What If We Reversed Bayes to Test How Robust Our Findings Are to Skepticism?

In this week's issue of Causal Python Weekly:

- Lu Qian's new blog post on Reverse Bayes and evidence

1/

#CausalSky #EconSky #EpiSky #MLSky #StatSky

When a research path becomes “predictable,” is it still research or product development? On NEJM AI Grand Rounds, Dr. Kyunghyun Cho draws a hard line between discovery and scaling. In health care, where should each live, and who should own the risk? Learn more: nejm.ai/ep40

#AI #MedSky #MLSky

Is the effect heterogeneity in your data real?

Testing effect homogeneity and confounding with mixed data

In their new working paper, Ana Paula Armendariz and Martin Huber propose a framework for testing the...

1/

#CausalSky #EconSky #EpiSky #StatSky #MLSky

If you not already have done so, I strongly recommend to read this beautiful, in-depth and hands-on article about how quantization works.

It may be the missing piece for how to understand how modern LLMs work. #MLSky 🧪

Complete Resolution of Persistent Globus Pharyngeus Using Cervical Plexus Block: A Case Report This case describes the successful treatment of persistent globus pharyngeus in a patient refractory to a multitude of traditionally used, mainstay interventions. This report is the first in the Engl...

A paper in The Laryngoscope presents the first successful use of a cervical plexus block for persistent globus pharyngeus, offering a new clinical pathway for chronic throat sensations.
bit.ly/4lOpOOh

#ArtificialIntelligence #BiomedicalInformatics #MedSky #MedAI #AISky #AcademicSky #MLSky

Figure 1. Strengthening the Reporting of Observational Studies in Epidemiology Inclusion Flowchart.

An AI-powered digital front door tool used in Portugal was associated with reduced patient uncertainty, meaningful shifts in care-seeking behavior, and a substantial improvement in the appropriateness of health care utilization. Full study results: nejm.ai/4rENohz

#AI #MedSky #MLSky

DeepFake Detection: Radiology Edition Easier said than done.

Turns out generative AI can sometimes make some very convincing radiographs, and radiologists (and LLMs) struggle to catch all the fakes.
#MLsky #medsky
www.evidencetriage.com/p/deepfake-d...

Beyond convolutions and supervised learning with transformers and representation learning for retinal image analysis Retinal image analysis has enjoyed groundbreaking advances in the last ten years due to seismic improvements in image analysis techniques from the fie…

New Review 💬 Beyond convolutions and supervised learning with transformers and representation learning for retinal image analysis: www.sciencedirect.com/science/arti...

#ArtificialIntelligence #BiomedicalInformatics #MedSky #MedAI #AISky #AcademicSky #MLSky #OpenScience

👀 @sebastianraschka.com wrote an epic blog post accompanying his LLM architecture gallery (see below):

"A Visual Guide to Attention Variants in Modern LLMs"

Highly recommended to check both out -> magazine.sebastianraschka.com/p/visual-att...

#MLsky

Star Warsologies 76: Wearable Tech in Star Wars On this episode of Star Warsologies, we talk to bioengineering professor Benjamin Smarr about technology integrated into the body. There’s tons of wearable tech in Star Wars! Luke has a prost…

My PI Benjamin Smarr breaks down how Darth Vader’s suit streams biometrics through a brain–computer interface.

Now I’m just trying to use wearable data to enhance my Force sensitivity.

skywalkingthroughneverland.com/star-warsolo...

#AcademicSky #MLSky #SciFiTech

A career in #AI doesn’t always start with a plan. Sometimes it starts with a recession, a random lab assignment, and a decision to keep going. On NEJM AI Grand Rounds, Dr. Kyunghyun Cho reminds us that serendipity favors the curious. Listen to the full episode: nejm.ai/ep40

#MedSky #MLSky

betterthanrandom.substack.com/p/turning-te...

#databs #MLSky

Dropbox Reallocated $25M of Their Marketing Budget After a Causal Study.

In this week's issue of Causal Python Weekly:

- Alberto D. Horner reviews a new End-to-End Estimation for Counterfactual Fairness by Yuchen Ma et al.

1/

#CausalSky #EconSky #EpiSky #StatSky #MLSky

ApplyPolygenicScore functionality. Application of ApplyPolygenicScore functions demonstrated in a case study of 1071 individuals from the TCGA database, diagnosed with bladder (BLCA), liver (LIHC), or uterine (UCEC) cancer. A. Recommended workflow when implementing functions provided by ApplyPolygenicScore . A set of preprocessing functions convert polygenic risk score model (PGM) weight files into BED-formatted genomic coordinate files for suggested use in filtering VCF genotype data to desired coordinates. PGM application functions facilitate genetic data importation and weighted sum computation. Visualization functions provide summary information on computed PGSs and phenotype data. Solid arrows indicate required inputs and dotted arrows indicate optional inputs. B. BMI PGS densities, cohort-wide and by categorical phenotypes, computed in the case study cohort and automatically plotted by the create.pgs.density.plot function. C. Correlations of PGSs from (B) with continuous phenotypes automatically plotted by the create.pgs.with.continuous.phenotype.plot function. D. Receiver-operator curves plotted by the analyze.pgs.binary.predictiveness function depicting the performance of the PGSs from (B) to predict obesity status as a sole predictor (top) and with covariates age at diagnosis, sex, and the first 10 principal components of genetic ancestry (bottom). Positive obesity status is defined as BMI ≥ 30. E. From top to bottom: percentile rank of PGSs from (B) for each individual in ascending order, decile and quartile covariate bars, categorical phenotype covariate bars, and continuous phenotype heatmaps. Nicole Zeltser; Rachel M.A. Dang; Rupert Hugh-White; Daniel Knight; Jaron Arbet; Paul C. Boutros

🚀 ApplyPolygenicScore encourages the research community to extend the findings of the statistical genetics niche, facilitating broader use of PGSs and subsequent novel discovery: bit.ly/41jRDnS

#ArtificialIntelligence #BiomedicalInformatics #MLSky #AISky #MedSky #AcademicSky #SciSky #OpenScience

Page 1 of "Response to Spillover Effects in Randomized Evaluations of Translational AI” Read the full letter at ai.nejm.org.

