Welcome to Real Talk on AI and I care with your host, Dr. Scott Morris. And I'm co host, Dr. Rehan Ahmed.
In each episode, we discuss current news, adding a little AI education, and debate innovative AI tools and topics that will change the industry. All right, well, welcome to Real Talk. Scott, I knew I knew you were just at the American Optimetric Association. You had a great time there, you were doing some panels.
So I'm super excited to hear about what you learned. And we'll talk about. Yeah, so I was very fortunate. I will thank the AWA for letting me kind of moderate two of the panels on something a new project called the Innovation Hub.
And it was really fascinating. So the first group was kind of my international panel. There really was talking a lot about akelamics about how what some of the new technologies that are embedded with AI are going to do in terms of using the eyes as not just the windows to the soul, but the windows to the rest of the body. And so we had Patrick Sauvillet over there from Zili who's talking about some of the work they're doing in a symmetry and looking at metabolites, oxygen metabolites.
And with an OTT kind of device that we'll be able to look at it in the coroid layer in the retinal blood vessels at the optic nerve and looking at what's happening is some of the metabolites for something like neurodegenerative disease and, you know, looking at it from in an eye perspective, kind of what macular degeneration, what metabolites is leaving behind that might be precursors to drosen, you know, especially when you think about GA and some of the things that are happened in terms of blood flow there. And that's Zili as EIL IA. Yeah, Z I L A. I'm fast, I'm almost similar or that attached your finger, but instead it's doing it in your eye, I'm happy I oxidative stress.
It's doing your eye and it's looking at metabolites and they really feel like, though I the eye care part of it's great and I was kind of hinting on my thing, I think that might be the future of glaucoma as well looking at blood flow and the optic nerve and what metabolites are happening there. I also think their their big push was they believe that they're going to be able to detect neurodegenerative diseases like Alzheimer's disease potentially years, if not decades before symptoms ever show up. I mean, there's truly where the eyes are the window to the future of healthcare. So fascinating stuff.
Catherine Bourbon was the CBO of redis back and she was talking about some of the Alzheimer's work they're doing and they're looking at a totally different way of looking at croitle blood flow and what's happening at macular levels in terms of looking as that a precursor to Alzheimer's from a totally different perspective. I'm using AI to build to do visual machine learning and look at these kind of, you know, visual models to do those types of things, but fascinating. I mean, I got done I was really excited as moderating the panel because it's like the future of neurodegenerative disease may not be in the hands of neuro it may be in the hands of eyes, you know, in terms of us being able to see it. That was pretty fascinating.
Then we had Laura piece of roll from Paris, who was there talking a little bit about her technology, which is called side view and and a refracting system that is kind of unlike anything else I've ever seen before. And I think we're going to talk about some of that kind of stuff and fireside chat here, but you know, in Europe where the model is very different in those not really even optometry there for the most part. Optimology is largely surgical medical care where optical is playing much larger role in terms of primary care and some of the things that her technology and the AI that's embedded in it is able to see and do and get people into the offices as kind of preventative care is pretty spectacular. So once again, I tied it all together, but it's all these really cool innovation pieces coming together in terms of oculomics.
The second one we did was we did a panel and we had my friend, Dr. Easy on the other talking with some of the stuff he's doing in terms of developing eyewickies. We had Colton from party was on there talking about some of the stuff they're doing in terms of trying to streamline with ambient scribes, which I know we've done a couple podcasts and ambient scribes. And what how that changes their model and also some of the stuff they're doing in terms of insurance and financial processing that was pretty fascinating stuff let little less AI but a little more innovative thinking about how we're going to change the way that part of the model works.
And then last but not least, I'm a grinch met who's there from ultra is talking about what they're working on in terms of their OCTs so all fascinating stuff like it was truly a talk about innovation for all morning and away and some really great interesting technologies out there that. You know, I think Ray on that you and I keep pretty up to date with all the AI technologies that are out there because it's kind of our job and we kind of have to. But it's always fascinating when here is people come in and talk about new ways that AI is being used from a machine visions perspective and other perspectives. That probably when we figure out how to merge and mesh some of these technologies together how and I think we're going to talk about it fireside chat a little bit today you know is that.
