Ari B Friedman

Ari B Friedman

@abfriedman.com

This is my personal account. Faculty at UPenn. Emergency doc, health economist, geriatrics and privacy researcher. Views are my own. RT does not imply endorsement.

7K followers554 following578 posts

Pinned

If you're not following Phillips P O'Brien you should. His background as a preeminent historian of strategy and economics of WWII, his moral clarity, and his realism have made him right about at startling number of things since the start of the most recent invasion of Ukraine.

Phillips OBrien@phillipspobrien.bsky.social

Free weekend update is out. Russia has a strategy for this winter--massive war crimes to make Ukraine unlivable. We need to speak the truth on that. One reason they need to do this is that they are struggling in other parts of the war--see Vivaldi. phillipspobrien.substack.com/p/weekend-up...

Ari B Friedman reposted

Free weekend update is out. Russia has a strategy for this winter--massive war crimes to make Ukraine unlivable. We need to speak the truth on that. One reason they need to do this is that they are struggling in other parts of the war--see Vivaldi. phillipspobrien.substack.com/p/weekend-up...

phillipspobrien.substack.comWeekend Update #205: Russia’s Strategy In Two Words: “War Crimes”Operation Vivaldi, Even More Robots; The Trump-Putin Quid Pro Quo

I'm truly shocked by all the claims that LLMs are intelligences or are brilliant at coding. They're useful tools, in some circumstances. But if you gave one a job exam for an entry-level programming job it would fail. LLMs try a strategy, fail, try again, fix the error, try again.

That's useful! And they can do some remarkable things. But the cost of un-entrustable code is very high. Here's an obvious mistake one just made on some basic data cleaning. Can you spot the error?

Some heuristics that have helped me think about LLMs clearly: 1) The גּוֹלֶם (Golem) myth: words have a power of their own. Write them on a clay figurine and it will imitate a human in certain ways.

Darren Dahly@statsepi.bsky.social

Lots of academics seem confused about the difference between a heuristic and a model.

View full thread

6) Just a predictive model that can interpolate well but can be quite sensitive to order and other minor details. Ask yourself, "Is the conversation I'm trying to predict likely to be anywhere in the training data?"

Apropos of sensitive to minor details, I'm playing with coding with one, and I had previously told it that the ZIP files were redundant with the CSV files in the data folder. Then I asked it to make a patient ZIP code variable and didn't capitalize ZIP. It next asked if it could unzip the files.

Some heuristics that have helped me think about LLMs clearly: 1) The גּוֹלֶם (Golem) myth: words have a power of their own. Write them on a clay figurine and it will imitate a human in certain ways.

Darren Dahly@statsepi.bsky.social

Lots of academics seem confused about the difference between a heuristic and a model.

2) Plagiarism of all of the world's smart words written by smart humans, and it's remarkable how close anything you could write is to something that's been written before. 3) A model trying to predict the next word that has some remarkable emergent properties.

4) The cost of generating reasonable-sounding bullshit has gone down way faster than the cost of detecting it. 5) A useful thing that we will figure out more uses for, if we can find applications where false positives and false negatives are acceptable or mitigated.

And if you think four hours is long.... jamanetwork.com/journals/jam...

Penn LDI@pennldi.bsky.social

How long can patients wait for inpatient care after they’ve already been admitted to the hospital? A study co-authored by LDI’s @abfriedman.com found 17.3% of admitted patients waited at least four hours for inpatient management, highlighting a #PatientSafety risk. Read more: https://bit.ly/4hlSm0T

Ari B Friedman reposted

Ari B Friedman reposted

One of the best examples of age heaping #demography Smth like this: - Who else lives in the household? - Grandpa - How old is he? - Ehm... 85

Ari B Friedman reposted

In residency, I had the habit of looking at the timestamps on critically ill patients who were transferred from outlying community hospital EDs. I was frequently in awe of the emergency physicians there and the two ED nurses they typically would have at night, who in an hour or so could...

sciencedirect.comsciencedirect.com
View full thread

In residency, I had the habit of looking at the timestamps on critically ill patients who were transferred from outlying community hospital EDs. I was frequently in awe of the emergency physicians there and the two ED nurses they typically would have at night, who in an hour or so could...

sciencedirect.comsciencedirect.com
View full thread

In residency, I had the habit of looking at the timestamps on critically ill patients who were transferred from outlying community hospital EDs. I was frequently in awe of the emergency physicians there and the two ED nurses they typically would have at night, who in an hour or so could...

sciencedirect.comsciencedirect.com

diagnose, stabilize, call for helicopter transportion, send labs, intubate, and place a central line. We know that critical care saves the lives of critical patients, and we know that minutes matter.

