Document :

UIT-D SG02 RAPPORTEUR GROUP MEETING  Contribution  80

Titre :

Scalable application of industrial artificial intelligence based on dynamic ontology and causal iteration

Reçu le :

2026-09-24

Source :

China (People's Republic of)

AI/Question :

Q5/2

Réunion :

2026-10-07

Disponibilité :

Réservé aux utilisateurs TIES [UIT-D]

Résumé :

At present, general large models have proliferated rapidly owing to their excellent capabilities in natural language understanding and multi-task coordination, yet their randomness and black-box nature make it difficult to meet the rigid requirements of highly reliable and highly complex industrial scenarios, becoming a core bottleneck that constrains the deep empowerment of the real economy by industrial AI. By introducing an approach to industrial large models based on 'dynamic ontology + causal iteration'-building a dynamic ontology knowledge base with physical mechanisms and process rules to achieve controllable reasoning under mechanistic constraints, and relying on a closed loop of production data to continuously mine and precipitate tacit causality-industrial AI is intended to support a shift beyond 'probabilistic generation' toward 'mechanism-controllable, causally-traceable, and experience-precipitable'. The approach has been applied in industrial production environments and is intended to address challenges associated with the controllability, traceability and reliability of industrial AI applications. The related exploration and practice may help lower the experience threshold for countries in developing their manufacturing sectors and may contribute to the upgrading and development of manufacturing industries worldwide.