Probability Space
summer 2023 | Immersive Lab Residency
Fall 2023 | Grad Lab Studio
Research + Interaction + Visualization
Generative AI to help explore the space of human cognition, perception.
Abstract
The 'Possibility Space' project utilizes generative AI by taking human memory as input to produce a collection of digital sculptures encapsulating blended memories. These sculptures intricately link to various stages and scenarios of human life. Grounded in an understanding of human cognitive processes, I developed a memory process flow. This involved employing generative AI, image processing techniques, and data visualization to craft sculptures that vividly represent the intricate tapestry of human memory.
Key Questions
- How does generative AI perceive and interpret the world?
- In what ways do memory and information shape the human cognitive system and world view?
- Can generative AI contribute to expanding our understanding of reality beyond traditional
- cognitive frameworks?
- To what extent can generative AI transcend, or challenge established human world views?
Research Collections
To comprehend the disparity between how AI perceives the world and how humans do, we must initially delve into the human process of understanding reality.
‘Blind Paradox’
Hume's skepticism has piqued my interest due to its demonstration, rooted in the traditional ontological perspective, that we cannot directly deduce any knowledge from reality but rather through inference. Our acquired knowledge is essentially an inference based on our interpretations of reality. The well-known "Blind Paradox" illustrates this concept, suggesting that the color perceived as 'Red' in one person's eyes could be perceived as 'Green' in another's eyes. Despite this difference, both individuals would label it as 'Red,' having reached an agreement within their linguistic framework regarding the representation of this color.
‘Possibilities’
In the realm of AI, a parallel process occurs during the training of Large Language Models (LLM) and generative AI. We may establish agreements with AI agents, trusting that they can produce outputs in a rational and responsible manner. However, there exists a potential disarray in the connection between representation, reality, and our linguistic frameworks—whether visual or textual.
This process can be seen as either a challenge or an opportunity, providing a means to transcend our cognitive system and linguistic framework, allowing for the exploration of a broader spectrum of possibilities related to reality or its representation. For instance, in my course paper titled 'Beyond the Algorithm: Unraveling the Influence of AI Models on Our Perception and Reality,'
I delve into the idea that the number of matches played by an AI player could potentially be exponentially larger than that of a human player.
Over the course of nine hours, the chess version of the program played forty-four million games against itself on a massive cluster of specialized Google hardware.
(The New York, December 28, 2018)
Thinking
This aspect of AI prompted me to contemplate whether we can employ generative AI as a tool not only to probe possibilities within strategic games like Go but also to venture into the realms of our imaginative spaces.
Humans utilize both textual and visual language to depict the world, acquiring knowledge through these representations. Thus, leveraging generative AI opens opportunity to explore the Possibility Space of human imagination.
Methods
To connect our perception to the representation of reality, we can create image, 3D spaces and go through the texture, shaders, material and the key elements from our cognitive system.
Three general features in human cognitive systems involves: Perspectival, temporal, indeterminate. (Madary, Michael. Visual phenomenology. MIT Press, 2016.)
We expending, implementing, and replacing elements in our memories to access new perception and memory.
Prototyping
Based on: TouchDesigner, Stable-Diffusion, Python
Process Overview
Imgae Generation
Segmentation, Image Regeneration
Stage 01 Output
![]()
Stage 02 Output : Image to Point Cloud
![]()
Stage 02 Output : Image to Point Cloud
Concept Developmet
Perception and Memory Segmentations
![]()
![]()
![]()
Perception and Memory Segmentations
From a phenomenological perspective, the constituents of our memory are derived from objects in reality.
We gradually form a holistic comprehension of reality through these representations, and this information traverses our cognitive system, guiding the development of our fundamental sense of space and time.
Our memory is constructed in a continuous and immediate manner.
Stage 03
The objective of this project will center on exploring the impact of our memory on perception, as well as the emotions and feelings arising from our cognitions.
The aim is to establish a comprehensive virtual space that unveils the intricate connections among memory, perception, and cognition.
We gradually form a holistic comprehension of reality through these representations, and this information traverses our cognitive system, guiding the development of our fundamental sense of space and time.
Our memory is constructed in a continuous and immediate manner.
Stage 03
The objective of this project will center on exploring the impact of our memory on perception, as well as the emotions and feelings arising from our cognitions.
The aim is to establish a comprehensive virtual space that unveils the intricate connections among memory, perception, and cognition.