World : IBM Wants To Make Artificial Intelligence Fair And Transparent With AI OpenScale

IBM has announced AI OpenScale, a service that aims to bring visibility and explainability of AI models for enterprises.

When it comes to adopting AI for business use, there are multiple concerns among enterprise customers. Lack of visibility of the model, unwanted bias, interoperability among tools and frameworks, compliance in building and consuming AI models are some of the critical issues with AI.

IBM AI OpenScale provides explanations into how AI models are making decisions, and automatically detects and mitigates bias to produce fair, trusted outcomes. It attempts to bring confidence to enterprises by addressing the challenges involved in adopting artificial intelligence. The service is available within the IBM Cloud Watson product portfolio.

AI OpenScale acts as an intermediary layer positioning itself between the application and the machine learning models that deliver predictions. Each time an application consumes the model, the inbound (data points) and output data (predictions) are logged in a centralized database. The service also comes with a database to store training and feedback data which is used to retrain and deploy the models. AI OpenScale uses both the databases to analyze bias, accuracy, and explainability of models. Users can zoom into a specific transaction to understand the factors that influenced the outcome from predictive analytics. The service continually monitors AI applications to prevent bias through an automated de-biasing technology.


Source :

Smart Grid Bulletin February 2019

View all SMART GRID Bulletins click here