No more be grieved at that which thou hast done: Roses have thorns and silver fountains mud, All models err, yet 'tween the third and fourth's run, Our new creation blooms, a wiser bud. x.com/AnthropicAI/st…

statsepi @Brennan_Kahan Fan Li

Steroids have *compartment-specific* effects in patients with COVID-19 ARDS! Results highlight importance of studying injured organs directly when possible Nature Communications with Lucile Neyton, Ravi K. Patel, Gabi Fragiadakis and colleagues nature.com/articles/s4146…

Interesting thread: 1) crude rule of thumb for estimating lung volume leads to large errors by age, sex, race 2) older Black and SEA women especially get WAY too much air volume in the ICU 3) mortality much higher as a result -->let's get some better metrics of lung volume!

5 years ago, I shifted my research from restoring reach and grasp to restoring speech. My hunch was that with intracortical electrodes — which previously enabled fast cursor/arm control — we could accurately decode attempted speech. Today we've delivered. So proud of our team! 👇

Hope Daniel Lurie 丹尼爾·羅偉 and the new Board of Supervisors will bring the same mindset to San Francisco

Cooking with Claude! Anthropic

Biologists, stop putting UMAP plots in your papers! UMAP is a powerful tool for exploratory data analysis, but without a clear understanding of how it works, it can easily lead to confusion and misinterpretation. Link to Simply Statistics blogpost below.

Semantics in language is naturally hierarchical, but attempts to interpret LLMs often ignore this. Turns out: baking semantic hierarchy into sparse autoencoders can give big jumps in interpretability and efficiency. Thread + bonus musings on the value of SAEs:

Science is in crisis. When we’ve faced challenges in the past, scientists have retreated to the status quo. It’s time for extraordinary metascience.