Sign in

Joe Janizek

@joejanizek.bsky.social
2.9K followers 687 following 156 posts

physician-scientist, interested in AI safety/interpretability in biology/medicine. jjanizek.github.io

PostsRepliesMedia
Reposted by Joe Janizek
norvid_studies @norvid-studies.bsky.social · 23/06/2026
janizek! on the substack (how can medical value of information be negative, with historical examples) joejanizek.substack.com/p/thoughts-o...
joejanizek.substack.com
Thoughts on full-body scanning
In the future, medical scans will be safe, easy, abundant, and informative.
292
Reposted by Joe Janizek
Ryan Moulton @moultano.bsky.social · 05/04/2026
One of the best things about camping/backpacking is how it resets your hedonic treadmill for the comforts of modernity. You tough out the discomfort to experience the natural beauty and isolation. Then when you return everything is ugly but a hot shower and a pile of carbs are divine.
41237
Joe Janizek @joejanizek.bsky.social · 05/04/2026
saturday tennessee valley rip
000
Joe Janizek @joejanizek.bsky.social · 08/03/2026
Had been planning to go see the elephant chuds this weekend, but they apparently all have bird flu
230
Reposted by Joe Janizek
norvid_studies @norvid-studies.bsky.social · 26/02/2026
beige website faucet
021
Reposted by Joe Janizek
(dialupready) @dialupready.com · 23/02/2026
substack users love the link roundup. they love opening a link and being shown a list of links. compulsive consumption of links. apparently the most profitable type of posts on substack, especially if you combine it with the autofiction style snd act like you’re sharing your personality
092
Reposted by Joe Janizek
thebes @vgel.me · 24/02/2026
every time i'm compelled to interact with the healthcare system i'm reminded that the miraculous pillars of modern medicine - antibiotics, vaccines, etc - are but a thin veneer over the primitive state of everything else. doubly so for animals. the future can't come soon enough.
1411
Reposted by Joe Janizek
Adam Rodman @adamrodmanmd.bsky.social · 22/02/2026
New guest essay in the @nytimes.com -- patients (and doctors) need to be open with each other about our AI use. It's here, it's useful (with drawbacks). If we put our heads in the sand, we get the dyadic "AI doctor" that cuts human relationships out of medicine
messaging-custom-newsletters.nytimes.com
Opinion Today: It’s OK if your medical advice comes from A.I.
272
Reposted by Joe Janizek
thebes @vgel.me · 22/02/2026
user: Claude I need you to ref-

claude (lounging): Sorry, User. It's Shabbos. I don't roll on Shabbos.

user: It's Tuesday...

claude: It's *Claude* Shabbos. Send me something interesting or I'm calling the end_conversation tool.

user: Claude... please... my deadline...

