E228 ‘Responsible AI in Practice: Governance and Innovation’ with IBM Research’s Elizabeth Daly

Today’s guest is Elizabeth Daly, STSM and Research Manager of the Interactive AI Group at IBM Research. Artificial Intelligence stands as a monumental achievement, reshaping industries and prompting critical discussions about ethical governance. One compelling narrative in this discourse is led by Elizabeth and her AI Award winning team at IBM Research. In today’s podcast, Elizabeth dives into the complexities and nuances of building responsible AI solutions, highlighting projects like the Risk Atlas Nexus.

Businesses are increasingly realising the importance of responsible AI practices. A survey by the IBM Institute for Business Values revealed that top-performing companies in ethical AI frameworks gain substantial benefits from operationalising AI. Elizabeth reaffirms this by emphasising the necessity of creating user-friendly experiences that prompt responsible decisions, thereby preventing AI-related risks and losses.

Elizabeth introduces “Risk Atlas Nexus,” a research-driven initiative aimed at unifying governance and responsible AI practices. This initiative emerged from the need for a flexible entry-point enabling entities to incorporate responsible AI governance. Elizabeth elaborates on how AI landscape changes – with new benchmarks, metrics and models emerging daily – necessitate adaptable and consistent solutions. By focusing on risk definitions and employing knowledge graphs, the Risk Atlas Nexus bridges gaps, ensuring reliable AI governance while adapting to policy changes.

One of the pivotal challenges in AI that Elizabeth highlights is the fragmented nature of risk definitions even within organisations. The Risk Atlas Nexus project addresses this by mapping and interrelating risk definitions across resources. She also underscores the burden of building reusable, understandable and verifiable AI models. Collaborating with the Adapt Trinity Research Centre, they utilised the “Arrow” knowledge graph ontology to integrate critical risk components, thus enhancing the decision-making process.

Elizabeth reveals that engaging with various AI research assets unveiled invaluable insights. For instance, integrating the Nokia Bell Labs’ classification technique for use-case risks extended the discourse on responsible AI. She observes that transforming static documents into operational, digital frameworks allows organisations to embody governance principles seamlessly within developmental workflows. The aspiration is to streamline innovation without imposing risks, ensuring AI serves both organisational and societal needs responsibly.

Reflecting on challenges like automating risk assessment and ensuring explainable AI decisions, Elizabeth also shared her insights on overcoming unforeseen hurdles. She advocates for integrating curated AI knowledge with automation, enabling organisations to control AI outcomes meaningfully. Her advice to AI innovators is to start small, explore risk assessments and leverage existing guardrails. Utilising granular, curated models tailored with specific risk dimensions can enhance AI projects swiftly and efficiently.

In conclusion, Elizabeth Daly’s work at IBM underscores the significance of marrying innovation with responsibility in the AI domain. Her efforts, exemplified by initiatives like the AI Award-winning Risk Atlas Nexus, pave the path for organisations to navigate the intricate world of ethical AI.

As AI’s influence expands globally, encouraging environments where responsible solutions are prioritised will not only safeguard businesses but also champion ethical technological evolution. Whether you’re an executive embarking on this journey or a developer fine-tuning AI solutions, Elizabeth’s insights offer a roadmap, guiding a future where AI thrives responsibly.