Cornell researchers studied 1.5 million petitions on Change.org after the platform introduced an AI writing tool. The study, published in Nature Human Behaviour, found that AI assistance made petitions longer and more uniform, shifting language toward words like 'implement' and 'mandate.' However, AI-generated petitions did not see improved success rates, and in some cases performed worse than those written without assistance.
Meanwhile, nurses are pushing back against AI deployment in healthcare. National Nurses United president Jamie Brown says hospitals are using AI to replace nursing staff and cut positions. Montefiore Hospital in New York City laid off 12 nurses and replaced them with AI software. HCA hospitals use a Palantir app called Timpani that nurses say disrupts their scheduling. EPIC health records integrated ChatGPT, raising concerns about patient data safety.
On the regulatory front, President Trump and executives from Google, Anthropic, Meta, OpenAI, Nvidia, and xAI signed a voluntary safety agreement at the White House on September 29. The Joint Commitment on Frontier Responsibilities asks companies to manage AI risks through internal controls and board-level oversight. Trump described the approach as self-policing, with no decentralized verification system or external audit trail in place.
Regulatory disagreements remain wide. Nvidia CEO Jensen Huang dismissed warnings about AI existential threats, while Bill Gates warned that AI in the wrong hands could cause a billion deaths, particularly through bioterrorism. Gates says self-regulation and kill switches are insufficient and wants governments to require monitoring systems within AI technology. The United Nations is exploring global coordination, but no agreement is close.
In other developments, OpenAI's Dots and Meta's Muse are competing as personal AI agents, each offering different approaches to user assistance. A new test called TrustSwap revealed that AI fact-checkers rely heavily on trust labels when making verdicts, and a training method called trust-swap augmentation reduced this bias at 4B parameters. The Department of Health and Human Services launched a five-year SURPASS program to speed up clinical trials using AI and computational models, aiming to compete with China in medical research innovation.
Key Takeaways
- Cornell study of 1.5 million Change.org petitions found AI writing tools made text longer and more uniform but did not improve success rates
- Montefiore Hospital in NYC laid off 12 nurses and replaced them with AI software
- Palantir's Timpani app used at HCA hospitals is criticized by nurses for disrupting scheduling
- Trump and executives from Google, Anthropic, Meta, OpenAI, Nvidia, and xAI signed a voluntary AI safety pact on September 29 with no external verification system
- Nvidia CEO Jensen Huang dismissed warnings about AI existential threats
- Bill Gates warns AI misuse could cause a billion deaths and wants government-mandated monitoring systems
- TrustSwap test shows AI fact-checkers rely too heavily on trust labels when rendering verdicts
- HHS launched five-year SURPASS program using AI to speed up clinical trials and compete with China
- OpenAI's Dots and Meta's Muse are competing as personal AI agents with different approaches
- Nigeria's GenusLab released an AI-powered learning platform teaching coding and design skills to African users
AI petitions on Change.org add words but not results
Cornell researchers studied 1.5 million petitions on Change.org after it introduced an AI writing tool. The tool made petitions longer and more uniform, changing words like 'implement' and 'mandate.' However, AI did not improve success rates and in some cases made things worse. The study was published in Nature Human Behaviour and compared countries where the tool rolled out at different times.
AI petitions on Change.org produce more text, less success
Cornell researchers analyzed 1.5 million petitions on Change.org and found that AI writing assistance changed petition wording. AI made petitions longer and more homogeneous but did not improve outcomes over time. The study used machine learning to compare petitions before and after the AI tool was introduced. Findings raise questions about AI's role in online petitioning and public opinion.
Nurses urge trust in their expertise over unproven AI tools
National Nurses United president Jamie Brown argues that hospitals are using AI to replace nurses and cut staff. Montefiore hospital in New York City laid off 12 nurses and replaced them with AI software. HCA hospitals use a Palantir app called Timpani that nurses say disrupts scheduling. EPIC health records integrated ChatGPT, which nurses worry is unsafe for patient data.
Countries and companies struggle to control AI growth
AI industry leaders and governments disagree on how to regulate AI. Nvidia CEO Jensen Huang dismissed warnings about AI existential threats. Companies like Nvidia, OpenAI, and Anthropic are building self-policing tools to monitor AI agents. President Trump favored self-regulation after meeting tech leaders. The United Nations is exploring global coordination, but no agreement is close.
New test reveals AI fact-checkers rely on trust labels too much
Researchers introduced TrustSwap, a test that swaps source trust labels in AI fact-checkers. They found that confidence and search decisions followed labels as expected, but verdicts often changed based on labels alone. A training method called trust-swap augmentation reduced this problem at 4B parameters. The main open question is whether this works at larger 8B scales.
Bill Gates warns AI could cause a billion deaths if misused
Bill Gates says AI in the wrong hands could cause a billion deaths. He believes bioterrorists could use AI to create dangerous pathogens. Gates says kill switches and self-regulation are not enough to prevent harm. He wants governments to require AI developers to build monitoring systems into their technology.
OpenAI Dots and Meta Muse compete as personal AI agents
OpenAI's Dots and Meta's Muse are both trying to become your go-to AI agent. Dots focuses on conversation and understanding emotions. Muse emphasizes learning and adapting to user needs quickly. A reviewer tested both and found each has strengths depending on what users want from an AI assistant.
AI brings serial killer Aileen Wuornos to life in controversial documentary
Investigation Discovery created a documentary using AI to recreate serial killer Aileen Wuornos. The use of AI to portray real people in documentaries is sparking debate. The technology raises questions about ethics and accuracy in filmmaking. Many actors and critics are voicing concerns about the approach.
HHS launches SURPASS program to speed up clinical trials with AI
The Department of Health and Human Services announced a five-year program called SURPASS. It uses AI and computational models to make clinical trials faster and more flexible. The program will run through the Advanced Research Projects Agency for Health. Officials want to compete with China and other countries in medical research innovation.
Nigeria's GenusLab launches AI platform for coding and design training
GenusLab Technologies in Nigeria released GenusLab Academy, an AI-powered learning platform. It uses quizzes, daily challenges, and projects to teach coding and design skills. The platform targets African users and aims to bridge the gap between education and industry. It is designed to be easy to use at any pace.
Trump and Tech Leaders Sign Voluntary AI Safety Pact
President Donald Trump and executives from Google, Anthropic, Meta, OpenAI, Nvidia, and xAI signed a voluntary safety agreement at the White House on September 29. The pact, called the Joint Commitment on Frontier Responsibilities, asks companies to manage AI risks through internal controls, external assessments, and board-level oversight. Trump described the approach as self-policing, with companies setting their own rules rather than following government mandates. The agreement does not establish any decentralized verification system or blockchain audit trail, keeping safety oversight within the corporations that build the AI models.
Sources
- AI-enhanced petitions on Change.org: More text, less success
- AI-enhanced petitions on Change.org: more text, less success
- Trust nurses, not unproven AI tools
- In Depth: What are countries and companies doing to keep AI from spinning out of control?
- Multi-Channel Mitigation of Source-Trust Shortcuts in Fact-Checking RL Agents
- ‘A billion deaths’: Bill Gates warns about AI in the wrong hands, saying a kill switch and self-regulation won’t be enough
- The Battle to Be Your Personal AI Agent Is Here
- New doc uses AI to bring serial killer Aileen Wuornos to life. The results are controversial
- HHS announces new efforts to speed up, expand clinical trials with AI
- Nigeria’s GenusLab Rolls Out AI Platform for Coding and Design Training
- White House AI Pact Leaves Safety Oversight in Corporate Hands
Comments
Please log in to post a comment.