March 9, 2024

This Week in Responsible AI, March 9, 2024

This Week in Responsible AI, March 9, 2024

Copyright/data

• A poster’s guide to who’s selling your data to train AI

• What Happens When Your Art Is Used to Train AI

• China court says AI broke copyright law in apparent world first

Labor

• Emotion-tracking AI on the job: Workers fear being watched – and misunderstood

• The job applicants shut out by AI: ‘The interviewer sounded like Siri’

Government

• NIST staffers revolt against expected appointment of ‘effective altruist’ AI researcher to US AI Safety Institute

• Making AI Work for Government: It All Comes Down to Trust

Evaluation

• A safe harbor for AI evaluation and red teaming

• Evaluating LLMs Through a Federated, Scenario-Writing Approach

• Why most AI benchmarks tell us so little

• Open Source Audit Tooling (OAT) Landscape

Fakes

• LexisNexis' Legal AI tool makes up fake legal cases

• Trump supporters target black voters with faked AI images

• FAKE IMAGE FACTORIES: How AI image generators threaten election integrity and democracy

• According to a test, GPT-4 produces more copyrighted text than other popular LLMs.

• Ads on Instagram and Facebook for a deepfake app undressed a picture of 16-year-old Jenna Ortega

Toxicity

• From One to Many: Expanding the Scope of Toxicity Mitigation in Language Models

• Microsoft begins blocking some terms that caused its AI tool to create violent, sexual images

Fairness

• OpenAI's GPT discriminates against certain race- or gender-coded names when used to rank resumes

• Dialect prejudice predicts AI decisions about people's character, employability, and criminality

• Queer = Bad in Automatic Sentiment Analysis

• "Our analysis identifies significant biases in the current state of sign language AI research, including an overfocus on addressing perceived communication barriers, a lack of use of representative datasets, use of annotations lacking linguistic foundations, and development of methods that build on flawed models."

Other

• How Automated Content Moderation Works (Even When It Doesn’t)

• What the digital streaming revolution of the 2000s can teach us about the AI revolution today, according to a former musician

• The Politics of Data Science: Institutionalizing Algorithmic Regimes of Knowledge Production

• Artificial intelligence and illusions of understanding in scientific research

• The Situate AI Guidebook: Co-Designing a Toolkit to Support Multi-Stakeholder Early-stage Deliberations Around Public Sector AI Proposals

Compiled by Leif Hancox-Li