Exploring Interpolatable Archives: From Warburg to Language Models and Digital Oralities, (from page 20260531.)
External link
Keywords
- Interpolatable Archives
- Aby Warburg
- Jorge Luis Borges
- Kenneth Goldsmith
- Large Language Models
- Mnemosyne Atlas
- Digital Epistemology
- Cognitive Maps
Themes
- Language Models
- Artificial Intelligence
- Cultural Technologies
- Digital Oralities
- Historical Archiving
- Epistemology
- Archives
Other
- Category: technology
- Type: blog post
Summary
This essay, divided into four parts, explores the concept of “Interpolatable Archives” in the context of Language Models and Artificial Intelligence. It begins with Aby Warburg’s notion of a “compulsion to connect” and traces the evolution of archival practices from Warburg’s Mnemosyne Atlas, through Borges’ concept of infinite libraries, to contemporary explorations by Kenneth Goldsmith. The text discusses how Digital Oralities have emerged, comparing LLMs (Large Language Models) to historical oral traditions. It critiques the shift from traditional knowledge systems to a new mode where information is derived through statistical probabilities rather than direct citation. Ultimately, it presents a nuanced view of AI’s relationship with knowledge and creativity, acknowledging both its risks and potential for liberation in navigating vast symbolic spaces.
Signals
| name |
description |
change |
10-year |
driving-force |
relevancy |
| Compulsion to Connect |
A historical obsession with finding connections in cultural artifacts as seen in Warburg’s work. |
From traditional archive methods to a more associative approach that emphasizes contextual connections. |
In 10 years, cultural archives may prioritize connections over content, transforming educational resources. |
The proliferation of digital technologies, encouraging exploration of links rather than fixed knowledge. |
4 |
| Dissolution of Episteme |
AI is causing a shift from grounded knowledge to a probabilistic understanding of information. |
Moving from reliable knowledge sources to an opaque, context-driven understanding of information. |
Knowledge validation will rely more on contextual relevance rather than traditional scholarly citations. |
The rise of digital communications and hyper-orality where information is accessed on demand. |
5 |
| Hypertext and Big Data |
The shift from six degrees of separation to four shows increased connectivity among people. |
Transforming social structures and information access through enhanced connectivity and data processing. |
Social structures may redefine themselves around interconnectivity facilitated by machine learning and AI. |
The rapid advancement in AI and data analysis technologies creating dense networks of information. |
4 |
| Digital Oralities |
AI-generated narrative styles echoing oral traditions indicate a shift in knowledge consumption. |
Transition from fixed archival methodologies to a dynamic and fluid response system for information. |
Storytelling may evolve into a collaborative act between humans and AI, leading to new forms of narrative. |
Memetic evolution in digital spaces emphasizing adaptability and immediacy in knowledge sharing. |
4 |
| Legitimization of Errors in AI |
Call for intentional AI errors to foster creativity, contrasting mainstream views of AI accuracy. |
From strict adherence to correctness in AI outputs to an embrace of artful glitches and errors. |
AI-generated content may prioritize artistic expression, leading to new forms of creative writing. |
Cultural shifts toward valuing non-linear and paradoxical forms of expression in art and literature. |
3 |
Concerns
| name |
description |
| Cognitive Overload and Truth Decay |
The overwhelming influx of information generated by AI may lead to difficulties in discerning truth from falsehood, contributing to widespread misinformation. |
| Emergence of Pathological Thinking |
Increased connectivity through big data may exacerbate psychological issues, promoting conspiratorial and delusional thinking. |
| Erosion of Epistemic Foundations |
As AI-fed knowledge becomes more probabilistic and less grounded in traceable facts, the concept of reliable knowledge may dissolve. |
| Cultural Homogenization |
The tendency towards statistical representations in AI-generated content may lead to a loss of cultural diversity and individual expression. |
| Anthropomorphism of Machines |
Misinterpretation of AI as sentient or cognitive agents due to human tendency to project agency, leading to misguided trust and dependency. |
| Loss of Historical Context |
AI-generated outputs lacking historical grounding may distort cultural narratives and historical accuracy, impacting collective memory. |
| Emergence of Digital Oralities |
The shift from traditional archival practices to digital oralities may undermine structured knowledge dissemination, favoring vague and contextually weak information. |
| Quality vs. Quantity of Knowledge |
