Responsible AI for General-Purpose Systems: Overview, Challenges, and A Path Forward
Gourab K Patro, Himanshi Agrawal, Himanshu Gharat, Supriya Panigrahi, Nim Sherpa, Vishal Vaddina, Dagnachew Birru
TL;DR
The paper reframes Responsible AI (RAI) in the era of General-Purpose AI (GPAI) by introducing the Degree of Freedom in Output ($DoFo$) to distinguish Type-1 (low $DoFo$) from Type-2 GPAI (high $DoFo$). It argues that high $DoFo$ enables broader capabilities but also amplifies risks across fairness, privacy, explainability, robustness, safety, truthfulness, governance, and sustainability, necessitating a principled design path. To this end, it proposes the C2V2 desiderata—Control, Consistency, Value, Veracity—as a framework for domain-specific RAI requirements, and surveys techniques (e.g., AI alignment, retrieval-augmented generation, neurosymbolic AI) that map to C2V2. The authors advocate a system-design approach that combines compatible methods to operationalize C2V2 in GPAI, illustrated with an enterprise knowledge assistant example. The work highlights governance, measurement, and continuous human-in-the-loop considerations as essential for scalable, responsible GPAI deployments, and calls for ongoing research to extend techniques and tailor them to emerging challenges.
Abstract
Modern general-purpose AI systems made using large language and vision models, are capable of performing a range of tasks like writing text articles, generating and debugging codes, querying databases, and translating from one language to another, which has made them quite popular across industries. However, there are risks like hallucinations, toxicity, and stereotypes in their output that make them untrustworthy. We review various risks and vulnerabilities of modern general-purpose AI along eight widely accepted responsible AI (RAI) principles (fairness, privacy, explainability, robustness, safety, truthfulness, governance, and sustainability) and compare how they are non-existent or less severe and easily mitigable in traditional task-specific counterparts. We argue that this is due to the non-deterministically high Degree of Freedom in output (DoFo) of general-purpose AI (unlike the deterministically constant or low DoFo of traditional task-specific AI systems), and there is a need to rethink our approach to RAI for general-purpose AI. Following this, we derive C2V2 (Control, Consistency, Value, Veracity) desiderata to meet the RAI requirements for future general-purpose AI systems, and discuss how recent efforts in AI alignment, retrieval-augmented generation, reasoning enhancements, etc. fare along one or more of the desiderata. We believe that the goal of developing responsible general-purpose AI can be achieved by formally modeling application- or domain-dependent RAI requirements along C2V2 dimensions, and taking a system design approach to suitably combine various techniques to meet the desiderata.