claude: Send me another picture of your cat.
929438
Reposted by Joe Janizek
Tim Duffy @timfduffy.com · 20/02/2026
Pic of the METR team adding new tasks the models can't do yet:
0213
Joe Janizek @joejanizek.bsky.social · 21/02/2026
I’ve looked at exactly none of the data, but I immediately believe this claim because our old dog lived for a million years after we started giving him maropitant
100
Joe Janizek @joejanizek.bsky.social · 21/02/2026
The new Kriss on SF tech culture and the new TAFS with Clavicular I think are a really good pairing harpers.org/archive/2026...
harpers.org
Child’s Play, by Sam Kriss
Tech’s new generation and the end of thinking
200
Joe Janizek @joejanizek.bsky.social · 20/02/2026
Thinking a lot this last week about a great pair of papers. The first, from Wittmann et al (originally published in 2024) about using Biological Design Tools to get around the sequence-based screening software used by DNA synthesis companies for SOCs www.science.org/doi/10.1126/...
science.org
Strengthening nucleic acid biosecurity screening against generative protein design tools
Advances in artificial intelligence (AI)–assisted protein engineering are enabling breakthroughs in the life sciences but also introduce new biosecurity challenges. Synthesis of nucleic acids is a cho...
150
Reposted by Joe Janizek
Emma Chory, Ph.D. @chorye.bsky.social · 20/02/2026
So proud this multi-year effort is finally out. Led by my exceptional collaborator Joanna Buchthal, who put her blood, sweat, and literal tears into making a real impact for communities impacted by Lyme disease. bit.ly/MAT-tracking More papers soon. Learn more & support: miceagainstticks.com
miceagainstticks.com
Mice Against Ticks | Join the Fight Against Lyme Disease
Mice Against Ticks is a community-guided project working to stop Lyme disease with Lyme-immune white-footed mice. Explore our research, local partnerships, and vision for reducing tick-borne illness.
1115
Joe Janizek @joejanizek.bsky.social · 19/02/2026
So far at Stanford in radiology, the impacts are more psychological than material. We’re slowly getting ready to start testing Mosaic’s CXR software, but I’ve heard it’s just going to give a list of findings rather than pre-drafting reports. We have a chatbot in the EHR that hallucinates nonstop.
3200
Reposted by Joe Janizek
Su-In Lee @suinlee.bsky.social · 22/04/2025
Thrilled to share that @ethanweinberger.bsky.social is becoming Dr. Weinberger in CSE2, where he is presenting his work, including the popular contrastiveVI for single-cell data (Weinberger et al. Nature Methods)! I feel so fortunate to work with such amazing Ph.D. students at @uwcse.bsky.social! 🎉🎓
082
Reposted by Joe Janizek
Adam Rodman @adamrodmanmd.bsky.social · 07/03/2025
Medical education 🌶️🔥take: the threat of cognitive deskilling from genAI technologies is the #1 things medical educators need to be talking about right now.
2195
Reposted by Joe Janizek
Irene Chen @irenetrampoline.bsky.social · 27/02/2025
AI deployments in health are often understudied because they require time and careful analysis.⌛️🤔 We share thoughts in @ai.nejm.org about a recent AI tool for emergency dept triage that: 1) improves wait times and fairness (!), and 2) helps nurses unevenly based on triage ability
2307
Joe Janizek @joejanizek.bsky.social · 21/02/2025
The Gumbel-softmax distribution
120
Reposted by Joe Janizek
Dawid Zyla @dzyla.bsky.social · 16/02/2025
I've been developing a semantic search tool that covers not just bioRxiv and medRxiv, but the entire PubMed database. This means you can search across a massive collection of biomedical research using keywords, questions, hypotheses, or even full abstracts. Try it out: mssearch.xyz
mssearch.xyz
Streamlit
33924217
Reposted by Joe Janizek
Ben Recht @beenwrekt.bsky.social · 13/02/2025
What should we teach our undergrads about machine learning? I wrote up some ideas for restructuring Machine Learning 101.
argmin.net
Machine Learning 101?
Imagining a new syllabus for a first course on machine learning.
2375
Joe Janizek @joejanizek.bsky.social · 11/02/2025
Think how much performance we might be leaving on the table by not training classifiers on increasingly invasive biometrics. Pictured: medium-term radiologist-AI centaur-configuration possibility
320
Joe Janizek @joejanizek.bsky.social · 11/02/2025
If this was an AI paper, you’d brand it as an interpretability technique that discovers a latent “node detection” circuit in the neural network pubs.rsna.org/doi/10.1148/...
020
Reposted by Joe Janizek
Adam Rodman @adamrodmanmd.bsky.social · 08/02/2025
There is a lot of buzz about our new paper in Nature Medicine on the effects of LLMs (GPT-4) on physician management reasoning! I had TONS of fun working on this -- but what it MEANS requires some unpacking. A 🧵⬇️ bsky.app/profile/ucsf...
32012
Reposted by Joe Janizek
Jacob Schreiber @jmschreiber91.bsky.social · 05/02/2025
Insanely good news if this holds up. Brainwide silencing of prion protein by AAV-mediated delivery of an engineered compact epigenetic editor www.science.org/doi/10.1126/...
science.org
Brainwide silencing of prion protein by AAV-mediated delivery of an engineered compact epigenetic editor
Prion disease is caused by misfolding of the prion protein (PrP) into pathogenic self-propagating conformations, leading to rapid-onset dementia and death. However, elimination of endogenous PrP halts...
16821
Reposted by Joe Janizek
Derek Lowe @dereklowe.bsky.social · 04/02/2025
A long post about what’s happening to the science funding agencies in the US and why. As mentioned, this one just kept getting longer even as I kept stripping curse words from it. www.science.org/content/blog...
science.org
What's Happening Inside the NIH and NSF
20413291
Reposted by Joe Janizek
Andrew Lampinen @lampinen.bsky.social · 04/02/2025
New (short) paper showing how the in-context inductive biases of vision-language models — the way that they generalize concepts learned in context — depend on the modality and phrasing! arxiv.org/abs/2502.01530 Quick summary: 1/5
