AI News Brief: today roundup
- Authors are disputing efforts by book publishers and agents to take a larger portion of Anthropic settlement payouts.
- H Company released NeoMME, a single-tower multimodal encoder family that processes text and raw image patches in one Transformer.
- OpenAI Chief Scientist Jakub Pachocki published an essay calling for global safety benchmarks and voluntary slowdowns across AI labs.
- Writers are challenging publishers and talent agents who seek to extract shares from Anthropic legal settlement funds.
- Meta FAIR introduced AI Research Preference Models to rank candidate machine learning experiments prior to consuming expensive GPU compute.
- Amazon announced it will permanently shut down its pioneer crowdsourcing platform, Mechanical Turk, on September 30, 2026.
- OpenAI announced its automated research interns now perform over three times more daily work than human researchers inside the company.
- OpenAI's Jakub Pachocki authored a post highlighting AI alignment challenges and advocating for stronger safety safeguards worldwide.
- Uber co-founder Travis Kalanick signaled that his latest startup, Atoms, may expand into the competitive robotaxi sector.
- OpenAI shared internal data detailing how autonomous coding agents are accelerating experimental velocity and task execution.
Sources
- Authors push back as publishers and agents make claims on Anthropic settlement
- H Company Releases NeoMME: A Family of 260M and 800M Single-Tower Multimodal Encoders That Drop the Vision Tower and Causal Decoder
- In “An Alien Mind,” OpenAI’s Jakub Pachocki Urges Shared Safety Bars
- Authors push back as publishers and agents seek share of Anthropic settlement
- Meta FAIR Introduces AI Research Preference Models (RPMs): Ranking ML Experiments Before Spending GPU Hours
- Amazon Sets Final Closure of Mechanical Turk, Its Crowdsourcing Marketplace
- OpenAI Hits Goal of Building an ‘Automated Research Intern
- An Alien Mind
- Travis Kalanick’s Atoms might be getting into the robotaxi business
- Research acceleration: The view inside OpenAI