The real magic of AI is that it caused everyone to forget the fundamental principles that govern how we probabilistically learn about the world. Small n / curse of dimensionality / no validation on well curated data? No problem. Visit my C seed startup at AI dot ai 💸 💸

Congratulations Dodgers, Cinderella team for the ages

Wake me up when gen AI can schedule a recurring meeting at a time convenient for both parties and proactively adjust future conflicts.

This is the back gate in my driveway which borders Camp Pendleton, I’ve circled the area where we see artillery drills. Does anyone have info about what time today USMC is shooting missiles over the 5 into this area? Want to make sure us and kids are away from home @mikelevin.org @governor.ca.gov

TIL 36.4 is "significantly larger" than 121.7 thanks google AI please build a giant data center in my backyard for more of this

Our family spends around $75 per month on prescriptions. Did a quick plot to find out how much the 650% is going to save us and hell to the yes the pharmacy is about to *pay us* $412.50 per month

This person sounds funny and cool but I am definitely worried about the “baby in placenta” part

After exits from padres and boston I had no horse in the world series race but this turned me. Seattle all the way

I'm convinced god's table must be more chaotic than this

I think this about yankees and dodgers fans all the time

Your wife is empirically correct.

Obv not the most important thing right now, but I am continuously astounded that Google invests heavily to push incorrect AI slop to the top of search results. In this easily verifiable case, the season series is actually 3-3. Even their own counting of the 2-2 description is wrong!

These ACIP comments from the public are brutal 💀 Kulldork was so clearly unhappy to field them, citing that they are "required by law"

Periodically tuned into ACIP 🗑️ Repeated calls for placebo controlled RCTs. Just shouting the obvious (🤡 ACIP excepted) into the void: we identify life-saving therapies through active control RCTs all the time. Eg effective therapies for cancer exist, we cannot withhold them to study new treatments.

no Apple News, I don't want to read about "Hooray! Stocks hit record highs because there are no jobs meaning interest rates may go down so that billionaires can borrow at lower costs"

Remember like a year ago when we devoted approx 0% of our effort wondering or talking about what dumb shit the president was doing

Facts. AI would have helped me with very little of the data infra work I’ve done over the past 10+ years because most is deployed within highly bespoke setups of large orgs with lots of tech debt

If only there was a study of, I don't know, 2.5 million children spanning 25 years that found no such link

This is what I've been saying since 2023 (image below) "prediction: use of "AI" [...] will come to be broadly associated with cheating, deception, lack of respect for other people, and low quality work that cannot be trusted in important settings"

“My package was accepted to CRAN”

This is extremely cool. Picked it up and made this DAG in like 15 minutes. Obviously pros/cons versus other DAG tools (e.g. low/no code cstructure and R's ggdag, both of which offer d-separation details and more) but if you have rigid aesthetic requirements this is a great tool to have in your kit

Good connection made by @adamjkucharski.bsky.social Take brain cancer: lifetime risk is around .5%. Picture an LLM which detects 100% of cancers and correctly calls "no cancer" 99% of the time when it is absent. Sounds good right? Not really: 67% of the time it says you have cancer when you don't.

Fascinating, accurate, and unfortunately terrifying. Great read.

☹️ frowning + 😒 unamused + 👊 fist-bump =

😹 laughing-cat + 🤪 crazy =

True story: in 2017 there was a serial killer in our neighborhood so we couldn’t go outside for a week, the toddler two doors down shot himself with his parents gun and died, one night we woke our toddlers to hide in the hall from gunshots outside, and a car crashed into our house. Tampa, Florida.

NIH now only funding Jeanome Wide Association Studies because woke

💯 every time I meet with a early career person trying to crack into the data science field, I spend a decent amount of time explaining how to organize their resume so that enshittified AI tools from large org HR won't autoreject them

🤩 @travisgerke.bsky.social and i have been blown away by our #rstats data science intern this summer, give her a follow!

Makary: "We have never talked about environmental exposures that may cause cancer" Sure would be cool if we had, say, an entire branch of the NCI devoted to this.

Why Most Published Statements By Ioannidis Are False

This is just like GWAS, minus the yell if you can imagine that

do u think the uruk-hai had orcIDs

INDIANA JONES AND VIRTUALENV USING THE OSX PYTHON

Same for Sunday-fil-a. Serves nuggets, waffle fries, only open on Sundays. Rainbow themes.

Reflecting on #ser2025 on flight back. Though an epidemiologist for 10+ years it was my first one! Many keynotes/talks were about how to preserve trust in science/reality. I can only imagine previous years were more “here’s what epi did last year, here’s what’s really exciting about this year.”

Thanks #ser2025 friends for hanging out to talk about DAGs!

TIL we can’t know anything without a randomized controlled trial. Up next: RCTs for smoking! RCTs for radiation exposure! Big Cancer just wants you to think these are bad for you, but really without RCTs they’re only “opining”

“Vaccine risks outweigh the benefits”

A masterpiece by @adamserwer.bsky.social. www.theatlantic.com/ideas/archiv...

Reading the Prasad/Makary article driving the FDA restriction of COVID vaccine access: "Third, natural immunity from Covid-19 against severe disease appears robust."

What is a p-value? A p-value is the probability that the alternative hypothesis is true given the observed data, suggesting that if the p-value is low, it actually indicates strong evidence in favor of the null hypothesis, leading you to confidently reject any meaningful findings in your research.

Prostate cancer epidemiologist here 👋 PSA screening not recommended for men over 70. Also de novo metastatic disease (i.e. very advanced at diagnosis) is 5-10% of new prostate cancers, so like 15k-30k in the US annually. Not rare unfortunately

The grant that funded a large part of my doctoral work - a 40 year cohort study which has been running successfully and learning all sorts of things about disease and wellbeing - was just terminated. Surreal.

With nearly all of Harvard Chan School’s direct federal grants terminated, we are relying on philanthropy to power our research and support our educational programs. Every gift, regardless of size, advances our vision of health, dignity, and justice for every human. Support our work: hsph.me/whygive

cStructure is proud to collaborate with @BeeKeeperAI_Inc and DREAM on The COVID Causal Diagram DREAM Challenge. Privacy-preserving compute + collaborative causal modeling -> the future of responsible AI development. Learn more at cstructure.net @travisgerke.bsky.social

"The average American has 3 friends" is why we prefer median to mean. Thanks Kevin Bacon

now considering vacation day for same day new albums from Propagandhi, PUP, Suzzallo, and Pet Symmetry

Overlay student debt -> plumbers are elitist

News bad. Just bike through the desert, clear head