Symbolica

Guide Labs is developing a new class of interpretable AI systems and foundation models that humans can reliably debug, trust, and understand. Our models are engineered to produce human-understandable factors for any output, provide reliable context citations, and specify which training data influences the generated output. This approach addresses issues in current AI systems, which often produce explanations unrelated to their outputs, are difficult to debug, and are challenging to control and align. The Guide Labs team comprises experts with over 20 years of experience in interpretable machine learning. We have developed the first interpretable generative diffusion model and large language model. We are rethinking the model architecture, loss function, and entire pipeline to constrain the model training process such that the models we get are more easily understandable, their errors easier to identify and fix, and easy to align.

​FalkorDB is an ultra-fast, multi-tenant graph database optimized for GraphRAG, delivering accurate, relevant AI/ML results with reduced hallucinations and enhanced performance. It leverages sparse matrix representations and linear algebra to efficiently handle complex, interconnected data in real-time, resulting in fewer hallucinations and more accurate responses from large language models. FalkorDB supports the OpenCypher query language with proprietary enhancements, enabling expressive and efficient querying of graph data. It offers built-in vector indexing and full-text search capabilities, allowing for complex searches and similarity matching within the same database environment. FalkorDB's architecture includes multi-graph support, enabling multiple isolated graphs within a single instance, ensuring security and performance across tenants. It also provides high availability with live replication, ensuring data is always accessible.