Could artificial intelligence transform the way we care for people with dementia? In this eye-opening conversation, neurologist and memory expert Dr. Jason Karlawish explores the promise—and the peril—of AI in dementia care. From tireless AI companions that ease caregiver burnout to the ethical dangers of handing caregiving over to machines, he shares candid insights every family, caregiver, and healthcare professional needs to hear.
Jeffrey Snyder, Broadcast Retirement Network
Joining me now is Dr. Jason Karlawish. He is the co-director of the PEM-PEN Memory Center, try to say that twice, and he’s also a columnist of neurotransmission for STAT.
Dr. Karlawish, it’s great to see you. Thanks for joining us in the program this morning.
Jason Karlawish, MD, Penn Memory Center
Pleasure to be here, Jeff.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, I’m a little tongue-tied. I don’t know why, doctor, but I’m really excited to talk to you about this. And before we get into the use of AI and whether or not it’s appropriate in dementia care, I wanted to ask you about the strains and stresses for both the patient with dementia and the caregiver.
I have to imagine there’s a big emotional component for both.
Jason Karlawish, MD, Penn Memory Center
Absolutely, and the topic of AI is very relevant to that emotional component. For patients, one of the strains and stresses is boredom, lack of things to do, desire to do something that’s engaging and meaningful. And for caregivers, there’s often a feeling of a lot of work and not enough time to get the work done, particularly the work of providing things for the patient to do and keeping them engaged in those things.
Those are just two of the kinds of stresses and strains, but I think they’re quite relevant to where AI has a lot of opportunities.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, I would imagine the caregiver, and it could be multiple family members, but it could be a spouse or some other loved one. You probably feel like a cruise director in some way, where you’re like, or a concierge.
Jason Karlawish, MD, Penn Memory Center
Yeah, it’s funny you picked on that. Yeah, in my writing for stat for neurotransmissions, I’ve quoted actually, that came from a caregiver, that’s not my creation, that they sometimes feel like either a wedding planner or an activities director on a cruise ship. And I picked up on that phrase that a husband told me because I think it captures kind of the, sort of the dark irony of what it’s like to be a caregiver sometimes, which is, it’s all about you, it’s all about you, and I’ve got to keep you engaged, and it’s work, and there’s also some of that distance that they’re beginning to feel from the person.
I mean, if you’re an activities director on a cruise ship, end of the day, you clock in, you’re done, and you’re good to go. And to say that as a caregiver, a family caregiver, I think captures some of the sad aspects of that work.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, I know Julie McCoy on the love boat had a very difficult time, so I can only imagine what it’s like to be a caregiver in all seriousness. Let’s talk about AI, because as we’ve talked about on this program, I know you talk about on your own podcast, you can’t help but go a day, not even a day, without hearing about artificial intelligence. How is that technology being leveraged in neurology, but more importantly, for caregiving of somebody who may be suffering for dementia, with dementia, excuse me?
Jason Karlawish, MD, Penn Memory Center
Yeah, no question that, even a quick search on the internet, oftentimes facilitated by AI, reveals a host of technologies that harness the basic logic of, and structure of a large language model, created by a generative pre-trained transformer, or in a word, AI. And I think it makes sense. One that strikes me as one of the more common ones that has been picked up on is some kind of role related to reminiscence, conversation, let’s talk, how can I keep you engaged in a topic, what would you like to talk about?
And the richness of large language models, to reach far and wide into a corpus of literature, provides that kind of opportunity. And in the essay I wrote in stat, I talk about a gentleman living with dementia, who takes great pleasure in getting on chat GPT, and chatting away with chat GPT. In particular, he got great pleasure out of talking about his past, with growing up in a small town in Maine, and the work that he did as a writer, and as a minister.
And for him, it was very rewarding. Chat GPT never got bored, didn’t mind if he was asking the same thing over again, or didn’t mind if it was going in directions that were interesting, which for this gentleman’s spouse, who was one who also called herself feeling like an activities director, was frankly quite relieving for her, and also I think gave her some degree of joy for her husband finding joy.
Jeffrey Snyder, Broadcast Retirement Network
So you mentioned chat GPT, in this individual’s case, were they literally typing on chat GTP?
Jason Karlawish, MD, Penn Memory Center
Yeah, which was a limitation, yeah. Because it required typing, and that was a task. And I think what I’m beginning to see is people who are developing technologies that require minimal amount of technological engagement, whether that’s typing, or even opening an app.
I’m beginning to encounter companies that are interested in telephonic, or extremely simple ways for you to get chatting away, for just these issues. So take away the hassles, it’s just like talking, yeah.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, I would think it’d almost be like a Siri or an Alexa. I don’t wanna give one brand over the other more priority, but you can have a conversation in some way with Alexa or Siri. Maybe this is more advanced, because sometimes Siri doesn’t always respond to my inquiries.
Jason Karlawish, MD, Penn Memory Center
Yeah, yeah, no. And I mean, where I see a lot of companies is trying to put together a large language model that really delves deep into the world. I think where I think things are gonna get very interesting is when things start to get personalized.
In other words, the more I know about you, the more I can have a rich conversation about you. And I think one of the great promises, and to some degree perilous, because now we’re starting to get into agentive aspects, but would be an AI-based technology that knows a lot about Jeffrey Snyder. Why?
Well, because I got into his iCloud and I got every single photo that he filled that thing with the last 30 years, along with whatever else I know about Jeff. And you could imagine a model trained to start talking with you about that trip to Poland that you took 10 years ago, et cetera. That starts to be extremely powerful, I think, for any of us, no matter where our cortex is, courtesy of a disease or not, as a way to really engage.