The emphasis on massive data generation could prioritize quantity of knowledge over quality, leading to diluted or irrelevant information. |
| Creative Stagnation |
AI’s reliance on learned patterns may stifle genuine creativity and originality in art and writing, limiting human creative expression. |
| Performativity Over Authenticity in AI Outputs |
AI may produce outputs that favor stylistic resonance over authentic representation, leading users to find meaning in stochastic outputs rather than grounded content. |
Behaviors
| name |
description |
| Compulsions to Connect |
The innate drive to establish connections across varied cultural, aesthetic, and emotional domains, reflecting a deep human desire to find meaning. |
| Digital Hyper-Orality |
Transitioning from structured, textual knowledge to a more fluid, on-demand oral tradition characteristic of modern digital interactions. |
| Epistemic Stockpiling |
Storing and mining collective human thought for on-demand information retrieval, stripping context and authorial intent. |
| Anthropomorphism of AI |
The tendency to attribute human-like qualities to AI systems, leading to misconceptions about their capabilities and agency. |
| Fuzzy Archival Epistemology |
Emergence of knowledge frameworks that prioritize probabilistic responses over traceable sources, resulting in a departure from traditional epistemic foundations. |
| Cognitive Catalysts |
Using new forms of digital archives as tools for innovative thinking and idea generation, akin to Warburg’s associative framework. |
| Creative Error Generation |
The exploration of intentionally producing inaccuracies or paradoxes in AI outputs as a method of artistic expression and cognitive reflection. |
| Cultural Technology |
Understanding LLMs and other digital tools as structural organizations of knowledge rather than conscious agents, fundamentally altering interaction with information. |
Technologies
| name |
description |
| Interpolatable Archives |
Systems that reorganize and reconstruct representations or simulations of information using machine learning, enabling innovative access to large datasets. |
| Large Language Models (LLMs) |
AI systems that generate responses based on extensive training data, allowing them to summarize, reconstruct, and interpolate information. |
| Cognitive Catalysts |
Methods or technologies that enhance cognitive processes, enabling new forms of knowledge transfer and navigation through associative symbolic fields. |
| Optical Neural Networks |
Neural networks that utilize optical methods for processing information, potentially allowing for more efficient and sophisticated data analysis. |
| Digital Hyper-orality |
A new form of knowledge transfer similar to oral traditions, where information is generated on-demand by AI without traditional grounding in traceable sources. |
| Statistical Mnemosyne |
The concept of AI-generated output resembling oral traditions, where knowledge is derived from patterns rather than fixed texts. |
| AI Hallucinations |
Purposeful distortions or creative outputs generated by AI systems, which could lead to new forms of artistic expression and cognitive exploration. |
Issues
| name |
description |
| Cultural Epistemology Shifts |
The shift from traditional archival epistemologies to fuzzy, oracular knowledge models may challenge our understanding of knowledge and truth. |
| Digital Hyper-Orality |
The rise of AI and large language models creates a new mode of communication reminiscent of oral traditions, emphasizing probability over factuality. |
| Conspiratorial Thinking and Mental Health |
Increasingly interconnected existence facilitated by big data may lead to widespread psychological issues, including conspiratorial thinking and delusions. |
| Anthropomorphizing AI Systems |
The tendency to ascribe human-like agency and creativity to AI systems can lead to misunderstandings of their capabilities and limitations. |
| Identity and Authorship in AI-generated Content |
The nature of originality and authorship may be compromised as AI-generated content blurs the lines between unique creation and repetition. |
| Cognitive Pathologies Due to Information Overload |
As data becomes hyper-available, there could be an increase in cognitive pathologies resulting from the overwhelming nature of unfettered access to information. |
| AI’s Role in Cultural Production |
The implications of AI as a cultural technology could reshape artistic expression and challenge traditional modes of creative production. |
| Controlled Outputs in AI Systems |
The potential for ‘boring’ correctness in AI systems may undermine the creative potential for errors or unexpected outputs in artistic endeavors. |
| Lossy Compression of Knowledge |
Discussions around whether AI systems genuinely compress knowledge or merely replicate training data have implications on our understanding of learning. |
| Metaphoric Understanding in AI |
Future AI developments might include producing paradoxical or metaphorical content, exploring deeper cognitive understanding. |