Responsible AI: For Whom, By Whom, and How? скачать в хорошем качестве

Responsible AI: For Whom, By Whom, and How?

Abstract: There have been multiple proposals on how to develop and deploy AI models responsibly. Nonetheless, each of these proposals comes with their own shortcomings. In this talk, I highlight the complexity of fully mitigating AI bias and harms. By going step by step through the data & model development-deployment pipeline, I highlight how different decisions made by different stakeholders about different stakeholders affect the model inferences. By scrutinizing the for whom, by whom, and how, I formulate concrete considerations and offer recommendations on how to develop AI models responsibly and inclusively, all while transforming unconscious actions and effects to conscious decisions. Responsible AI is not a one-size-fits-all process, therefore there is a need to create flexible and effective frameworks to mitigate unwanted outcomes. Speaker Bio: Orestis is Professor for Societal Computing at the Technical University of Munich and head of the Civic Machines Lab within the TUM Think Tank. Orestis’ research provides ideas, frameworks, and practical solutions towards just, inclusive and participatory socio-algorithmic ecosystems. He builds tools and performs foundational research on platforms and artificial intelligence. Orestis analyzes new and old media by the application of data-intensive algorithms, as well as the political and social impact of the use of data-intensive algorithms themselves.