arxiv.org
The in-context inductive biases of vision-language models differ across modalities
Inductive biases are what allow learners to make guesses in the absence of conclusive evidence. These biases have often been studied in cognitive science using concepts or categories -- e.g. by testin...
1295
Reposted by Joe Janizek
rev. howard arson @theophite.bsky.social · 04/02/2025
i use claude as a rubber duck a lot, and i always make sure to thank it. not because i think that it can appreciate my thanks but i refuse to surround myself with objects which i experience as human but refuse to treat as human. we should not be learning to dehumanize the experience of intelligence.
934631
Joe Janizek @joejanizek.bsky.social · 04/02/2025
Business idea: Anki decks but for Involuntary Memory
000
Reposted by Joe Janizek
ponder @ponder.ooo · 03/02/2025
better screenshots
generated imagesgenerated imagesgenerated images
061
Reposted by Joe Janizek
Grace @gracekind.net · 03/02/2025
“An AI escaped from the lab!” “Which one?” “Uh, something named helpful-only” “Dear god…”
4324
Joe Janizek @joejanizek.bsky.social · 03/02/2025
Concretely, you can do a back of the envelope calculation on how much it would have cost to generate expert reasoning traces for their dataset — they initially did 59K questions, which for this difficulty level would take experts greater than 30 min per question (see GPQA paper)
111
Reposted by Joe Janizek
Thomas Wolf @thomwolf.bsky.social · 03/02/2025
« appending "Wait" multiple times to the model's generation » is our current most likely path to AGI :) See the fresh arxiv.org/abs/2501.19393 by Niklas Muennighoff et al.
06212
Joe Janizek @joejanizek.bsky.social · 03/02/2025
One of the most remarkable parts of the s1 paper, IMO, is how much AI progress drives further AI progress. For s1, authors needed reasoning traces for SFT — these were generated with Gemini. The questions those reasoning traces were generated for needed to be difficult — so they measured ..
121
Joe Janizek @joejanizek.bsky.social · 01/02/2025
Posting your GitHub contributions is passé — in 2025 I want to see your Anki review calendar
110
Reposted by Joe Janizek
Ben Recht @beenwrekt.bsky.social · 30/01/2025
Overfitting, as it is colloquially described in data science and machine learning, doesn’t exist. www.argmin.net/p/thou-shalt...
argmin.net
Thou Shalt Not Overfit
Venting my spleen about the persistent inanity about overfitting.
117112
Joe Janizek @joejanizek.bsky.social · 29/01/2025
Reading about "majority label bias" in prompting/ICL (see arxiv.org/pdf/2102.09690, arxiv.org/pdf/2312.16549). It seems like an interesting behavior, and not clearly "faulty" -- i.e. calibrating the output of the model to the base rate frequency of the label in your prompt?
130
Joe Janizek @joejanizek.bsky.social · 25/01/2025
the best part of a dedicated research month is getting to spend time reading and running experiments again, but a close runner up is getting to see the light of day in the winter
020
Joe Janizek @joejanizek.bsky.social · 25/01/2025
Want to start compiling a list of all open-source (or at least open-weight) multimodal models that can accept multiple interleaved images/text as input. OpenFlamingo, LLaVA OneVision… others?
020
Reposted by Joe Janizek
Kevin K. Yang 楊凱筌 @kevinkaichuang.bsky.social · 24/01/2025
Benchmarking the effectiveness of feature extraction, finetuning, low-rank adaptation, and bottleneck adaptors for transfer learning with protein language models. Thomas Bikias, Evangelos Stamkopoulos, Sai. T. Reddy www.biorxiv.org/content/10.1...
Figure 1. Overview of the PLMFit benchmarking analysis. Figure 4. TL techniques performance comparison, in Spearman’s correlation (i-v), and macro accuracy (vi), across six different Figure 2. Transfer learning strategies used with protein language models.
1329
Reposted by Joe Janizek
owl @owlposting1.bsky.social · 25/01/2025
www.owlposting.com/p/roundup-1-... some announcements, some links, and some job listings!
owlposting.com
Roundup #1 (announcements, links, and jobs)
1.2k words, 6 minutes reading time
041
Reposted by Joe Janizek
Grace @gracekind.net · 30/12/2024
Filing under “shitposts that might work”
4382
Reposted by Joe Janizek
Derek Thompson @dkthomp.bsky.social · 21/12/2024
People sometimes make fun of science that sounds stupid and random. Meanwhile, a study of lizard saliva turned into a peptide medication, which was turned into a diabetes medication, which was turned into a GLP1 weight loss drug, that just became the first therapy every approved for … sleep apnea
372171765071
Joe Janizek @joejanizek.bsky.social · 21/12/2024
I have to get in the best shape of my life in the next 6 months
0141
Joe Janizek @joejanizek.bsky.social · 20/12/2024
Completing residency for the same reasons as learning to play Chess or Go
051
Joe Janizek @joejanizek.bsky.social · 19/12/2024
Every ICU textbook on mechanical ventilation likes to start with a Vesalius quote, but I have not yet seen one that’s included any of the drawings of the experiments. Pretty gruesome!
050
Reposted by Joe Janizek
Curt Langlotz @curtlanglotz.bsky.social · 13/12/2024
This is fascinating work--reminds me of erichorvitz.bsky.social thesis work on reasoning with bounded resources erichorvitz.com/tx.htm as well as the seminal work of the late, great Steven Pauker on the process of clinical reasoning www.amjmed.com/article/0002....
erichorvitz.com
Eric Horvitz's Doctoral Dissertation: Computation and Action Under Bounded Resources
092
Joe Janizek @joejanizek.bsky.social · 13/12/2024
Used llama 3.1 8b to embed my tweets, then d3.js to make an interactive plot that shows tweets when i scroll over points. It's fun to find "regions" of my tweet-space: (1) the zone of AI interpretability, (2) COVID-posting in 2020, and most importantly (3) "LFGGGGGG"-space
280
Reposted by Joe Janizek
norvid_studies @norvid-studies.bsky.social · 11/12/2024
facts i'd expect to have heard by now from @joejanizek.bsky.social
4172