About the symposium

Artificial intelligence (AI) has emerged as a promising technology with the potential to transform domains ranging from healthcare to defense. However, the use of AI also poses significant risks and challenges that need to be addressed. To mitigate these risks, it is critical to ensure that AI systems are designed to operate safely, effectively, and reliably (e.g., in accordance with the Department of Defense’s ethical principles for AI). Additionally, it is crucial to perform comprehensive test and evaluation (T&E) of these systems at multiple stages before deployment to identify and address any potential vulnerabilities or flaws.

This symposium will bring together three groups of adjacent, relevant stakeholders from industry, academia, and government to address the use of AI in different domains. The symposium will consist of a session on assured AI, a session on T&E to verify and validate assured AI, and a session on the need for and use of assured AI to address emerging technologies such as generative AI. Each session will fill one day of the symposium, with the last session given the half day. All sessions will have specific objectives to drive forward the field of AI assurance and enable safe and effective adoption of AI in different settings. This will include aims to define key technical terms and to identify the existing and future methods and best practices for ensuring the safety and effectiveness of assured AI. By establishing a common understanding of these concepts, using a combination of keynote talks and interactive panel discussions, the workshop will help to promote successful adoption and integration of assured AI principles to enable development of more robust and reliable systems.

The proposed schedule is designed to promote meaningful dialogue and exchange of ideas between industry, academia, and government. One concrete outcome of the symposium is a planned collaboration on the part of the organizers and any other interested participants in writing a survey or opinion paper on the critical issues covered during the symposium. Potential topics of interest include, but are not limited to:

1. 1. Methods and tools for assurance of AI

   2. AI for assurance in high-risk applications

   3. Interoperability/composability of AI T&E tools

   4. System/infrastructure perspectives on AI T&E

   5. Assurance of emerging technologies (e.g. large foundation models)

   6. AI assurance implications for development of policy and regulatory strategies

   7. Technology transition to address, facilitate, and accelerate current government/industry needs for AI assurance

The call for papers can be found here.

Reach out to brian.hu@kitware.com or akapusta@mitre.org if you have any questions about the symposium.

AAAI Fall Symposium Series Website: https://aaai.org/conference/fall-symposia/aaai-2023-fall-symposium-series/ 

Organizing Committee