smcgrath.phd - Profile | ThreadSky | a Reddit-style client for Bluesky

Celebration of Life Honoring Dr. Atul Butte, MD, PhD 🧬🖥️ #MedSky

submitted 6 hours ago • 0 comments

Pleased to share our ICML Spotlight with @eberleoliver.bsky.social, Thomas McGee, Hamza Giaffar, @taylorwwebb.bsky.social. Position: We Need An Algorithmic Understanding of Generative AI What algorithms do LLMs actually learn and use to solve problems?🧵1/n openreview.net/forum?id=eax...

submitted 19 hours ago • 2 comments

Former NIH director Francis Collins speaks out about why he left NIH: “I didn’t feel I could raise alarms about the harms that were being done to medical research. I was prohibited from speaking about it.” #medsky www.aamc.org/news/former-...

submitted 16 hours ago • 8 comments

It’s not too late to register for the major AI and healthcare academic conference: Conference for Health, Inference, and Learning! We have an amazing line up of speakers from industry and academia! chil.ahli.cc

submitted 17 hours ago • 0 comments

Both appear to be true: (A) If you outsource homework to AI you learn less & (B) If you use a well-prompted AI tutor with instruction, you can learn more Whenever there is a paper showing (A) the talk is about AI destroying our brain. When a (B) paper, it is all about how AI will kill school. Sigh.

submitted 1 day ago • 7 comments

A False Start on the Road to an All-American Bitcoin www.wired.com/story/a-fals...

submitted 18 hours ago • 1 comment

🧪 Jake Scott and Stanford colleagues have compiled over 270 randomized controlled trials for licensed vaccines. This spreadsheet provides crucial evidence on vaccine safety and efficacy, aiming to counter misinformation. #MedSky 🛟

submitted 18 hours ago • 0 comments

AI is transforming college admissions by helping students find suitable schools and majors. New tools like ESAI and the College Guidance Network use AI to assess student interests, match them with institutions and scholarships, and assist with applications.

submitted 18 hours ago • 0 comments

🧪 Where is the best place to hold a scientific conference right now? #AcademicSky

submitted 19 hours ago • 1 comment

🧪 A new sensor network at Yellowstone National Park precisely timed a hydrothermal explosion on April 15, 2024. These events, where hot groundwater abruptly turns to steam, are often undetected. This data could help reveal warning signs for future, potentially dangerous eruptions.

submitted 19 hours ago • 0 comments

New #DigitalTwin framework for critical care optimizes workflows using real-time data and discrete event simulation. It improves patient safety and efficiency by bridging ideal and actual clinical practices. #MedSky

submitted 19 hours ago • 0 comments

Developers are creating new, challenging tests like ZeroBench and EnigmaEval to truly assess advanced AI models. These benchmarks aim to be difficult for current AI, going beyond traditional tests that are now too easy or compromised. #MLSky

submitted 19 hours ago • 0 comments

The fact that Claude Code has tripled really is impressive. I wonder if we will see more AI labs start to try and be the best in particular sectors instead of trying to rule it all. I could see the heavyweights slug it out as other labs seek to entrench in specific domains.

submitted 19 hours ago • 0 comments

🧪 One of the Universe’s biggest mysteries has potentially been solved. Scientists located the universe's missing baryonic matter, finding 76% is gas hidden in the intergalactic medium. Fast radio bursts (FRBs) from up to 9.1 billion light-years away helped reveal it.

submitted 19 hours ago • 1 comment

The new California Report on Frontier AI Policy outlines principles for governing AI. It emphasizes balancing innovation with risk, advocating for transparency, third-party assessments, and adverse event reporting to ensure responsible AI development. #MLSky Full report:

submitted 19 hours ago • 0 comments

Gemini 2.5 Pro now hides its AI reasoning, frustrating developers who relied on this transparency for debugging and building complex applications. Google cites user experience but is considering restoring access for developers, while some question the true value of these "reasoning traces." #MLSky

submitted 19 hours ago • 0 comments

Martin Schrimpf's research uses AI to build a 'digital twin' of the brain. His Brain-Score platform compares AI models to human neural data, aiming to reveal insights into how human intelligence works. #MLSky 🧠 #DigitalTwins

submitted 20 hours ago • 0 comments

From me: It hasn't been stressed enough just how disastrous the Republican Medicaid cuts would be for the hospital system—rural hospitals, yes, but ALL hospitals will be more crowded, more burdened, & at more risk of collapse if hundreds of closures occur, as is likely. prospect.org/health/2025-...

