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, so welcome to News of the Week. And this week we're talking about now their aspect of Google on what they should bring to the market. So last week's got you, you talked about the B03, which is their really interested AI video generating tool, which I use it is fantastic.
This week we're talking about another of them which is that they're the world, I own conference and it's called metgema, me Jim. So metgema is a multi open source model, trained on again, master data sets that are purpose guilt for medicine. And there's really true reasons that we should care and the truth for things that I think are really important. The first three reasons as I just mentioned is that it is trained entirely on medical to limit.
So you know, we know chat, you tea on it, every night. I like playing, all the polls, net, books, YouTube videos, all sorts of other stuff. Netgema and comes in two variants. One is a four, four billion gram version and the other ones are 27 here, 27 billion gram version, which is their text only version.
It's trained on entirely medical images and text. Basically, it's about contextualizing five, and clinical languages and images is basically what we used. Physicians, write notes, replications, all that sort of stuff. So it's being chat, GPT or plot entirely medical.
So that means it is very, very high quality and it's basically a domain expert. Can you medicine? The other thing, I think this may be a figure to deal. It is open scores.
So you know, Scott, I know you work with civil startups. I would you advise several and what's really interesting, why do the deal it, you know, it means that start academic, you know, much individual dollars can fine tune their home. See it. So you have your own bread, you're ready and grab your own bread and all this or you have really get to some start from your heart.
And rather than building stuff scratch, it can allow you to read thought that sort of big area where you need just a bunch of data to get started and then just start by tuning based off your data. It also sort of allows us to unlock occupants, which again is basically the idea of using the eye as a central disease or looking at the whole body. So what we put in red and show because it has 27 billion parameters and text, it sort of opens up the field of economics in a way, especially for new startups, the way that really was possible before. So I mean several scenarios that we can come up with and I recently did you look in a yard, you know, I picked it all I'm super article about this, but imagine you have a fund is for right we go with sort of like the classic thing we've got a tech or diabetic right now, but instead of that, suppose now you just say just summarize this on this photo in clinical pros, fly anything in you.
Is there anything around it is this patient at a risk or any neuro gender disease this and put it in a patient from the expedition. So done it you could do this, you know, fly location sees you have all this ready to go this comes short sort of practical now I see with note that metgemon is not wonderful great. It requires a bind truly by by you know by the start of talking about or the active group. So it's not which is good in my it needs to hack regulatory type of validation, but it's a huge first step I'm super excited about this I think this is the start of even more new company from the space and really excited or this and I'm really going to give it or to climb with this and having this move that feel for their events, oxygen and core.
Yeah, I think it's just the next evolution we always talk like everything seems to be an evolution anymore, but I think it's the next evolution of taking data now still where we get the data still the problem, but taking data and moving in multiple steps down the process. This week for the knowledge bite I want to talk about as soon as we have in this conversation a lot the last couple weeks and maybe think if I'm having this conversation this many time with my friends and colleagues maybe it's something that would be worth sharing. I think there's a lot of confusion out there about avatars and agents and now kind of the new term agentech AI and I think that many times they're overlap and maybe confused and what I really wanted to do was talk about those are really different concepts that may or may not overlap so let's kind of go we're going to break them down a little bit at times so avatars you know avatars it's not the movie avatar we're not eight foot tall blue people, but you know this is essentially a digital representation in a virtual environment now maybe that's a representation of yourself like myself and it's like my online persona or kind of proxy. You know or maybe it's a person of someone else you know I think avatars can be using lots of different ways they can look very very humanoid or human like these days or maybe they can be more anthropic where they look more like a cartoon and you're trying to represent some brand and just really it's a graphical depiction you know it could be 2 day it could be 3 day it is really just a identification that you use somewhere else to represent you or somebody else.
