KARPATHY SHARES 'LLM KNOWLEDGE BASE' ARCHITECTURE THAT BYPASSES RAG WITH AN EVOLVING MARKDOWN LIBRARY MAINTAINED BY AI
https://images.ctfassets.net/jdtwqhzvc2n1/5rACVd0XGWlMS5e6o1F14i/b0e0dd19d82776e22bffed439593afbc/Carl_Franzen_a_humanoid_robot_in_elegant_cravat_and_velveteen_3a76f138-c488-4d03-8016-c5db26b6a025_3__1_.png?w=300&q=30
โ๏ธ Andrej Karpathy, former Director of AI at Tesla, has proposed a new approach to managing research projects using a 'LLM Knowledge Base' that uses a persistent, LLM-maintained record of projects to solve the problem of 'stateless' AI development.
๐ LLM Knowledge Bases approach uses a persistent, LLM-maintained record of projects ๐ Rejects the complexity of vector databases for mid-sized datasets ๐ Relies on the LLM's increasing ability to reason over structured text ๐ System architecture consists of three stages: Data Ingest, Compilation, and Active Maintenance (Linting) ๐ Approach allows for a self-healing, auditable, and entirely human-readable knowledge base
-- #AndrejKarpathy #CharlyWargnier #VamshiReddy #OleLehmann #EugenAlpeza #JasonPaulMichaels #StephAngo #LexFridman #technology
Loading quoted postโฆ




















