College of Engineering Unit:
Artificial Intelligence is everywhere; it has helped speed up many aspects of the modern working environment, but as with everything there are problems that occur when human error is allowed in. Partnership on AI has developed an online database to collect news articles about such incidents with artificial intelligence, in order to direct the practice of equitable, ethical AI design.These articles range from AI controlled cameras, to Military weapons, and everything in between as artificial intelligence becomes more common in every industry. A database like this is important to ensure developers have easy access to this history of AI development in order to not repeat these sometimes harmful mistakes.
Our team is working on the Natural Language Processing for AI Incident Resolution project to extend the functionality of a pre-existing database with new text-to-incident utilities. The AI Incident Database (AIID) is a collection of artificial intelligence system mishaps and accidents, allowing for new incidents to be cataloged and linked to other similar accidents for collective reading. During our initial meeting, our stakeholder Sean McGregor listed several problems with the existing system. He hopes to add functions that analyze the text of articles and categorizes them as incidents based on their content. He hopes that we can then add functionality in order for the system to correlate and recommend articles categorized as similar.