But I think they really kind of interesting point about is it's under complete human control meaning that you asked the human user we control what the avatars actions movements it's appearance it has no autonomous action it is totally a predefined set of who out where why went. And so you know can think about those as bitmoji was we talked about that once before was kind of the original or someone you're using for your chatbot or or somebody using my kids use it you know who they are on Minecraft or whatever you know and those are the avatars and I think that's kind of we're transitioning away from a world of avatars to maybe start to be a world of agents. Now let's talk about that agents is really just a piece of software or some system that can designed is really designed to perform some specific task or action on behalf of somebody else now they can be super simple they can be super complex it's just an operation it's some kind of program software it's code that operates with some degree of independence to achieve the end goal it was programmed for it is really just a simple task. So it's a software program well at least for now you know and though while they act independently and they don't need human supervision because they're pre program to do certain things they counter very task specific and there's lots and lots of agents and more and more we're seeing multiple agents working together one agents good at doing a one agents could it be and one agents could it see and when you attack them together now you're this new super agent is good at a BNC.
You know and we were very familiar chatbots chatbots are a type of agent you know there's assistance like Siri or Alexa those are agents that either play music or answer questions or control your devices or whatever the case may be but those are all just agents. Now we get to what I think is the really fascinating kind of evolution that's happening which is called agentic AI and this takes that concept of an agent says well instead of it being pre program to do a certain thing maybe now we're going to give it a target a goal and it will have to use sophisticated reasoning or planning to figure out how to get from a to see it this is really using artificial intelligence and so it's not just performing. It's actually learning how to plan sequences adopt strategies form different objectives of how to get to that endpoint goal. Now I think that we're taking avatars and you could have an avatar that is the face of an agent or the agent could just be a chatbot and doesn't have a face to it but I think when you prime that or now position agentic AI in front of it now you have really some pretty interesting things and I mean I was thinking the other day we're talking to you and I've talked before around about GPS you know GPS is really AI it's been AI since for years and years and years and it can tell you you go to a certain place and you may choose to follow that path you may choose to do something different whatever the case may be when we get to agentic AI it may say hey I'm going to drive you from point A to point B to point C in your autonomous car and it's making those decisions for you and maybe even at some point it will reroute traffic lights and timings and vehicles to make so everybody can get there equally faster, equally better.
I think you can look at it from a game or whatever the case may be and I think you know I look at agentic AI and I think we're just I'm trying to build that right now with some of our avatars and our agents in my program you know about how we I think agentic agentic AI is going to completely change and influence everything the avatars the agent and it's going to almost give avatars or avatars being the face of an agent the ability to maybe not have independent thought I don't I think maybe that's going a little far but at least have complex dynamic interactions with other humans or even other agents so maybe one agent we'll talk to another agent and we as humans will get the output from that discussion. I think it's fascinating and I think this one of those things six months from now or a year from now we're probably going to do another one of these and it's going to be agents will almost be something we expect and it'll be very fascinating to see where it's at but hope everybody learn a little bit today about the different sweet avatars and agents and really where the future goes is putting some intelligence into those avatars and agents called agentic AI. Yes, I think this probably really well into our fire science chat because we're going to be talking about the future works for some eye care sort of in an AI power world and that world is going to be dominated by by eight or what we what we are calling ages at least right now that term is very well to keep the next six months and make out something else but what we're sort of thinking about when we think about agents and so I would love to get a little bit more of an agent. And so I would love to get your thoughts.
I just got my sort of view of is you know the file you know you always hear about this call we said it anytime it's like what the I take my job is what the I take you know our job and bar view is that AI won't AI will only take jobs people walk using AI so it will be an AI plus human world and I think one of the ways to sort of be about this is like should take the role of the world so I think about when you check when you check the mic on the local you have receptions and there's also a whole bunch of people in the back where to me and sure and insurance verification and scheduling and billing early on you know all those searches and send it reminders and so what I think was that another style of your thought is that you know we're going to have AI to do a lot of that. There's a fight night checking insurance. I think you're still going to have first sin and their role is going to expand they're going to be for see all these different tasks potentially women AI self in the YouTube has itself looking at all the past and maybe not other substance setting or in such a variation specialist they're called something like a patient experience that's the whole part you stole the words right out of my mouth there that that's exactly what I think is going to be it's going to be a patient experience advocate or patient advocate totally agree. That's what I think one so it's not going to take you have a fantastic reception and someone in terms of you will be up skill and they're going to be doing a lot more I think for real type of work and oversee everything much more all the stick and empathetic checking people and all the scouts thoughts.
