Building more trustworthy and ethical AI systems is not only a research question it’s a business, legal and societal imperative

Andrea Martin, the leader of the IBM Watson Center Munich and Member of the Bavarian AI advisory board, was the guest of the IEAI’s Speaker Series that took place virtually on January 21,2020. In her presentation, she explored in depth the topic of trustworthy AI and the importance of implementing mechanisms to increase trustworthiness in AI systems being developed in order to increase consumer acceptance.

As part of her discussion, she outlined IBM’s engagement with developing trustworthy AI solutions. This starts by understanding the prerequisites for trust in AI by analyzing the best practices available from previous research in software development and engineering.

To this end, IBM published core principles for the development of AI solutions, such as augmenting human intelligence, the protection of the intellectual property of data (and the insights retrieved from the data), and last but not least, transparency. In order to implement these guidelines in reality, IBM founded an AI Ethics board that provides centralized governance by reviewing the new solutions being developed. For instance, IBM took the bold step of deciding not to pursue the advancement of facial recognition systems until the extent of their use is clearly defined, particularly in law enforcement contexts.

Andrea Martin also introduced the methodologies and tools implemented by IBM to achieve the development of ethical solutions. This includes, for example, solutions used in data-mining to analyze the quality and the sources of the data being analyzed. Furthermore, they initiated open-source toolkits that can contribute to the analysis of important factors such as fairness, explainability andmanipulation of data prevention. This is an important step in achieving practical solutions to the complex dilemmas emerging from the development of AI systems.

She concluded by analyzing the impact of AI solutions in the day to day world and how AI enabled tools may impact employees in their daily tasks, such as the use of chatbots that answer frequently asked questions, allowing HR personel to focus on more challenging tasks.

We want to thank Andrea Martin for her thoughts and time, as well as the insightful discussion that followed her talk. The discussion session touched heavily on the current efforts of technology companies in building trustworthy AI systems, as well as where gaps in AI governance still remain.

We also had the pleasure of speaking with Andrea Martin in an IEAI Q&A: Reflections on AI. We thanks her again for her time and the interesting discussion and hope to have her back to the IEAI soon.