Ari B Friedman reposted

Eye popping. Court just unsealed Summary Judgment motions in New York Times v OpenAI / Microsoft. Easy to see why the AI companies had so many redactions (pink). Their own people wrote the lede for NYT, the "largest theft of labor in human history"... "more and more substitutive..." /1

Ari B Friedman reposted

Absolutely surreal to have lived through the Napster days and watched as the titans of industry moved hell and high water to lock down any possible chance that someone's song might get swapped online without giving the studios a cut and now those same people are just like "well we gotta steal music"

Ari B Friedman reposted

Ari B Friedman reposted

The fallacy that University leaders pushing AI are falling into is thinking that the point of undergraduate research is to find things out. It isn't. What students find out is typically already generally known. The point of undergraduate research is to learn how to find things out.

Ari B Friedman reposted

Amazon is officially shutting down Mechanical Turk next month. It feels like the end of an era. I've written about the platform on and off for close to a decade. We wouldn't have the AI tools we do now if it weren't for the low-paid, largely invisible labor of MTurkers

Ari B Friedman reposted

Is AI being used in your healthcare? 🤖 🩺 Our new survey asked Americans about their views on #AI at the doctor’s office. We found that they’re uncertain about whether it’s used in their healthcare, and they want to be informed if it is.

Ari B Friedman reposted

we all have mental rolodexes of patients we never forget as an ER doc, I've taken care of quite a few patients where we found new -- almost always advanced -- pancreatic cancer this is great news and a nice reminder of why we all chip in and hire the best scientists to find new & better tools

Kevin M. Kruse@kevinmkruse.bsky.social

My grandma died of pancreatic cancer and, as I recall, it's one of the swifter, more devastating forms of cancer. She was gone within a couple months of the diagnosis. This is absolutely incredible.

So many things in healthcare and policy are like this chicken coop: surely it can't matter if the siding ends an inch too high? Yet the details matter in the long run, and that missing inch of siding costs an entire coop years later.

This post is unavailable.

How will you know without expertise? As we rush to deconstruct systems that have long functioned, we should be thoughtful about not causing unintended harms along with the significant potential benefits. Move fast and break things has its place, but in healthcare the "things" are humans.

Ari B Friedman reposted

In a new NEJM Voices blog post, @iwashyna.bsky.social explains why physicians should make targeted, moderate-scale philanthropy a core task in their career development and why doing so is particularly urgent and valuable now. Learn more and sign up for email alerts: https://nej.md/3UzTlBt

Ari B Friedman reposted

"By writing, you are literally firing up your brain and therefore stirring up your conscious thoughts and something new emerges. You are not, or at least not always, transcribing something already present in your conscious thoughts.“ LOVE this piece. www.historians.org/perspectives...

historians.orgHow Writing Leads to Thinking (and Not the Other Way Around) – AHAWriting is stressful. Sitting in my computer chair my neck and shoulder muscles almost immediately tense up as I dig around in my brain for the best phrase or even any coherent string of words.

Ari B Friedman reposted

Does postpartum visit attendance vary by birth timing? LDI/CHOP experts found that people with preterm or late/postterm births were more likely to miss postpartum visits than those with full-term births. Risk was highest after periviable births (22–23 weeks). https://bit.ly/4hyOzhf

bit.lyOBGYNIntroduction Approximately 40% of obstetric patients do not attend a postpartum visit. Postpartum parents of preterm infants (<37 weeks) are often sicker than parents of full-term infants. Nonethele...