RAIDO

While solutions for Trustworthy and Green-AI do exist, they are mostly fragmented and myopic focusing on providing very narrow and isolated solutions functioning only within their own boundaries without ensuring interoperability. To alleviate the limitations of existing solutions, RAIDO aims to make transformative steps towards Trustworthy and Green-AI by offering a holistic and comprehensive framework solution covering all data- and model-related aspects of these problems within an integrated platform.

Trustworthy AI is driven by the need to ensure that AI systems are reliable, safe, and unbiased in their optimisation and deployment processes. The increasing use of AI in critical applications such as healthcare, industry, finance, and criminal justice has highlighted the importance of trustworthy AI. Additionally, with the increasing amount of data being generated, there is a need for AI systems that can make the best use of this data while also being transparent and accountable in their training and deployment stages.

The RAIDO platform will offer methods for automated data curation and enrichment, including Digital Twins and diffusion models, providing a pipeline for high-quality, representative, unbiased and compliant training data which ensures highly performing and trustworthy AI model development. This will include 4 real-life demonstrators that cover key application domains (namely smart grids, computer vision based smart farming, healthcare, and robotics for bio-based composites) with notable societal and market impact.

RAIDO will clearly define the data used by all its activities and will clearly define the aspects, nature, and utilisation of the data ensuring fair, transparent, and lawful conduct. It will limit data collection to the most attainable degree that has no negative impact on the project outcomes and results, thus, mitigating any potential misuse or undesired utilisation. The project will clearly define, through the related deliverables, the flow of data and will specify rules on data collection, storage, and usage. In addition, when necessary, all data will be properly pseudorandomised or anonymised.

RAIDO brings together a team of high experienced and skilled partners (including 14 SMEs) from 9 EC member states, UK and India, dedicated to delivering the outcomes promised. The majority of the SMEs are data driven and will contribute data sources.

There’s a number of tangible results that we are planning at this initial stage. In terms of data these are the RAIDO Model and Data Repository Search, Innovative KPIs that measure how AI-enabled systems perform, and Green-AI & Data Orchestrators. In terms of ethical and unbiased data for Trustworthy AI training, and AI explainability we are looking at a decentralised blockchain mechanism with IPFS support for decentralization and efficient AI models optimisation that guarantees data privacy and transparency for Regularity and Progress Monitoring, RAIDO feedback-based Reinforcement Learning solutions, and monitoring for continuous AI model & Data optimisation and Improved GDPR legislations for privacy associated with data synthesis.