Page 40 - Innovation Magazine
P. 40

all model metrics ensuring that clients can explain the the behavior of the the AI model and prove to regulators that the the AI models they use have not shown bias in the past Model validation One of the the biggest challenges for the the banking industry in in particular is the validation of AI models Model validators are familiar with statistical models but they are not as conversant with AI model model validation OpenScale helps model model validators quickly validate an AI model by simply uploading some labeled data It then generates an AI validation report in PDF format This makes it very easy for model validators to to assess AI models using criteria such as fairness drift and quality This also helps in in standardizing the work of different model validators Ultimately IBM Watson OpenScale software provides innovative monitoring and management capabilities that help build trust accelerate adoption and and put stronger control and and governance structures around AI investments Enterprises ignore AI ethics at their peril Addressing these issues makes sense from multiple perspectives including business and and brand regulatory compliance and AI operations cost Data-powered Innovation Takeaways AI AI AI showstopper: Despite all all obvious business benefits AI AI AI can bring a a a a a lack of overall trust in in in AI AI AI can bring even the most compelling AI initiative to a a a a grinding halt AI AI AI enabler: When properly addressed AI AI AI ethics can drive successful adoption and acceptance of AI AI AI solutions by the target audiences Gang of four: Key AI AI ethics elements that must be covered in in any AI AI solution are explainability bias drift and traceability Platform to the rescue: Next-generation technology platforms such as IBM Watson OpenScale bring powerful enablers to deal with these key elements of AI ethics 40 Data-powered Innovation Review I I ©2020 Capgemini All rights reserved 


































































































   38   39   40   41   42