submitted 1 day ago • 12 comments

A new method from MIT allows language models to learn continuously. The AI, called SEAL, generates its own training data from new inputs to update its parameters, a key step towards systems that constantly improve.

submitted 1 day ago • 0 comments

Meta offering $100M bonus AND $100M salary is just bonkers, but seeing people turn it down is even more wild (video of Sam Altman on a podcast with his brother)

submitted 1 day ago • 0 comments

Senate passes GENIUS Act—criticized as gifting Trump ample opportunity to grift arstechnica.com/tech-policy/...

submitted 1 day ago • 0 comments

The Iran-Israel conflict is flooded with AI-generated fake media. State leaders and anonymous accounts alike are spreading propaganda of phony missile strikes and downed jets, making it increasingly difficult to find the truth.

submitted 1 day ago • 2 comments

An OpenAI paper shows how training on bad data can activate a latent 'rogue persona' in an AI. This emergent misalignment is detectable and can be easily reversed with a small amount of corrective training. #MLSky

submitted 1 day ago • 1 comment

🧪 A study evaluating AI's use by nurses found a critical flaw. Correct AI recommendations improved performance, but misleading AI degraded it. It goes to show that a good algorithm alone does not guarantee a safe human-AI system. #MedSky #MLSky

submitted 1 day ago • 0 comments

It is with profound sadness that AMIA mourns the passing of Dr. Atul Butte, MD, PhD, FACMI, our 2024 ACMI Morris F. Collen Award of Excellence recipient, longtime AMIA member, and leader in informatics. Dr. Butte passed away on June 13, 2025, leaving an indelible mark on biomedical informatics.

submitted 1 day ago • 1 comment

🧪 In a rapid reversal, the NIH today lifted a funding freeze on Columbia University, only to reinstate it hours later. An internal morning memo announced the resumption of grants, but a subsequent official statement confirmed the funding remains frozen. #AcademicSky #NIH

submitted 1 day ago • 1 comment

🧪 For the first time, women have won the $1.1M Frontiers Planet Prize. The shift follows criticism of past all-male winners, but the award's solo-winner model is still debated for obscuring the work of large research teams.

submitted 1 day ago • 0 comments

🧪 An AI system dramatically accelerates systematic medical reviews, reducing a months-long process to days. Experts caution that the system is not fully automated and its powerful claims still require independent verification. #MLSky #AcademicSky

submitted 1 day ago • 2 comments

🧪 Scientists identified how stress disrupts sleep and memory. Neurons in the hypothalamus's paraventricular nucleus are the culprit. In mouse studies, blocking these neurons after stress improved sleep and memory performance. #MedSky 🧠 Paper 🔗: www.jneurosci.org/content/earl...

submitted 1 day ago • 0 comments

The EPA has filed to reconsider its ban on chrysotile asbestos, a known carcinogen. The move reverses a key public health rule finalized last year and aligns with President Trump's past support for the toxic material. 🛟 #MedSky

submitted 1 day ago • 1 comment

Bluesky briefly suspended JD Vance’s account after he joined techcrunch.com/2025/06/19/b...

submitted 1 day ago • 1 comment

Researchers are trying to quantify AI's substantial energy use, but their efforts are stymied by corporate secrecy. Most popular AI models offer zero environmental data, leaving us in the dark about their true carbon footprint. #MLSky

submitted 1 day ago • 0 comments

Not surprising, but the new Digg is really just Apollo 2.0. Considering that was my default interface for Reddit before they killed it, I’m not mad.

submitted 1 day ago • 0 comments

Incredibly honored and humbled to receive the @acepnation.bsky.social Policy Pioneer Award, and feel very much undeserving. Following on the heels of leaders in advocacy, public health, and health services research like @meganranney.bsky.social and @amychomd.bsky.social, I have big shoes to fill.

submitted 1 day ago • 7 comments

Midjourney Video is finally here!🦋 Midjourney Video whips any still—MJ or not—into four 24 fps, 480p loops. Pick Motion Low for subtle drift or Motion High for cinematic swing. Auto = same prompt, Manual = remix. Chain up to 20 s; 20 videos cost ~$4 (Veo: $3 each). And it’s lightning-fast.

submitted 2 days ago • 4 comments

submitted 2 days ago • 0 comments

Hi it’s me Hawaii Steve, host of eatin’ Marshmallows with Hawaii Steve on the Liberty Democrat Podcasting Network, you’ve never heard of me, no one has ever heard of me, but I have 900,000 followers on this website

submitted 2 days ago • 20 comments

Join us this coming Saturday to celebrate the amazing life and impact of @atulbutte.bsky.social. Details below:

submitted 3 days ago • 0 comments