The Artificial Intelligence Incident Database (AIID) is an innovative and valuable resource dedicated to indexing and understanding incidents involving the deployment of artificial intelligence (AI) systems. Similar to databases used in aviation and computer security, AIID aims to learn from past incidents to prevent or mitigate future harm. It operates on the premise that AI will only truly benefit humanity and society if we collectively record its failings and learn from them.

AIID offers various ways to discover and examine incidents. The features include a spatial view, a table view, a list of involved entities, taxonomies, word counts, and the option to submit incident reports. A leaderboard showcasing the contributions from various individuals and entities is also available. For those interested in keeping up with the latest incidents, there’s an option to subscribe to a newsletter.

Interestingly, AIID also provides data on the most frequently involved entities in AI-related incidents. For instance, Facebook has been involved in 48 incidents, allegedly causing harm to 86 entities. Tesla has been implicated in 38 incidents, reportedly causing harm to 47 entities, and Google is associated with 30 incidents, allegedly harming 42 entities.

Contributions to the database come from various sources. Anonymously submitted reports lead the pack, followed by contributions from Daniel Atherton of BNH.ai and Khoa Lam. The database encourages more individuals and organizations to submit incident reports to expand the resource and increase the opportunities for learning.

The AIID is a project of the Responsible AI Collaborative, an organization dedicated to advancing the database. The governance structure of the Collaborative revolves around participation in its impact programming.

The database has gained recognition in various media outlets, including the PAI Blog, Vice News, Venture Beat, Wired, arXiv, and Newsweek. This further attests to its growing influence in shaping the narrative around the real-world impacts of AI systems.

In summary, the AIID is an invaluable tool for anyone seeking to understand the real-world implications of AI systems. By collecting, categorizing, and learning from past incidents, the AIID provides a unique resource for policymakers, researchers, and the public to help shape a safer and more responsible future for AI.