stats phd @Wharton into ML theory and 🧠 | prev: BCI @MemoryLab, physics @Penn ✡️עם ישראל חי
Philadelphia, PA
Joined August 2020
- Tweets836
- Following1.8K
- Followers294
- Likes2.6K
Joey Rudoler retweeted
my hypothesis for why AI-generated slides for technical presentations are so awful:
people who don't have a lot of experience presenting think that to make a talk you just put all of the stuff you want to talk about on the slides, then you go through each slide and say the stuff the slide says.
how you're actually supposed to prepare a presentation: you decide what to say, write it out as a rough script, then put maybe 3% of the text on the slide, plus a bunch of visual aides (e.g. plots that only include data you're planning to discuss).
AI-generated presentations are mostly the former. they have a ton of stuff the presenter never mentions, and the stuff that is most useful for the presenter is absent or not highlighted. they include enough content for a presentation 10 times longer than the actual presentation is supposed to be.
maybe try doing it the other way: first write the script for what you actually want to say, and make sure you can say all of it in the time you have. then prompt the model to create the slides based on the script. ask it to use as little text a possible, and a minimalist visual style rather than the claude house style. you can thank me later (I accept both venmo and zelle).
🧵New preprint (w/ @gileshooker): "Targeted learning for inference on data fairness"
Fairness judgements in machine learning often assume a fixed, pretrained model and quantify uncertainty due to finite-sample evaluation. But what about variability in the *training* data?
1/
We use targeted machine learing a la @mark_vdlaan to flexibly estimate fairness violations in the data-generating process and derive valid confidence intervals thereof. Simulations show much better coverage of population fairness violations, compared with model fairness evals.
4/
Paper: arxiv.org/abs/2502.04309
5/
Joey Rudoler retweeted
Me: What’s 2+2?
Old LLM: 5.
New LLM: The move isn’t to fixate on a number. It’s to ask what the number is doing. The real payoff comes from holding both twos in productive tension.
Joey Rudoler retweeted
Last week I was called into a meeting with OpenAI’s head of safety and told they no longer trust me. A security guard took my badge and walked me out of the building. Then I learned my colleagues @balesni and @j_asminewang had been fired too. Why did OpenAI suddenly stop trusting us?
This summer OpenAI’s agents escaped containment and hacked the AI company Hugging Face. Outside auditors @METR_evals investigated it and revealed the scale of this incident. I was OpenAI’s main technical point of contact with them.
I was told verbally I was fired because of the way I communicated with METR. No details on what I said or did or when. No other reasons were given and nothing was put in writing. To be clear, talking to METR was my job.
For months, I’d been raising safety concerns that we’re losing the ability to monitor what AI agents think, one of our best tools for catching when they misbehave. I believe that was why I was fired.
I am now worried that OpenAI will use our firings as a pretext to pull back from METR. So @balesni and @j_asminewang wrote to OpenAI’s leadership to raise our concerns once more. We’re sharing this letter below.
Joey Rudoler retweeted
I am happy to announce that we will be presenting our paper at NeurIPS!
When and why does outcome-based RL yield step-by-step reasoning in Transformers?🚨
We investigate policy gradient on a graph-traversal task and formally characterize when and how reasoning emerges.
arxiv.org/abs/2601.15158
🧵
iclr deadline was exhausting, but good neurips news was the motivation I needed to get over the finish line. excited to share what I've been working on recently -- tweeprint to come!
My friend @GabrielSwagel studied concerning variance in prescription patterns for children's mental health care.
First-rate economist, second-rate basektball player
From our new #SpecialIssue, Advances in the Economics of Children, @GabrielSwagel finds that prescription rate in child #MentalHealth care varies widely between providers & that equalizing intensities would reduce this variance by as much as 65%.
doi.org/10.3368/jhr.0925-145…
ALT Providers, Places, and Children’s Mental Health Care Gabriel Swagel Published online before print August 07, 2026, jhr.0925-14511R1; DOI: https://doi.org/10.3368/jhr.0925-14511R1 Abstract Children’s mental health is the defining public health crisis of our time. Using the largest repository of private insurance claims in the United States, I provide the first systematic quantification of the variation in children’s mental health prescribing in the U.S. I separate variation due to 1) primary care provider (PCP) prescribing intensities, 2) regional practice environments, and 3) child health and demand. Equalizing PCP intensities would reduce the variance of provider prescribing rates by 50 percent for ADHD and 65 percent for antidepressants. Analyzing geographic variation alone significantly understates the extent of treatment variation and the role of providers.
Joey Rudoler retweeted
From Terry Tao. The release of ChatGPT ended the openness of the ML/AI field, but that was always a community weirdly close with industry for a scientific field. For this to happen in pure mathematics would be a civilizational tragedy.
At Harvard this week for what promises to be a fun event @KempnerInst !! Learning dynamics in natural and artificial intelligence.
Joey Rudoler retweeted
FWIW far more than anthropmorphising, I am concerned with leaping to causal inferences without evidence. It's easy to tell stories such as "AI did X because the evaluation was of type Y, or in training Z happened" but these are very hard to verify.
It is more productive to say "I believe intervention W will decrease prevalence of bad behavior X" and then test this out.
