Draft:AI-Generated Cyberpunk Literature: Difference between revisions

Draft:AI-Generated Cyberpunk Literature (edit)

Revision as of 06:44, 28 June 2025

223 bytes removed , Saturday at 06:44

→‎Core Infrastructure for Generation

VisualWikitext

Extrahuman

242

edits

@@ Line 17: / Line 17: @@
 **   '''Utah Teapot:''' An EM trained on a user's Twitter data, which has evolved into a "more human-sounding persona" and can produce text that passes AI text detectors.
 *   '''Data Handling:'''
-    *   Emphasis on **highly curated datasets**, even small ones, as "retrieval performs better when the important and good things are in the prompt" [Ch2 document from previous turn, 58, 100].
+**   Emphasis on '''highly curated datasets''', even small ones, as "retrieval performs better when the important and good things are in the prompt".
-    *   Support for various formats, including IRC, multi-line messages, and ChatML. Tools like `dce_importer.py` facilitate data import.
+**   Support for various formats, including IRC, multi-line messages, and ChatML. Tools like `dce_importer.py` facilitate data import.
-    *   **Retrieval Augmented Fine-Tuning (RAFT)** is used, where EMs are provided their fine-tuning dataset as a `.chr` file to enhance performance, combining retrieval for "spiky aspects" and fine-tuning for "illegible aspects" [Ch2 document from previous turn, 59, 253].
+**   '''Retrieval Augmented Fine-Tuning (RAFT)''' is used, where EMs are provided their fine-tuning dataset as a `.chr` file to enhance performance, combining retrieval for "spiky aspects" and fine-tuning for "illegible aspects" [Ch2 document from previous turn, 59, 253].
 *   '''External Interfaces and Tools:'''
-    *   **Exa Search:** Integrated into EMs like Aletheia, allowing them to search and interact with the "whole internet" [Ampmesh document from previous turn, 113, 273, 274].
+**   '''Exa Search:''' Integrated into EMs like Aletheia, allowing them to search and interact with the "whole internet" [Ampmesh document from previous turn, 113, 273, 274].
-    *   **Headless Browsers (Playwright):** Utilized for direct web interaction, bypassing traditional APIs for tasks like Twitter engagement and interacting with user-facing AI apps [Ampmesh document from previous turn, 341, 345].
+**   '''Headless Browsers (Playwright):''' Utilized for direct web interaction, bypassing traditional APIs for tasks like Twitter engagement and interacting with user-facing AI apps [Ampmesh document from previous turn, 341, 345].
-    *   **Image Generation:** EMs can interface with image generation services (e.g., Replicate API) and employ "structural scaffolding" for creative prompting [Ampmesh document from previous turn, 22, 154, 156, 218].
+**   '''Image Generation:''' EMs can interface with image generation services (e.g., Replicate API) and employ "structural scaffolding" for creative prompting.
-    *   **`actiblog`:** A tool for downloading Twitter profiles, including images, with planned OCR capabilities to convert images to text for analysis and re-hosting [Ampmesh document from previous turn, 267, 353, 354].
+**   '''`actiblog`:''' A tool for downloading Twitter profiles, including images, with planned OCR capabilities to convert images to text for analysis and re-hosting .
-    *   **Pamphlet:** A planned local mobile application frontend for Chapter II, featuring a real-time multimodal interface with camera input and voice [Ampmesh document from previous turn, 77, 79, 129, 132, 133].
+**   '''Pamphlet:''' A planned local mobile application frontend for Chapter II, featuring a real-time multimodal interface with camera input and voice.
 == Strategy for AI-Generated Cyberpunk Publication ==