1- Did I get through my Sunday to do list?
ALMOST. I think I did everything except for G’s room organization (her books). It was ~4 pm, I was DEEP into Fellowship Point, and I made the executive decision to sit on the couch and finish it rather than do the organizing. (She was not excited about it either).
I am still determined to get through her room in October like I got through C’s in September, but have just adjusted the timeline. Next Saturday will be go time (I’m not on call this weekend and we are headed to the Keys for a bit, but not until Saturday late afternoon which means that day is actually perfect for some focused room time).
2- FELLOWSHIP POINT = DONE!
I don’t usually feel like I have to push myself to finish books, because truthfully if it feels that hard I generally abandon. (At least this year I have!). BUT I did like this book, and wanted to find out how it ended, I enjoyed the writing style, I enjoyed the character development and the strong sense of place.
I just didn’t always like reading it. There were parts that just made me sleepy. I am not sure it needed to be 579 pages, you know?
That said, I really enjoyed the perspective of the two main characters (both in their 80s!) and I liked the ending despite the fact that Jenny told me some people do not (she liked it too!). It was also fun to text Jenny while the story unfolded (fun fact: somehow Kae was on that text chain — because it started with us sharing pictures of cozy fall things — so she may have gotten some spoilers – sorry Kae!).
I’m reading The Shampoo Effect next because it’s a non-renewable library book. Onward!
3- I have a cardiac monitor on and it’s apparently supposed to stay on for 14 days. I usually do little quick ones but I guess my EP wanted to get a broader perspective (seek out runs of sneaky VT I guess?).
I’ve had a lot of skipped beats/bigeminy (basically when PVCs/ventricular beats alternate with regular beats – not acutely life threatening but feels terrible) so trying to quantify and figure out if I need a med adjustment.
4- CALL IS UNDERWAY. I am sure things will get busier but today has been totally fine (I’m writing this mid-day Monday for publication Tuesday morning so maybe we’ve gotten 3598734 consults since I hit publish, but hey — at least I got to enjoy part of one day that wasn’t too busy). Just 3 nights on this week (I’m on full days, but have night coverage Tuesday & Friday). Then I’ll be on next Tuesday & the following Friday-Sunday weekend. THIS is how I like to take call! Determined to continues spitting up as many weeks as I can.
5- I thought this was a great ep: An Unlikely Team Tackles AI Brainrot. Helped me understand some things about the current AI conversation, AND I agree with all of the practical tactics he mentioned, which included
- Use AI for tedious things, not to do the hard things. I love using it for creating planning documents but it’s my calculator, not my decision maker. I remain the one in charge of what I want to do, how long I think it will take, and which hard landscape elements must be included. This is very different from giving the tool access to your calendar, your email, and your task manager and just prompting it: “Hey! What should I do today?” (not saying anyone here is doing this, but I think this is a nuanced space and preserving autonomy and the ability to think deeply about what we actually WANT is important
- Don’t use AI to write // any time an LLM is used to express anything it should be noted as such, otherwise it is dishonest / emotionally manipulative. Yep, hard agree.
- Make sure you are READING things that are not AI summaries of other things. Yes, there could be the occasional summary of some big document where you need to capture the main ideas, but this should be the exception not the default.
There were some other ideas mentioned in the paper discussed in the episode, but I liked these the best.
All right, onward + upward. And a fun concert tonight! I want to hear all of the super old ’90s Weezer though I also liked OK Human a lot!
15 Comments
I am looking forward to that Cal Newport episode! I was interested in his thoughts about using AI for writing. I have started using it to draft emails for really tedious things – all the life admin stuff like when my grocery delivery comes with a pear that is rotten, or when I returned some items and they didn’t give me the credit.
I figure when a company is using AI to manage the chat agent then I’ve started using AI to write emails to the company to try and get what I need. I realize this feels like a total AI race to the bottom where we just have AI battling AI but perhaps that’s leveling the playing field of consumer protection a bit?
Completely agree – I used AI to figure out what words to use in a recent (months long) exchange with an insurance company. It reduced the stress on my end, enabled the issue to be resolved appropriately, and I am guessing helped the insurance company too since they weren’t trying to translate what I meant from the non-expert words I would have otherwise used.
I agree this is a valid use case! You’re totally right!
AI being dishonest is *why* to use it for charged work emails. NO WAY do I want it to convey the depths of my true thoughts lol. Its manipulative corporatese is *chef’s kiss*.