And indeed, our phones do that. They pull up photos like dogs that you have owned. I got that one recently and saw all my past dogs and whatnot.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, I wonder if it’ll be able to do a Baltimore accent, doctor. I guess we’ll have to, maybe eventually it’ll have those long O’s.
Jason Karlawish, MD, Penn Memory Center
Well, it’s very interesting to see how much these technologies are going to adopt a personality because that’s where things start to get frisky, which is, the gentleman I recounted named his chatbot. He called her Jane. He just made up that name.
So things start to get a little personal and this is where things get perilous. They can get weird. And I recounted in the essay some actual cases of individuals who got driven down ratholes of conversation courtesy of these chatbots.
Jeffrey Snyder, Broadcast Retirement Network
Well, look, I think it’s great. I think it’s good. Anytime, from what I’ve read and I’m sure you had a lot of interaction with caregivers, they’re under a lot of strain.
Sometimes they have jobs, full-time jobs, that they leave or maybe they step away from to take care of a loved one. This could give a lot of space to those people. It may give some time back.
But you also, on the flip side, don’t want, you know, I don’t think Siri or Alexa or even a large language model focused would be able to take over full, at least currently, take over the full caregiving responsibilities.
Jason Karlawish, MD, Penn Memory Center
No, because there you’re getting at the need for caregiving to be agentive. By agentive, I mean, having intentions that are real and being able to act on those intentions. You know, in its fully realized form, you could imagine, you know, a vibe coded house for someone, you know, that just kind of updates you on the news, asks what do you want for breakfast?
Has that delivered from the food delivery service? You know, sees the need to make sure the housekeeper comes in again and, you know, in other words, you could start to see agentive AI. You know, I use the word vibe coding.
People have created these sort of AI platforms that sort of do things for them that provides, does the work of caregiving. And their promise is really quite spectacular. You know, put it in a robot, literally a thing that moves around, you know, and it gets even more interesting.
You know, one of the upcoming plans from Mr. Musk after his cars and spacecraft is a home robot. And it’s not a question of whether, but when those get rolled out. And they’re intended to be able to do household tasks.
Make that agentive, that it can think and decide and things could get really interesting for the care of people with dementia.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, I recollect a movie with Robin Williams, I think called Centennial Man or Bicentennial Man, where he played a robot that eventually became more and more human-like. And I’ve also recall seeing like The Outer Limits, an episode similar. But, you know, is it a perilous slope?
I mean, let’s talk a little bit about, you mentioned some of the privacy, the concerns there. Can AI know too much about Jeff Snyder or Dr. Jason Karlois or anybody?
Jason Karlawish, MD, Penn Memory Center
Well, I think it certainly can know enough that if depending on who gets all that information, that could cause great harm. And that stands for all of us. And, you know, corporations can be quite frisky with who they share data with.
Witness, you know, go online, search for shoes. Two days later, go onto the Washington Post website and look at all the shoe ads that appear. You know, okay, I got it.
The cookies ate each other. But, or the program ate the cookies, whatever. I think where things get particularly concerning though for people with dementia is where a gentive personalized AI can go down dangerous paths.
Because, you know, in the care of someone with dementia, there are some tough choices that have to be made, you know, and or problems that need to be aborted. I gave one simple example. It’s the person who forgets they eat, you know, and wants lunch again.
You know, a human caregiver kind of said, well, you know, we’ve just had lunch. Lunch was an hour ago or whatever. Serves a very small second lunch just to kind of keep things going, but doesn’t make it another big meal.
You know, could we have a robot that reminds, I keep on using robot, but, you know, it reminds you that you’ve already eaten lunch and would it override further demands for lunch or would it say, okay, fine, let’s have more lunch, you know. That’s a very simple example, but I think it, you know, any caregiver will tell you it’s not just being a butler. You know, I’m just not there to serve the commands of the person who is paying me, which is, you know, the model of a butler or a maid.
You know, I’m there to curate the mind of this individual and help it thrive and keep it from harm. To be more direct, you know, I have patients who without their caregiver were, you know, falling for romance scams. You know, would an agent of AI have the judgment to prevent a romance scam or even more wild, could an agent of AI engage in a romance scam and take someone down those paths?
So the rules that will govern agent of AI for people who need someone to step in with aspects of frankly ethics are rather notable and as exciting as it is, you could find it taking people down very dark, dangerous paths.
Jeffrey Snyder, Broadcast Retirement Network
Doctor, I can’t let you go. I know you’re a professor of medicine and you have a lot of areas of expertise. Are we close to at least identifying the cause of dementia and if we can identify the cause, is it at some point in the future we’ll be able to blunt stop or even reverse the- So we’ve identified several diseases that alone or together cause dementia.
Jason Karlawish, MD, Penn Memory Center
The most common disease that causes dementia is Alzheimer’s, though not the only disease. We’ve identified at least four other pathologies that can help cause people to develop disabling cognitive impairments. Alpha-synuclein or Lewy body disease, TDP-43 or LATE or FTD, vascular disease.
So we’re pretty good now with labeling the pathologies and in the case of Alzheimer’s, we now have treatments that target some of that pathology that slowed down the pace. I think the pace of innovations that’s occurred as I wrote recently in stat, dementia finally has a business model. Our ability to both detect the diseases with biomarkers and now to discover drugs that alter those pathologies is really quite impressive.
So I think we should anticipate in the next several years that for some of these diseases, we’ll be taking the drugs before we have dementia to prevent us or to slow down the onset of dementia. I think that’s a very foreseeable future.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, absolutely. Well, Dr. Carlawish, we really appreciate you making some time. Great article, great column.
And look, we look forward to having you back on the program again soon, sir.
Jason Karlawish, MD, Penn Memory Center
Take care.