I think it's fascinating how all these innovation things are coming together and AI is going to be the brains behind a lot of it. Yeah one of the things you said was that really struck me was is that idea like AI as a connector of various fields right in terms of occupants you said well it's going to be. The the fund is photo the right the phone that the neurologist is going to use now and you know you may go to next time you go see neurologist we don't have to see a neurologist but they may do a rental scan on you. I think we may be co managing neuro generative disease with neurology yeah.
And you know and the fact in another way is like you mentioned the technical side you right they they are connecting what we do at AI and I think we're talking about the three O's optical optometry optimology and and AI is in a way connecting all of us. And so truly, truly disrupted and you had finished by talking about sort of a vision language models and that's a good a good tie into what we're going to be talking about today in the knowledge right. So on the knowledge right today we'll be talking about VLN for visual language models and this is kind of a good extension from a kind of extent on what we were talking about Scott and where we talked about. LLNs are large language models and before that we talked about nlp's and we then we talked about neural networks and AI and so this is kind of like we're just taking you know our are the step by step sort of the frontier of what what's happening now.
And so the next really what's really exciting for talking to the agriculture is about visual language policy and we just start to finish up with all of them so now we're ready the next thing but the LN so what is the beyond. It's really an extension of what we're talking about that it's a multimodal. And so it's two key components the first component is like a vision encoder and Scott this is what you talked about before and it's a convolutional neural network or vision transformer that basically turns an image into you know like a numerical representation right but the second thing is is the language models that we know involved like. Maybe T or long that process of text and understands context sequence and you know hope maybe probably have but maybe meaning or something a process.
So can I can I paraphrase this I was yeah I I was listening to a book called sapiens this last couple days and what you said was almost like word for word what they were saying in terms of I think we. Correct me if I'm wrong we're going to take a picture we're going to digitize it we're going to turn into a series of zeros and ones that's digital language that these VLMs can then read categorize process is my on the right track there for that you're on the right track that's a secret sauce though is that there's a fusion mechanism it's across attention module and so. What it does it that focal point here that in mind what it reads or writes and then it's trained on data sets where images are paired with text so like it's awesome on the ground commentary that you think about red line images it's an example use a lot are off down reports you know chest X raise you know picture by healthy object right and this goes really well with what we talked about last week with net gem right because it has 27 billion parameters of text and and and vision that are full four billion parameters of vision. So so how does it so like you have a picture of a retinal photo and you just say describe describe this photo not this is photo have diabetes or not not does this photo have thought home or not it's just a retinal photo it is broad which is different than some of the die different outfits that we've talked about here it's it would just describe it like a normal person great in this case that be you know.
In large cup to this with potentially some you know some scattered about hemorrhages consistent with my own tortuosity with my blood vessels in the same way I mean you can imagine a system where you just drag a fund this photo and it just sort of does it for you another way which is maybe even more abstract you may just say hey this is it was CT why does this patient happy these vision 42 years old 82 years old OK and then maybe. Say something like oh just sub right now fluid seen in the OCT and that's suggested that active CMV based on that xyz right so those are the other and then it combines that into the rest of the note right you're going to have these combined everything's like well just a 66 year old with the worst thing several fluid and metamorphosis and then give you a treatment plan like consider and I bet you're right this is like structured actionable intelligence and you know it's sort of the models right now if you go into a TFT or. Lama actually Facebook certain methods called lava it's not lamas that's a large language and vision assistant call lava and these are already sort of available and you know if you're using chat to you to your probably already using the vision language vision language model approach I know I do actually a picture of something I don't like I was looking at a plan you know and there's an app for this is where. Well I had to be so sorry oh my gosh I literally downloaded that this morning for my greenhouse because I have a plan I have a couple of plan I just don't know what's wrong with them I cannot get them to grow no matter what I do and only you guys have seen that TV you know I take a picture and it goes add epsem salt I'm like really sure enough you put it on and two days later it looks great you know so I it's going to help my gardening skills I might actually have a garden that looks presentable yeah it so it's helping you.