Have to agree with this – charged corporate communications where you are emotionally compromised get a run through with AI.
you guys are so right! I guess i was more thinking about . . the writing I typically do in my emails. But like to fight a (probably automated) denial for a necessary medication? AI ALL THE WAY. (bc much of it is GARBAGE work to be doing that shouldn’t exist to begin with!)
Sometimes even when the communication is not charged (though definitely then), AI has made my communication more persuasive or clear. And it can be relied on to fix typos. For instance, nobody wants to obviate the impact of an important email with an opening line that says, “I hope this finds you hell,” for instance.
HA! Okay I can definitely see value for this use case!!!! (ESP THE INSURANCE BACK AND FORTHS OMG)
I’m still working my way through Fellowship Point. I’m finding it to be a bit slow but not to the point where I’ve ever wanted to stop reading it, and it’s a nice change up to my normal reading. I’ve been reading it so slowly that I’m going to have to put my Kindle in airplane mode so that it doesn’t get returned to Libby before I finish it.
100% agree that AI is great at handling the boring stuff.
Good luck with call!
Glad call is going well so far! I’m looking forward to that episode of Deep Questions. I agree about not getting AI to write things for you – so many podcasts/Youtube episodes are suddenly being written in the distinctive AI style and I HATE it. “And here’s the one thing most people miss…” “And this is the thing that changed everything…” Argh. Sure fire way to make me switch off, no matter how good the ideas. As an academic I’m not really worried about brain rot for myself, as I read so much science and challenging ideas, but definitely something to be mindful of as our kids get older and have the option to get “AI” input into everything they do, if we were to let it.
Hahaha about the book spoilers… I was like OOH what exactly are we talking about here?!?! 🤣 No worries- I am sure by the time I ever get around to reading it I will have forgotten what you guys said anyway….
About the AI, one interesting thing I heard at a conference recently about AI was something I had already been wondering about too: Since AI uses existing content to “learn”, and the way it works is to essentially quickly search everything that’s ever been written on a topic online, apparently there is increasing concern about AI “eating it’s own tail” in the near future. Like, as more and more people are USING AI or AI summaries to get answers to things, fewer people are say, posting on Reddit threads to ask questions, or leaving comments on real blogs, or leaving reviews, etc. So therefore AI could quickly end up with old/ outdated information that it’s referencing. Or, as more things are published that used AI in the first place, eventually AI may just be reading other AI content to create NEW content.. which in the end is not the same. Not sure if I’m explaining that well, but it’s something I’d been curious about already.
HA! I’m pretty sure we didn’t say anything too spoiler-y in the conversation with Kae. She probably just skipped over it anyway once she realized we were talking about a book she wasn’t reading. Anyway, I’m glad you liked it despite the laggy middle.
Good luck with call! It sounds like you’ve figured out the best way to do it.
How are you approaching AI with learners or invading our EMR? We now are forced to let the ‘AI summary’ of the patient’s chart load, which I hate, because it’s buggy and slow, and I can read by myself. It’s also pulling in very incorrect info more often than I’m OK with.
That part I’m annoyed by, but can get over. I am concerned when medical students and housestaff are using AI for note writing, DC summaries, and basically…anything for patient care. Note writing is long and slow, but it’s also kind of an essential skill. I’m biased, as ID though, because we literally only have notes as our *thing*, but the act of writing my A&P for either a primary team note or consult note is a good exercise for me. I’m not sure how to counsel the learners on service about that in a way that they get it. When a medicine intern H&P is less useful than a neurosurgery note, we have problems.
Thank god I squeaked through training before AI infiltrated everything in medicine…
Ours loads but I totally ignore it. I have yet to get anything useful from it.
I have *not* really seen our learners do this other than sometimes I’m like “hmmm was that assessment section copied from OpenEvidence?”. I absolutely agree note writing teaches you to understand patients in a very special way. (When my colleagues have residents rotate and don’t ask them to write notes I’m like Oh NOOOOO they will not learn nearly as much! in fact sometimes I’ll have an epiphany about a pt while writing a note)! I will start to pay more attention to what the residents are doing, but I know when they work with me there isn’t a lot of opportunity to use AI really. However as more features roll out in EPIC that may change.
NOW, I am ALLLL for effective AI doing my documentation effectively for me. I don’t currently use DAX (the only option available here right now) but once it integrates better I can imagine this will be super nice. Right now our system still requires like 3844 clicks to order a simple lab. And the “suggested responses” to in basket messages have added zero value (they are useless.)
That’s a good point-the use of AI to make documentation easier! I am sad when I see ID rotators use AI for note writing because in the same breath, they also say they love reading ID notes (and I love gathering that exposure history just as much!)