Richard K. Leuchter, MD, William B. Turner, BA, and David Ouyang, MD, respond to a letter about their Perspective “Evaluating Translational AI: A Two-Way Moving Target Problem.” Read the full response: nejm.ai/4cQzu8d

#AI #MedSky #MLSky

Folks believe LLMs will just keep getting better forever, as if transformer-based models have no limits.

Wait until they’re profitable… they’re gonna suck.

Facebook, Google, YouTube, celebrated when they were losing money. Now the robber barons of a new gilded age.

#AcademicSky #MLSky #AIEthics

More Little AI Malpractice Clues Or, at least what 282 people on the internet thought.

Ignore AI in #radiology at your peril, unfortunately.

Interestingly, however, as this (very small) report notes – *where* AI fits in the workflow might matter as far as a perceived "duty of care" to the a patient harmed by a missed diagnosis.
#medsky #mlsky
www.evidencetriage.com/p/more-littl...

Lithium and the Brain–Bone Axis: A Bridge between Osteoporosis and Alzheimer’s Disease - Current Osteoporosis Reports Purpose of Review We evaluate the converging evidence positioning lithium as a systemic modulator of bone and brain health through shared molecular pathways. This review examines the molecular basis, ...

New Review in @springernature.com 💬 Lithium offers a unique paradigm for understanding and potentially treating age-related decline in multiple organ systems at subclinical dosage and concentration.

bit.ly/4siRZqL

#ArtificialIntelligence #BiomedicalInformatics #AISky #MLSky #MedSky #MedAI

AI Grand Rounds Episode 40 AI’s Next Frontier with Dr. Kyunghyun Cho Photo of Dr. Cho

In the latest episode of AI Grand Rounds, Dr. Kyunghyun Cho discusses his wide-ranging career spanning fundamental #AI research, co-founding Prescient Design (acquired by Genentech), and driving applications of AI in health care. Full episode: nejm.ai/ep40

#MedSky #MLSky

It can be difficult to know which models you can actually use with your setup, when you want to work with LLMs on your local setup.

The other day I came across **llmfit**, which has a nice interface that lets you estimate which models can run on your computer: github.com/AlexsJones/l...
#MLsky

5 Causal Inference Ideas for the Age of Vibes

I'm working on a talk for an event this spring, and I want to ask you for your thoughts.

I collected these 5 ideas in response to various queries I got from people across industries and...

1/

#CausalSky #StatSky #EpiSky #EconSky #MLSky

betterthanrandom.substack.com/p/three-hund...

#databs #MLSky

Rule 🥇: temporal leakage is a sin. You don't use the future to forecast the past.
Rule 🥈: don’t forecast what you can measure.
Rule 🥉: counterfactuals don’t get spoilers. If it knows what happened next, it's not a counterfactual. It’s fanfiction.
#forecasting #causalinference #mlsky

Auditor models to suppress poor artificial intelligence predictions can improve human-artificial intelligence collaborative performance AbstractObjective. Healthcare decisions are increasingly made with the assistance of machine learning (ML). ML has been known to have unfairness—inconsiste

New Article: Bridge2AI researchers discover suppression of poor-quality ML predictions through an auditor model shows promise in improving collaborative human-AI performance and fairness

🔗 academic.oup.com/jamia/articl...

#ArtificialIntelligence #BiomedicalInformatics #AISky #MLSky #MedSky #MedAI

Bayesian-calibrated global sensitivity analysis for mathematical models using generative AI Author summary In this research, we introduce a novel approach for conducting global sensitivity analysis in biological models using generative AI. Our method is fully compatible with Bayesian inferen...

Bayesian-calibrated global sensitivity analysis for mathematical models using generative AI

journals.plos.org/ploscompbiol...

#MLSky

Page 1 of the letter "From Psychological Metaphors to Mechanistic Framing in Describing Errors in Large Language Models" Read the full letter at ai.nejm.org.

In a letter, Daniel I. Ro, MD, comments on the Perspective “Faulty Artificial Intelligence, or the Sleep of Reason.” Read the full letter: nejm.ai/4rBU9Br

#AI #MedSky #MLSky

Not every day you get a walkthrough of AdaBoost, an algorithm taught in nearly every intro ML class, from the professor who developed it.

Seeing the mix of rigor and intuition behind something that’s now standard curriculum was pretty priceless.

#MLSky #HDSI #AcademicSky

Robust Reinforcement Learning via Leveraging Historically Optimal Policy With Regulation of Performance Most existing adversarial training methods in reinforcement learning (RL) offer limited robustness and remain vulnerable to novel attacks. To address this limitation, an approach that enhances policy ...

Robust Reinforcement Learning via Leveraging Historically Optimal Policy With Regulation of Performance

ieeexplore.ieee.org/document/114...

#MLSky

LLM Architecture Gallery A gallery that collects architecture figures from The Big LLM Architecture Comparison and related articles, with fact sheets and links back to the original sections.

Fascinating LLM architecture gallery by @sebastianraschka.com -> sebastianraschka.com/llm-architec...

#MLSky

What changes when researchers get foundation models? On NEJM AI Grand Rounds, Seth Hain, senior VP of R&D at Epic (@hey.epic.com), sees broader clinical insight. Learn more: nejm.ai/ep39

#MedSky #AI #MLSky