November 4, 2023

This (past two) weeks in Responsible AI: Nov 4, 2023

This (past two) weeks in Responsible AI: Nov 4, 2023

Apologies for the delay and jumbo edition!

New Research

• Digital resignation and the datafied welfare state

• "One-size-fits-all"? Observations and Expectations of NLG Systems Across Identity-Related Language Features

• Auditing Fairness by Betting

• Bridging Systems: Open Problems for Countering Destructive Divisiveness Across Ranking, Recommenders, and Governance

• Dual Governance: The intersection of centralized regulation and crowdsourced safety mechanisms for Generative AI

• Data Refusal From Below: A Framework for Understanding, Evaluating, and Envisioning Refusal as Design

• Seamful XAI: Operationalizing Seamful Design in Explainable AI

• How Robust is Google's Bard to Adversarial Image Attacks?

• Composite Backdoor Attacks Against Large Language Models

• "generative AI outputs (at least the speech-like ones) are likely entitled to First Amendment protections"

• Automated Tax Planning: Who’s Liable When AI Gets It Wrong?

• People Perceive Algorithmic Assessments as Less Fair and Trustworthy Than Identical Human Assessments

• The Authoritarian Data Problem

• Reimagining Democracy for AI

• Building the Epistemic Community of AI Safety

• Improving Diversity of Demographic Representation in Large Language Models via Collective-Critiques and Self-Voting

• Making AI Less "Thirsty": Uncovering and Addressing the Secret Water Footprint of AI Models

• Datasheets for Digital Cultural Heritage Datasets

• "both humans and preference models (PMs) prefer convincingly-written sycophantic responses over correct ones"

• The Sentiment Problem: A Critical Survey towards Deconstructing Sentiment Analysis

• RoCourseNet: Distributionally Robust Training of a Prediction Aware Recourse Model

Talks

• Arvind Narayanan | Resistance or harm reduction?

• Sasha Luccioni: AI is dangerous, but not for the reasons you think

Policy

• Supporting Open Source and Open Science in the EU AI Act

• "any kind of effective AI regulation would need to regulate personal computing"

• Commentary on the White House's executive order on AI: Biden seeks to rein in AI, Tate Ryan-Mosley on the executive order's emphasis on watermarking and content authentication, Three things to know about the White House’s executive order on AI

• UK government report (PDF): Safety and Security Risks of Generative Artificial Intelligence to 2025

• The fingerprints on a letter to Congress about AI

• Belt and road forum: China launches AI framework, urging equal rights and opportunities for all nations

• Inside ICE’s Database for Finding ‘Derogatory’ Online Speech

• "Companies including Meta, Google DeepMind and OpenAI have agreed to allow regulators to test their latest AI products before releasing them to the public"

• NSF invests $10.9M in the development of safe artificial intelligence technologies

• EDPB issues binding decision banning Meta's targeted advertising practices

• How AI Can Be Regulated Like Nuclear Energy

• Joy Buolamwini: “We’re giving AI companies a free pass”

• Funding is available from the NEH for Humanities Perspectives on Artificial Intelligence

• The Case for Including the Global South in AI Governance Discussions

• AI RED-TEAMING IS NOT A ONE-STOP SOLUTION TO AI HARMS: RECOMMENDATIONS FOR USING RED-TEAMING FOR AI ACCOUNTABILITY

Fairness

• "the algorithm's ability to identify the risk of depression varied by up to 15% between different groups."

• Can AI Be Fair?

Transparency

• A method to interpret AI might not be so interpretable after all

• How the Foundation Model Transparency Index Distorts Transparency

• AI Cameras Took Over One Small American Town. Now They're Everywhere

• Could Cruise be the Theranos of AI?

Tools

• Data Provenance Explorer

• How advances in AI can make content moderation harder — and easier

• This new data poisoning tool lets artists fight back against generative AI

Privacy

• Privacy in the Age of AI

Generative AI

• AI Modi started as a joke, but it could win him votes

• Hackers Are Weaponizing AI to Improve a Favorite Attack

• An AI-generated poll speculating on the cause of a woman's death appeared next to a Guardian article on the same death. In response, Microsoft has "deactivated Microsoft-generated polls for all news articles".

• Propaganda or Science: Open Source AI and Bioterrorism Risk

• AI doomsday warnings a distraction from the danger it already poses, warns expert

• Riley Reid on AI: ‘I Don’t Want Porn to Get Left Behind’

Compiled by Leif Hancox-Li