The Landscape

CLAIS Knowledge Graph

The CLAIS Knowledge Graph System (CKGS) is a software platform being developed by CLAIS. It provides views through which a Member can grow, edit, explore, and query the CLAIS Knowledge Graph (CKG). A large directed acyclic graph (DAG) of semistructured content, the CKG can also be queried using the CKGS to dynamically produce guidebooks, checklists, application safety reports, technical diagrams, and more.

Topics

The CLAIS Knowledge Graph is intended to cover topics granularly across safety engineering, assurance, trustworthiness, robustness, both shorter- and longer-term AI safety, control theory, and a host of other related topics and disciplines. On our AI safety page, find an early overview of some of the key areas covered. Documented subtopics currently number in the hundreds.

Multiple Perspectives

AI safety is a very interdisciplinary community of interest. As different disciplines often bring disparate terminology for very similar ideas, and different organizations bring different perspectives regarding centrality or hierarchy of concepts, we have found it useful to enable the full expression of these views, or contexts, while also facilitating dense interconnections among them. Each context can be visualized as a 2D planar diagram, and would in turn be richly connected with other contexts to form the larger 3D CKG.

CLAIS Default Taxonomy

We encourage each member organization to express and put its stamp on its unique perspective on the AI safety space. When that is not possible or there is only some one-off contribution regarding a single topic, those additions can live on the CLAIS Default Taxonomy (CDT). The CDT can be thought of as a catchall scaffolding, a default view when one is needed, but explicitly not a primary view overshadowing other sources, contexts, or perspectives.