AI Trust Engine for Multi-Vendor Pipeline Interoperability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies lack effective methods for implementing trustworthiness management of artificial intelligence (AI) and machine learning (ML) models, particularly in interoperable and multi-vendor environments, where control and evaluation of AI/ML model trustworthiness are necessary but not adequately addressed.
Innovation Solution
The introduction of a Trustworthy AI (TAI) framework for cognitive autonomous networks (CANs), which includes an AI Trust Engine and AI Trust Managers, facilitates the definition, configuration, monitoring, and measurement of AI/ML model trustworthiness through defined interfaces and APIs, ensuring fairness, explainability, and robustness across AI pipelines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional AI/ML models are deployed in multi-vendor environments, then system complexity and interoperability challenges increase, but trustworthiness control and evaluation become ineffective
Solution Approach 1:
The patent introduces an AI Trust Engine as an intermediary component that mediates between AI/ML models from different vendors and the trustworthiness evaluation framework. This engine provides standardized interfaces and trust metrics, enabling effective trust control without requiring direct complex interactions between multiple vendor systems, thus resolving the contradiction between reliability improvement and system complexity reduction
2Reliability
If comprehensive trustworthiness evaluation is implemented, then fairness, explainability and robustness are improved, but computational overhead and processing time increase
Solution Approach 1:
The patent implements preliminary trustworthiness evaluation by assessing fairness, explainability, and robustness metrics during the AI model development and deployment phases rather than only during operation. This allows trust metrics to be pre-computed and cached, reducing the computational overhead and processing time required for real-time trustworthiness evaluation while maintaining comprehensive assessment coverage
3Adaptability or versatility
If standardized trust interfaces are introduced for multi-vendor environments, then interoperability and trust management are improved, but implementation complexity and integration effort increase
Solution Approach 1:
The patent designs the AI Trust Engine with universal, multi-functional interfaces that can work with AI/ML models from different vendors through standardized protocols. These interfaces perform multiple functions including trust metric collection, evaluation, and reporting in a unified manner, improving interoperability while reducing implementation complexity by avoiding the need for separate integration solutions for each vendor
Data Source
AI summary
There are provided measures for trust related management of artificial intelligence or machine learning pipelines. Such measures exemplarily include, at a first network entity managing artificial intelligence or machine learning trustworthiness in a network, transmitting a first artificial intelligence or machine learning trustworthiness related message towards a second network entity managing artificial intelligence or machine learning trustworthiness in an artificial intelligence or machine learning pipeline in the network, and receiving a second artificial intelligence or machine learning trustworthiness related message from the second network entity, where the first artificial intelligence or machine learning trustworthiness related message includes at least one criterion related to an artificial intelligence or machine learning trustworthiness aspect.