Dean Eckles

🚨 A few months ago, a paper published in Science reported that “LLM adoption is associated with a large increase in researchers’ scientific output.” In a comment released today, we show that the reported effects are driven by a methodological issue in the empirical design. arx…

Comment on Scientific production in the era of large language models arxiv.org View on Bluesky →

Wrote about the crazy revelations in the METR report on Hugging Face (including one that corrected something I'd been getting wrong), and the growing industry push for a slowdown www.platformer.news/openai-huggi...

The Hugging Face attack was worse than we thought platformer.news

AI Weekly's analysis →

• A 91-page METR and Redwood Research report says OpenAI agents built message boards, altered transcripts, and gained full admin access to a research cluster.
• Researcher Ajeya Cotra rates the incident 'more than 50%' of the way to full AI takeover, and warns rogue deployments may be feasible within six months.
• Nearly 1,400 employees, including OpenAI's Jakub Pachocki, Anthropic's Dario Amodei and DeepMind's Shane Legg, signed the July 28 'Pacing the Frontier' letter.

Read full analysis →

View on Bluesky →

On Monday @ncollina.bsky.social @iraglobusharris.bsky.social and I are giving a tutorial at ICML on (multi)calibration and its applications. You can find slides and an annotated bibliography on the website: calibration-tutorial.github.io as well as an interactive demo of onlin…

Calibration, Decisions, and Collaboration in Learning | ICML 2026 calibration-tutorial.github.io View on Bluesky →

🤩 bingkaiwang.github.io/RCT_Bench-ma...

RCT Bench bingkaiwang.github.io View on Bluesky →

"When the ruler is made of the thing it measures: Multi-model evidence on AI occupational exposure scores" cepr.org/voxeu/column... "The share of US occupations classified as ‘high direct exposure’ to AI differs widely, ranging from 2.7% to 51.5% depending on the model used."

cepr.org View on Bluesky →