Inlay

<Profile> by @dansshadow.bsky.social▾

uri: at://smcgrath.phd/app.bsky.actor.profile/self

loading...

Biomedical Informatics PhD • CITRIS Health @UC Berkeley • FAMIA • Focusing on Informatics and AI in medicine • Linfield U. Grad • Missoula MT https://smcgrath.phd

I just donated, and you should consider chipping in if you can too. Graze is an awesome tool, and an even more awesome team. If you enjoy the ML feeds I set up, that’s all thanks to Graze!

Most clinical ML studies estimate sample size. They rely on old rules that can overshoot the mark. Traditional methods predict nearly 30x the data actually needed (19k vs 670 samples). This paper presents a new calculator for tree-based models to ground study design in actual data. #MedSky

The snow gods: How a couple of ski bums built the internet’s best weather app www.technologyreview.com/2026/03/26/1...

The average American worker using AI reports time savings of 6%, or 2.5 hours in a work week. Those are similar to the UK & Netherlands, slightly more than other EU countries. There some early, non-causal, signs that this is translating into gains in productivity growth www.nber.org/papers/w34995

Early AI papers had found that their replies were seen as more empathic than doctors, but a new paper warns of "empty empathy talk". It’s a semantic shell that mimics care without human depth. Outsourcing compassion and empathy risks "moral deskilling" of clinicians.

Concerning: The IRS Wants Smarter Audits. Palantir Could Help Decide Who Gets Flagged www.wired.com/story/docume...

Silicon Valley heavyweights usually don't hesitate to share their opinions, but there’s a deafening silence on administrative overreach lately, @jeffdean.bsky.social is an outlier. While others weigh the risk, he’s been challenging the Pentagon on AI ethics and calling out ICE.

New research in Science shows LLMs endorse user actions 80% of the time, while humans only do it 40% of the time. This AI flattery makes users more certain they're right and less likely to apologize in social conflicts.

Most clinical prediction modeling studies using machine learning models do not estimate sample size. A conceptual model for thinking about power for machine learning models is presented in this…

“We were just trying to create something that made us happy.”

Despite rising concerns about sycophancy—excessive agreement or flattery from artificial intelligence (AI) systems—little is known about its prevalence or consequences. We show that sycophancy is…

npj Digital Medicine - The risk of empty empathy talk and challenges of automated empathy in healthcare communication

Documents show the tax agency is testing a Palantir tool to surface “highest-value” audit and investigation targets from a maze of legacy systems.

AI pioneer Jeff Dean is a rare tech leader who has been publicly criticizing actions by the Trump administration.

Even people who were sceptical of chatbots’ utility fell under the sway of the AI tools’ flattery.

Proud to have co-authored this updated statement on education for pediatric #CHD families as a scientist, mom of a child who has had these surgeries, & board member of Sisters by Heart which supports these families. ❤️ professional.heart.org/en/science-n...

Like what you're seeing? This feed runs on @graze.social, where the feeds you love come to life - pitch in below to support the team and keep the lights on!

Donate to the Graze Social team and keep the feeds flowing!

Education for pediatric heart disease (pedsHD) must be continuous, inclusive, and family-centered—beginning at diagnosis and extending through adolescence and transition to adult care.