February 20, 2024

This Week in Responsible AI: Feb 20, 2024

This Week in Responsible AI: Feb 20, 2024

Climate

• Measuring AI’s Environmental Impacts Requires Empirical Research and Standards

• How much electricity does AI consume?

• Understanding And Curbing Generative AI’s Energy Consumption

AI Romance

• "Mozilla found the AI girlfriend apps used an average of 2,663 trackers per minute"

• When Love and the Algorithm Don’t Mix

Bias

• What is Bias?

• The Illusion of Artificial Inclusion

• “AI will cure cancer” misunderstands both AI and medicine

• A causal perspective on dataset bias in machine learning for medical imaging

• "gender bias is consistently more prevalent in images than text for both female- and male-typed categories"

• Are you talking to [‘xem’] or [‘x’, ‘em’]? On Tokenization and Addressing Misgendering in LLMs with Pronoun Tokenization Parity

Errors

• L.A. Is Changing How It Scores “Vulnerability” of Unhoused People

• Chicago to stop using controversial gunshot detection technology this year

• Waymo recalls software after two self-driving cars hit the same truck

Fakes

• Why Big Tech’s watermarking plans are some welcome good news See also: Watermarking the future

• Hugging Face's collection of tools for provenance, watermarking and deepfake detection

• Air Canada's chatbot gave a B.C. man the wrong information. Now, the airline has to pay for the mistake

• AI is now making up the deaths of still-alive people and causing their friends to grieve unnecessarily

• What's Wrong With This Rental Listing? The Furniture Is AI.

• Study Featuring AI-Generated Giant Rat Penis Retracted, Journal Apologizes

• 'Deceased leaders are in vogue for political campaigns “because they continue to be more popular than the living ones”'

• How do we feel about AI-resurrected Suharto?

Security

• Staying ahead of threat actors in the age of AI

• Buffer Overflow in Mixture of Experts

Law/Policy

• Preprint on governing AI by governing its compute

• The Problem With Public-Private Partnerships in AI

• Beware misleading AI hype and ‘AI-washing,’ SEC chair warns

• Justice in Forensic Algorithms

• Big tech vows action on 'deceptive' AI in elections

• The Brussels Side-Effect: How the AI Act Can Reduce the Global Reach of EU Policy

• "Health insurance companies cannot use algorithms or artificial intelligence to determine care or deny coverage to members on Medicare Advantage plans, the Centers for Medicare & Medicaid Services (CMS) clarified in a memo sent to all Medicare Advantage insurers."

• FTC Commissioner Alvaro Bedoya on Algorithmic Fairness, Voice Cloning, and the Future

• AI Outputs Are Not Protected Speech

• Framing Online Speech Governance as an Algorithmic Accountability Issue

Other

• Can AI porn be ethical?

• The way AI will be sold to a skeptical public is starting to become clear

• MIT Tech Review interview with Gary Marcus

• A Roadmap to Pluralistic Alignment

• Best Practices for Text Annotation with Large Language Models

• Women in AI: Irene Solaiman, head of global policy at Hugging Face

• Women in AI: Eva Maydell, member of European Parliament and EU AI Act adviser

• Peering into the Black Box of AI Medical Programs

• What would it take to trust an intelligent assistant?

Compiled by Leif Hancox-Li