February 11, 2024

This Week in Responsible AI: Feb 11, 2024

This Week in Responsible AI: Feb 11, 2024

Harms and mitigations

• Check out "the world's most responsible AI model"

• "By prompting a reverse LM with a problematic string, we can sample prefixes that are likely to precede the problematic suffix"

• Red-Teaming for Generative AI: Silver Bullet or Security Theater?

• The persuasive effects of political microtargeting in the age of generative artificial intelligence

• Social media algorithms ‘amplifying misogynistic content’

• Waymo driverless car strikes bicyclist in San Francisco, causes minor injuries

• Beyond Behaviorist Representational Harms: A Plan for Measurement and Mitigation

• "participants engaged in more biased information querying with LLM-powered conversational search, and an opinionated LLM reinforcing their views exacerbated this bias"

Data

• Implications of Common Crawl’s popularity for trustworthy AI

• Best Practices for Text Annotation with Large Language Models

Transparency

• Faithfulness vs. Plausibility: On the (Un)Reliability of Explanations from Large Language Models

• Rethinking Interpretability in the Era of Large Language Models

• The AI tools that might stop you getting hired

• Feature Importance for Uncertainty Quantification In Agent-Based Modeling

Law/Policy

• Biden-Harris Administration Announces First-Ever Consortium Dedicated to AI Safety

• Stratospheric safety standards: How aviation could steer regulation of AI in health

• What can we learn from the FDA model for AI regulation?

• The EU AI Act has been approved

• EU AI Act will fail commitment to ban biometric mass surveillance

• In India, an algorithm declares them dead; they have to prove they’re alive

• Los Angeles Becomes First U.S. City to Outlaw Digital Discrimination

Fakes

• Finance worker pays out $25 million after video call with deepfake ‘chief financial officer’

• 'The practice of re-dressing proudly sexual women in AI-generated, modest clothing and putting them into fake scenes of domestic child-rearing is emerging alongside the growing “tradwife” movement, which envisions the popular resurgence of 1950s-style gender roles.'

• Addressing Computer-Generated Child Sex Abuse Imagery: Legal Framework and Policy Implications

• Iran-backed hackers interrupt UAE TV streaming services with deepfake news

• AI-generated voices in robocalls can deceive voters. The FCC just made them illegal

• Images generated with ChatGPT will now contain metadata indicating that provenance---but that metadata can still be removed

• Meta Oversight Board supports leaving up manipulated Joe Biden video but questions existing policies

Other

• Ethics Review at ICML

• When it Comes to Facial Recognition, There is No Such Thing as a Magic Number

• Algorithms of Resistance: The Everyday Fight against Platform Power (Open access book)

• Schools install toilet sensors that ‘actively listen’ to pupils

• Inside OpenAI’s Plan to Make AI More ‘Democratic’

Compiled by Leif Hancox-Li