientific way. One aspect that makes the world of humans or human thinking different from the world of machines is the frivolity or the nonsense of everyday life. We are constantly talking about the age of AIs and how Artificial Intelligence is supreme to humans and might take over the world some day, it is important to also talk about what makes us great as humans. It’s the small things, the imperfections in our nature that inspires such humor that is a significant part of our everyday lives.
From the execution point of view, I’m planning to approach this narrative by using machines to tell the story of machines. I’m training different machine learning models on different source materials. For example, to one model I’m feeding lots of archaeology books, to another, I’m giving more research journals, while to another I’m planning to feed a lot of science fiction novels. With this, I’m planning to make a set of machines with different personas and hence different understanding of Memes. These models will imitate machines with different areas of specialties. After I have a set of different machines, I’ll initiate a conversation between these models/machines and build a narrative out of it. The narrative will be about these machines interacting with and interpreting different Memes in their own way and finally reaching a conclusion about it.
Reconstructing history though Memes by AIs.
Through this project while I’m aiming to create a narrative experience, I also wish to comment on the importance of humor and frivolities in our lives. Otherwise there would be no difference between humans and machines.
I guess my main design question here would be, what narrative do my AI models build? Or how do I use AIs to build a narrative that makes sense? Even as my project is more about the comments on machine-interpretation of humor, I feel that the overall narrative should make more sense so that this comes up as a complete project. One thing that I feel would work here is experimenting with a lot of data that I feed to my AI models. I’ll have to constantly work on the technical part while I look back and form a complete concept for my narration.
I have divided my topic into different parts. The first is the narrative, or the storytelling aspect of the project background. The one that sets up the premise for my project, about how some machines in future are trying to recreate history through using memes. The next is the technical part which is the conceptualization and creation of those so called machines, their personalities and the training of ML models based on their personality. The final part is the narration of the conversation that happens between these machines and how they deduce history. I feel that I should treat these parts differently in terms of how I present them. My first prototype, which was a series of blogs narrated from the point of view of one such machine, that has been reading some memes and trying to decipher the meaning behind it. I’ve added snippets from my published blogs below. The full blogs can be found on my medium page here.
I found the creation of this prototype to be highly exciting and entertaining for me while it also gave me an insight into how to look at this narrative from a scientific perspective, or the perspective that a machine might have.
The second prototype that I made which addressed the technical part of this was based on training a machine learning model on an archaeology book. I then fed this model text from a meme about the events of 2020 and it produced a very analytical style narrative about that text which was really interesting. I’ve attached a snippet of the prototype below.
While creating this I realized how machines would interpret anything based upon what they have been trained to do, no matter how ridiculous it sounds. This made me draw parallels between machines and humans and how both interpret comic elements like sarcasm, exaggeration and humor. This also led me to think of how through my project I could also comment on frivolities of human life and how it makes us great.
I made another prototype which is based on the third aspect of my project, which is the conversation between different machines. The prototype is a comic strip about two machines who are talking about their research about the “ancient beings (humans)” and discuss how they were extremely intelligent. The direction of this prototype (a comic strip) is not necessarily the direction I want to pursue for my project, but it definitely was an interesting exercise and helped me explore another direction for my project.
As a designer, I currently feel quite excited about the direction that I’m pursuing. I feel that this will have a lot of scope in narration and storytelling which would be very interesting. I feel my concept of building a world of machines who are working towards a common goal provides me with a lot of context to work with. For the storytelling part, the thing that I feel I need to focus on is the bigger narration. I feel quite confident with the smaller snippets or the small conversations and comments that the machines produce on the memes but I’m not sure where would this lead to? What would be the sum total of those conversations? Similarly from a technical point of view, though I feel excited to train different models imitating different machines personas, I feel challenged in trying to train them in a way that makes the conversation that they produce significant and different from each other.
I figure I have three tasks to pick up as my next steps. First one would be to define personas for my machines. What are the different personalities (archaeologist/ scientist/ engineer) that I want to be taking part in these conversations? I want to figure out which set of personas would be the best fit for this project, the personas that make the discussion most interesting. The second step would be to experiment with the technical part. To feed these personas into my models and test how are simple conversations forming up with these personas and if needed tweak the models accordingly. And then the third step that I think of is figuring out the broad narrative of the entire project. Where do I see my project going? I feel that through these experiments and steps, I’ll be able to figure out what I wish to make.