-
The Promise and Peril of Artificial Intelligence -- Violet Teaming Offers a Balanced Path Forward
Authors:
Alexander J. Titus,
Adam H. Russell
Abstract:
Artificial intelligence (AI) promises immense benefits across sectors, yet also poses risks from dual-use potentials, biases, and unintended behaviors. This paper reviews emerging issues with opaque and uncontrollable AI systems and proposes an integrative framework called violet teaming to develop reliable and responsible AI. Violet teaming combines adversarial vulnerability probing (red teaming)…
▽ More
Artificial intelligence (AI) promises immense benefits across sectors, yet also poses risks from dual-use potentials, biases, and unintended behaviors. This paper reviews emerging issues with opaque and uncontrollable AI systems and proposes an integrative framework called violet teaming to develop reliable and responsible AI. Violet teaming combines adversarial vulnerability probing (red teaming) with solutions for safety and security (blue teaming) while prioritizing ethics and social benefit. It emerged from AI safety research to manage risks proactively by design. The paper traces the evolution of red, blue, and purple teaming toward violet teaming, and then discusses applying violet techniques to address biosecurity risks of AI in biotechnology. Additional sections review key perspectives across law, ethics, cybersecurity, macrostrategy, and industry best practices essential for operationalizing responsible AI through holistic technical and social considerations. Violet teaming provides both philosophy and method for steering AI trajectories toward societal good. With conscience and wisdom, the extraordinary capabilities of AI can enrich humanity. But without adequate precaution, the risks could prove catastrophic. Violet teaming aims to empower moral technology for the common welfare.
△ Less
Submitted 27 August, 2023;
originally announced August 2023.
-
Approximating Vertex Cover using Structural Rounding
Authors:
Brian Lavallee,
Hayley Russell,
Blair D. Sullivan,
Andrew van der Poel
Abstract:
In this work, we provide the first practical evaluation of the structural rounding framework for approximation algorithms. Structural rounding works by first editing to a well-structured class, efficiently solving the edited instance, and "lifting" the partial solution to recover an approximation on the input. We focus on the well-studied Vertex Cover problem, and edit to the class of bipartite gr…
▽ More
In this work, we provide the first practical evaluation of the structural rounding framework for approximation algorithms. Structural rounding works by first editing to a well-structured class, efficiently solving the edited instance, and "lifting" the partial solution to recover an approximation on the input. We focus on the well-studied Vertex Cover problem, and edit to the class of bipartite graphs (where Vertex Cover has an exact polynomial time algorithm). In addition to the naive lifting strategy for Vertex Cover described by Demaine et al., we introduce a suite of new lifting strategies and measure their effectiveness on a large corpus of synthetic graphs. We find that in this setting, structural rounding significantly outperforms standard 2-approximations. Further, simpler lifting strategies are extremely competitive with the more sophisticated approaches. The implementations are available as an open-source Python package, and all experiments are replicable.
△ Less
Submitted 31 October, 2019; v1 submitted 10 September, 2019;
originally announced September 2019.
-
Beyond Confidence Regions: Tight Bayesian Ambiguity Sets for Robust MDPs
Authors:
Marek Petrik,
Reazul Hasan Russell
Abstract:
Robust MDPs (RMDPs) can be used to compute policies with provable worst-case guarantees in reinforcement learning. The quality and robustness of an RMDP solution are determined by the ambiguity set---the set of plausible transition probabilities---which is usually constructed as a multi-dimensional confidence region. Existing methods construct ambiguity sets as confidence regions using concentrati…
▽ More
Robust MDPs (RMDPs) can be used to compute policies with provable worst-case guarantees in reinforcement learning. The quality and robustness of an RMDP solution are determined by the ambiguity set---the set of plausible transition probabilities---which is usually constructed as a multi-dimensional confidence region. Existing methods construct ambiguity sets as confidence regions using concentration inequalities which leads to overly conservative solutions. This paper proposes a new paradigm that can achieve better solutions with the same robustness guarantees without using confidence regions as ambiguity sets. To incorporate prior knowledge, our algorithms optimize the size and position of ambiguity sets using Bayesian inference. Our theoretical analysis shows the safety of the proposed method, and the empirical results demonstrate its practical promise.
△ Less
Submitted 20 February, 2019;
originally announced February 2019.
