Enframing and Disclosure

Disclosure names the process of the opening of the world to intelligibility. Reasoning is the process of making sense of and determining truths from the open intelligibility of the world. Machinic reasoning fails not when the world is unintelligible, but rather when the process of Disclosure has failed in opening the world as such. At Gestell, we call Disclosure the process by which Enframed data is made available to and used by LLMs

What does this look like? Let us begin with Enframing. Enframing can be compared to traditional ETL processes. These processes, often programmatic, form the basis of modern data pipelining and ingestion. These processes weren‘t built with LLMs in mind, they were rather built to accommodate the modern enterprise data stack and business processes. The focus was on unification, ease of access and legibility across platforms. All of these things were good in their time, but that time has now passed

Enframing is the process of ingesting data, reading its contents and relating these contents to other portions of the data. Enframing is an all-encompassing ingestion and canonicalization process, it includes everything from table-friendly OCR, to intelligent chunking, to vectorization and canonization. Importantly, the stack for Enframing is holistic, each piece of the process relates to another and optimizing for whatever ingestion task is required according to the instructions placed upon it. The database is the world of the LLM. The world is Enframed for Disclosure

Why does Enframing need to operate as an ETL? Categories underlie understanding, structuring underlies cognition, both for humans but also for LLMs. Gestell is not an LLM but rather the architecture of understanding for LLMs. Data must follow schema in order to become intelligible - Gestell as ETL then is the process of the implementation and execution of rules for the conversion of ‘raw’ unstructured data into Enframed data. Gestell as ETL is unlike any other ETL seen prior. Gestell does not simply transform and validate instead it perceives, signifies and Enframes for Disclosure

What is Disclosure? Disclosure is the opening of the world to an LLM. An LLM that ‘reasons’ requires content for reasoning. Discursive reason is the processing of the relations of ‘facts’ into ‘truth’. Recursion unto falsity is the result not of ‘error’ but of a model’s discursive reasoning that cannot process the ‘world’ of the data. Lacking ‘means’ for processing data, the ‘ends’ are thus determined

What does this mean? LLMs weren’t made to interact with modern data pipelines and search methods. ‘Warehouses’ and ‘Data Lakes’ don’t contain readily intelligible data for reasoning but present a challenge in even beginning the search process to find the information required to cognize. The modern response to this would be to setup Retrieval Augmented Generation (‘RAG’) pipeline but RAG fails to understand that reasoning is not the result of relevance and structuring but the fruit of presentation and relation

Disclosure is the process of strategizing, searching and thinking through the contents of data. Disclosure is an all-encompassing search and judgement process, it includes everything from devising search strategies, to creating and analyzing citation and refining outputs. Disclosure maximizes computational efficiency because judgement is applied at each point in path determination. Search thus becomes a method of perception forming the presentation of the world, and its frames and relations, to an LLM. The Enframed world is opened by Disclosure

Want to transform your AI application with Disclosure? Enframe the world with Gestell