After a long hiatus, we return with one of our most interesting episodes yet! Arham and I were delighted to chat with David Stutz, Ph.D., whose work at Google Deepmind has been absolutely trailblazing in the field of medical A.I.
Unlike most A.I. conversations where I walk away with existential dread about going to medical school in the age of LLMs, I left incredibly inspired about the new heights healthcare may reach when A.I. is leveraged intelligently and thoughtfully. Our conversation spans from the process of building smart models to the insights David's research has found about how to best implement them. We hope you enjoy it!
If you want to read more about David's work, here's some of their recent preprints:
https://arxiv.org/abs/2507.15743
https://arxiv.org/abs/2505.04653
If you have any feedback or ideas for future guests, please send us a message at baselineobservations@gmail.com.
After enough "DOCTOR'S DON'T WANT YOU TO KNOW THIS" headlines, it's about time we flip the script. Let's start here: red light therapy does not do much to fix hairlines. Trust me because...well, it's obvious.
It is impossible to practice (let alone pay for) everything you see online that’s claimed to be good for you, and we’re not even sure that it would be a good idea to try. Separating out the useful from the useless is crucial - in this episode, Arham and I talk about some of the things we’re thinking about when we try to do that. If you’re enjoying the podcast, please give us a rating! It goes a long way in helping us grow larger. We always welcome any and all feedback - leave a comment or send us an email at baselineobservations@gmail.com.
Applying medical research at the bedside can be difficult. After all, you don't inspire the most confidence in patients when you tell them their treatments are backed by p values of 0.049.
Dr. Anil Makam schools me here on how medical evidence should actually be applied in practice. We cover how doctors can make sure that their everyday decisions are based on the best evidence, when clinical context overrides published literature, and how new technologies and data sources can provide novel value to the medical evidence base.
You can find Dr. Makam at @AnilMakam on X. If you have any feedback for the show, please leave a comment or reach out at rjhawar227@gmail.com.
As a general rule, lowering your standards is almost never a good idea. But when you're giving relationship advice to a humongous national corporation that dictates nearly everything in American healthcare, maybe the general rules don't apply...
We tackle an increasingly pressing issue here - how much of medicine should be standardized and centrally regulated? It's easy to say both that guidelines are important and that individual context should remain supreme, but how do we draw the line on what's apples and what's oranges here? Beneath this all - is competition in healthcare actually good? This conversation touches on all of the above, and truly was one of the episodes where both Arham and I decisively changed our minds.
If you have any feedback, please leave a comment or reach out at rjhawar227@gmail.com.
All premeds take o-chem, some premeds take p-chem, but should any take...no-chem???
As medicine has become more evidence-based and biologically sophisticated, the job description of a doctor has changed dramatically over the past half-century. Good clinical practice involves a lot more of inspecting p-values and effect sizes than it ever did before, and there's an argument to be made that the time has come for hard sciences to be kicked out of the pre-med equation and more rigorous statistical training to be added in. While discussing this, we confront the larger question of whether understanding first principles will even be relevant as the world continues to progress.
All feedback is always appreciated! Please leave a comment here, or reach out at rjhawar227@gmail.com.
Physics haters be warned…
Dr. Travis Zack is a practicing oncologist and assistant professor at UCSF, where he heads a research lab working on implementing A.I. tools in the electronic health record. He is also head of A.I. at UCSF's Comprehensive Cancer Center, and a clinical consultant at OpenEvidence, one of the world's fastest growing and most innovative A.I. powered clinical decision support startups.
Dr. Zack talked to us about how his unique background has proved extremely useful in his work, and gave us his thoughts on the best paths forward for leveraging medical A.I. at scale for the next 10 years.
We appreciate any and all feedback! Please reach out if you'd be interested in being the next guest on the podcast, or if you know someone who might. Thanks for listening!
There may be only one thing that lasts longer than ketchup: clinical trials from the 1990's. But at a certain point, even these must expire. How do we know when it's time to put old knowledge to new tests, and is there a way to do this ethically and efficiently? Arham and I tackle this as we grapple with an hilariously unsettling truth:
All roads may lead to Rome, but all healthcare startup ideas lead to "What if you made an insurance company, but instead of being horrible, you just did this"
As AI/ML methods progress at an incredible pace, what advancements in medical therapies can we look forward to? To explore this, Arham and I discuss some of the fundamentals behind how new drugs are discovered and some of the seminal advancements that machine learning has brought to this field. We then predict how this will impact both clinical medicine and the economics of the pharmaceutical industry.
Many widespread practices in medicine have never been directly shown to make people live longer, happier, or healthier. Arham and I discuss why this is the case. Is this a fair standard, or is it impossibly difficult to reach? I present the stances held by a few different camps in medicine, and Arham responds with a statistician's perspective on them.
With all the media recently about United Healthcare and its widespread claim denials, we discussed why health insurers deny care (HINT: it's not because it saves them money). We also discussed why current regulations in the health insurance market actually incentivize health insurers to increase healthcare spending, not decrease it.
We would love to hear any and all feedback in the comments.
We discuss whether medical training should have a more dedicated emphasis on teaching statistics, and broadly touch on the implications of statistics in medical decision making.