So it's helping you in you know getting a green agree robotic and so this is this is the next step so these are I hope to talk about these are again vision language models the next tension from l l and it's it's essentially multimodal and you know you could sort of tease next steps and think about what's after this and that what I think that this is that we're going to go into a robotic. Revolution you know and maybe that's our next but that's where I think there's now we're going to get some controversy you know I think you know that's right here some of the people say that for next time talk about yeah I think I think so I think so you know as I was traveling this week and I was in the United Club and you know the robot comes and picks up my dishes and says thank you so much for putting your dishes away I was like okay I I can't remember what movie from the jets what what that wasn't it what was it named from the jets and. Mark. No I can't remember the name was but the robot that came and did the housework I was like oh my gosh we're really not that far away you know so crazy stuff well that's a good transition so let's talk about what do we think some of the transitions are going to be and you and I've been talking a lot about headsets right and wearables where the wearables I think maybe are I often get myself pigeon hold or nail.
I'm not really thinking about wearables in terms of like a ring or a watch or you know I'm in meadow or whatever we're going to talk about heads up you know display and I think that that's that is such a small little consumer driven part of where wearables and how that's going to impact I mean I think about you know you and I've talked about some of the diagnostic technologies that are going to be a big device. I'm not sure topography is going to be a big device I'm not so sure that. Refraction you know for optical is going to be a big device I mean I think that we're going to see wearable heads that we're already seeing that right I mean you think about what was the name that radius right I mean they're doing that's a good revolution of hey everybody's got a smart watch everybody's got a smart ring you know and and I think we're going to see the revolution go into headsets wearables that kind of stuff we're going to see it come back to our world which is I care but then I think you know I'd love to get your input on where do you think this is going diagnostic technology wise. Yeah and so yeah there are a number I have a headset you know we use our clinic there number of really great devices from you mentioned radius.
There is he grew all live virtual field and several others that are all quite good I think their value proposition is how portable they are and they're increasingly getting better it's the idea and you mentioned this too we use a number of different technologies and the bioplin and they're kind of big and they're expensive they take up a lot of space and the topography you know not like you mentioned prior to refraction and as we get smaller more modular mobile devices headsets being one of them and what happens when you attach AI to it so I think in in Prematory for example you know we do it for typically for glaucoma screening I definitely see a value and staging the disease what diagnosing a disease based on a visual field and other elements. Seeing if there's progression or not like a like a GP like I had a progression analysis that's available like you know why can't have a headset that would AI that's built in. And then you know prediction when if you have a head if you're using a headset and especially at scale and the people patients are using it at home why not throw it in and being able to predict disease before you get happens because you know the the primary end point for glaucoma is it pressure pressures just a body. It's visual field damage and so yeah but what better way of testing actually testing it so I think I see a ton of AI applications and the headset space and that's just Perman tree if you know headsets are doing more and more and it's going to be interesting to see how they are they continue to develop.
Well and I think everything emerged from being just agnostic testing from a patient perspective to education right I mean I was on the plane on the way back and I have to say I'm a total tech geek I love my tech you know I spend way more money on crazy stuff that I probably should my wife would probably agree to that you know but but the reality is that there was a guy sitting next to me with one of the new you know headsets that gives you 400 you know whatever it is. 200 centimeter view he's sitting there watching a movie and I could see him like just smiling and doing all kind of stuff I couldn't hear what he was hearing I couldn't see what he was seeing and I said is that one of the new movie cameras he goes Scott he didn't say Scott he said he was sitting next to me he goes he's like this is the coolest way to fly ever he goes I am watching it he can't what was this I'm actually was watching an I max movie in 3D on his head. And I was like oh my gosh I'm getting one of those I'm never flying the same way again right I'm just like I want to watch Gladi or two and 3D on an you know that kind of stuff I have to fly for hours I'm not as well enjoy my four hours but I thought that I thought to myself was like how is that going to change patient education now because instead of you know you and how are tonight didn't want not too long ago Dr. Priscilla I do one of these I think you're on vacation we did one about the future of Asian.
