Sonita Lontoh, Directors & Boards; Ethical and Responsible AI: A Governance Framework for Boards
"Boards must understand what gen AI is being used for and its potential business value supercharging both efficiencies and growth. They must also recognize the risks that gen AI may present. As we have already seen, these risks may include data inaccuracy, bias, privacy issues and security. To address some of these risks, boards and companies should ensure that their organizations' data and security protocols are AI-ready. Several criteria must be met:
- Data must be ethically governed. Companies' data must align with their organization's guiding principles. The different groups inside the organization must also be aligned on the outcome objectives, responsibilities, risks and opportunities around the company's data and analytics.
- Data must be secure. Companies must protect their data to ensure that intruders don't get access to it and that their data doesn't go into someone else's training model.
- Data must be free of bias to the greatest extent possible. Companies should gather data from diverse sources, not from a narrow set of people of the same age, gender, race or backgrounds. Additionally, companies must ensure that their algorithms do not inadvertently perpetuate bias.
- AI-ready data must mirror real-world conditions. For example, robots in a warehouse need more than data; they also need to be taught the laws of physics so they can move around safely.
- AI-ready data must be accurate. In some cases, companies may need people to double-check data for inaccuracy.
It's important to understand that all these attributes build on one another. The more ethically governed, secure, free of bias and enriched a company's data is, the more accurate its AI outcomes will be."
No comments:
Post a Comment