Yeah I mean I think it's going to I think where we're going is going to completely automate all those routine tasks that nobody really wants to do and I agree I mean I firmly believe that our patient experience our patient advocates are going to have a completely different role and I think it changes the entire workforce I think that I'm not saying it's going to replace anybody's duties and get rid of a lot of that routine monotonous. And really human error prone type of situations when my agent can talk to the insurance companies agent verify everything and do all that send out to the patients agent and do all that where we can actually spend more time being human and it was funny you mentioned I don't even know if you knew this my article for next month is called H.I. Plus AI human intelligence plus artificial intelligence you set it out of perfectly I guess we're going to talk about the entire columns about that but I think that it is going to automate that and it's going to change how our staff works and it's going to redefine a lot of those roles. But I also think they're going to redefine all the roles for clinicians to I think it's going to be less and especially as we start incorporating more and more AI into machine vision and and you know sensory perception and all those types of things I think clinicians going to be less about doing we're going to do less doers testing and we're going to become more.
The cerebe as you said the cerebral part clinical interpreters of data between not just what the system saw but we're going to help interpret it in a very human empathetic carrying weight to our patients I think that the skill sets that we're going to use provider in terms of our digital literacy in terms of our you know how our interpersonal skills work. I think it's going to completely change what all of us do it's not going to eliminate as you said it's not going to eliminate us but the people who can adapt and become better at literacy and interpersonal skills are literally just going to have a significant or heads up leg up whatever you want to call it on providing care to people who are going to expect that that's the level of care they're supposed to go once you've been to one practice like that if you move you're going to be able to do that. I think it's going to be not willing to accept anything less than that in my opinion. Yeah, I took 100% agreed I'm really excited, curious maybe a little scared of how it's so that you mentioned how it's going to change sort of our experience as providers as you know as encounters and not monitors one thing you know I spent and I needed to do it like memorizing all these rare.
That's gone yeah all that's gone that's all going to change it's going to be about being able to. I'm going to go to. Fosterly communicate complex find a space and a way that connects that and this is going to be where I helped our humanity I know that's why I like to think. Because there will be much more emphasis on the things that the machine can't do that even that can't do really up and what of those things is connecting to another life being in front of you this year as I finished.
Yeah, I just think that we're going to I think the future of the workforce is going to change from a single provider who has all the knowledge in their brain. And may or may not be able to communicate that knowledge to their staff to any HR to patients to a very collaborative H.I. human intelligence plus A.I. agents as we talk before and but even more importantly how do we as you mentioned how do we create that improve the human element of doing that I think it's going to be our A.I.
will be our intelligence of where to memorize all those weirdly weird you're going to see it once in your lifetime condition of course you got to pull it up when you most need it. But I do think that was going to enhance that's going to augment our medical decision so we can be more human and be provide better care to human human individuals right so I think the workforce is going to change I think it's going to change what kind of doctors we need it's going to change what kind of staff we need I think that everybody's going to have to be more data sensitive more A.I. literate which is kind of the reason why you and I do these talks. So I would try to increase everybody's A.I.
literacy I think the future we're going to all have to be like that I think if you're not it's going to be very, very tough to compete out there in the real world. And then that's a great way to end this real talk we're talking about increasing human intelligence and human connection as well as I see. So please check us out at a I I here at dot com. You've been listening to real talk an AI and I care you're weekly podcast to keep you informed about AI technologies revolutionizing I care.