And we didn't really get into the subject but it's been on the back of my my back burner for a while is I think education as we get more and more into the wearable lens type of evolution revolution where you have augmented reality you have virtual reality in maybe a clear pair of glasses or maybe they're able to shift and turn into something different people get more comfortable with receiving information that way I see a day where we're not sitting in a room educating people the way we're currently educating them which is let me sit and talk to you. But maybe we're going to be able to pre-educate in their headsets multimedia in a way that we can't currently do right now. Yeah, and then the next episode of AI and actors going to be in the metaverse so there you have it's got in the metaverse I'm not so sure I believe in the whole metaverse thing but I agree with you I mean I think that the next piece of that is is that maybe we're going to bring education in a much more interactive tick tock micrural learning kind of perspective so that's another thing I would like to cover one of these days is micro learning because I think there's a huge opportunity in that. But I also think you know I as I sit there next to the guy in the plane I thought to myself huh maybe we're thinking about all this is wrong or maybe I'm only being very limited in my thinking so when I think I head up display I'm thinking about it from patient diagnostic testing and patient education and wearables in terms of getting biometric data from them but then I started thinking well how is it going to affect what we do ray on as a provider.
And I started thinking oh my gosh you know talk about me squaring up we'll say you're patient me squaring up with you putting on my headset which are really just my glasses right I mean you can have your meta glasses on as they evolve and it'll be other companies than met I'm sure but you do this but we put it on and all of a sudden we say well you know you have a bacterial corneal mid peripheral ulcer and what pops up on my feed in my head and I'm going to have a head up display is all of the different treatment options for a bacteria corneal ulcer with the references now no longer do I lose track and not look at you as a patient I'm still focused on you but the data stream of all the potential evidence based medicine is right there and you know I've been working on that with my project now for a couple years and I think we're very very very close to doing exactly that we're we're going to have heads up augmented care augmented have a helper who says here's all the data right I mean I do think that it's going to change our lives when we're wearing heads up and we have access to all that information and we don't have to be pulling out of our brain because it's showing up on us the display is like do you remember that did you ever see that show at movie gosh this was years ago minority report oh yeah Tom Cruise that was the first one that we ever looked at when oh my god there's a screen you can touch and move and now I think we call it everything we are around you know I mean your surface your phone your whatever everything's touch screen you know when he put the headset on and all that information about whatever the twins or the triplets or whatever they did all show up in front of him I was like oh that's cool well I don't think we're that far away from that I mean I think we're going to start seeing them the next year or two how does that change our job as a provider when all of a sudden all of that information is there and we get to spend our time focusing on human care caring for the human sitting in front of us instead of trying to remember every piece of data that we're supposed to remember just a thought teach on looking forward to that and definitely looking forward to that day so we'll hope everybody today learned a little bit we talked a little bit about what was going on some of the innovation things ran did a great job talking about visual language models and how those are probably the future just like you know we have LLMs right now VLMs are going to be just as important part where you start to digitize images instead of just text and then we got a little bit about what the headsets and wearables look like in the future so you know I think those are all fascinating things hope you guys learned something about that if you want to learn more we have a whole series of both our innovators podcasts this podcast as well as lots of references you can go to ai and iCare.com and look up the journal articles or have referenced any of the podcast that we have you can watch them on the site itself or you can listen to your podcast player of your choice we pretty much have all of the different platforms that you can listen to us so everybody I hope you learn a little bit and stay tuned next week I think we have all kinds of good stuff that we're going to talk about you've been listening to real talk an AI and iCare your weekly podcast to keep you informed about AI technologies revolutionizing iCare