AI’s adoption is having an impact on every day’s life: mass surveillance with facial recognition in Ethiopia, predictive justice in the USA,
refusal of a mortgage, scholar orientation, disease detection, political manipulation with fake news, etc.
All those aspects have given the rise to ethical AI, trustworthy AI, AI regulations and numerous guidelines on AI usage and system building.
Problem : Ethical AI is vast and confusing, many notions intersect and regulations seem vague and hard to apply.
Hence we ask the following research questions:
– RQ1 How can AI practitioners implement ethical AI principles and regulations in their systems ?
– RQ2 How can institutions audit AI systems and verify compliance ?
In this presentation, we propose a framework to clarify the ethical AI ecosystem.
We decompose the broad field of AI research into challenges, from utility to resource consumption.
For each challenge, we identify technical, legal, societal and environmental aspects.
The goal of this work is to build a framework for ethical AI that takes into account the reality of the field.
We want system builders to be able to rely on our framework to understand regulations.
We want institutions to use our framework to conduct audits of AI systems.
- There won’t be (too much) math .
