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. We are looking forward to this week's podcast. We talk about one of the new important processes and news of the week, the release of chat GPT 4.5. Rehonsen had covered the different types of AI.
We promised you guys last week that every week we will cover a little bit of some basics so you can get a better understanding of concepts and terms that people are using. So when you have those conversations, you are well informed. And then we're going to spend a few minutes today talking about what we think AI is going to go and how it's going to affect clinical research and what clinical trials of the future might look like. So we'll go ahead and get started.
I'm going to kind of cover news of the week. So news of the week this week is there was the release of chat GPT 4.5. And for all of you who are currently using chat GPT, you're probably using 3.5. Maybe even someone using version 4.
Well, the kind of the big news about this one is it's not such a big deal. Chat GPT 4.5 is not releasing any more. It doesn't have any more power to do any other things. It's using the same language, large language model that version 4 used.
But 4.5 is going to be much more customer friendly to create a better user experience for everyone. So instead of you having to know how to prompt a chat something like chat GPT, it's going to have an interactive feature where it helps you figure out how to use it. So we're moving from a place where really generative AI becomes commonplace and somebody who is very well versed in it to starting to explore how people are going to use it and how people are commonplace common folks are going to interact with it to get the most out of it. And I think it would I would guess that they're going to track it and see what are people really asking how is the best way to use it.
And I think that they're actually going to be gaining synthetic data on this as they try to figure out how to better utilize large language models. So I think it's interesting that we're moving from let's keep expanding and improving its capabilities to maybe we need to figure out a better way to interact with it from a user interface standpoint. Yeah, there's something I've been thinking a lot about recently and I use chat GPT at home. And in the beginning, I would use chat GPT, Lodge, M&I, and I'll have them each talk to each other sometimes and see how they would be different.
But I found myself recently just relying on one and the reason is and maybe I disagree with this a little bit. I think it is a kind of a big deal. I think personalization is going to be huge because it has almost like a sticky network effect. And I'm like, I don't want to switch now because that chat GPT I asked it the other day, hey, like what do you know about me?
And it gave me three paragraphs of like where I'm from, my interest, you know, why I ask it about different projects I'm working on. And I've been working on it for a really long time and it knows a lot about me and to transport all that to a different LM now, it's kind of like leaving a social network. You know, it would be hard for me to leave. If you're leaving, you know, I'm an Android guy.
If you're an iPhone guy, you know, we're going to have that argument about who's going to switch and we're probably going to lose that because we're like, hey, this is what we're comfortable with. It's kind of weird. It's like this is something like a friend who knows you really well and you really just want to leave them with all that sort of institutional knowledge that you're going to have to restart. So I don't know, maybe this is a maybe it's not a big deal in terms of 4 to 4 or 5, but the idea of this becoming much more personalized and really knowing you.
I think that's it. I think that's cool. I think that's going to be a bigger and bigger deal. Like a lot of things in AI time will tell.
Exactly. So yeah, this week we promised you every week will be giving you a bit of the basics of AI. And so today we'll be talking about the types of AI. And so as everyone knows, AI is everywhere these days.
It's Netflix recommending movies as we just said, chat GPT, helping you draft an email, maps in terms of car directions in health care and sort of in AI is obviously bigger. It's changing how we diagnose diseases, manage patients conduct research with the talk about. So there's three basic types of AI. There is machine learning, which is one.
There's deep learning and natural language processing. And there's good examples for all three of them. I'll go through them quickly. So we just are everyone's at a similar starting point.
So ML or machine learning is the foundation. It's sort of what we traditionally think about when we think about AI. It's about recognizing patterns and data and making predictions. So okay, a good example for I can imagine an AI model and these already exist that predict a patient's risk of say developing glaucoma or macular generation based off by an exact visual field, I op, OCT scans.
These learn from massive massive data sets and they help doctors flag these highest patients long before damage ever happened. So that's the promise of the first type of AI, which is called machine learning. Now the second one is deep learning and that takes steps sort of step further, hence it's called deep. So this is where AI shines in image analysis.
This is where you think like radiologists, you know, they're going to lose their jobs or retina specials are just going to turn off the brains when they look at retina scans. That's what you know, those are tongue and cheek that's not going to happen. But it's about the AI assisting you in recognizing patterns and using its deep learning models to sort of mimic how the human brain thinks and sort of read your images, grading, not the and detecting signs of a macular generation, for example. So good examples here.
This is like the common stuff and diabetic pornography with the IDXDR, the LumaDex core, the AI health of the world's Toku's model. This is all based off deep learning AI. So the last one is I think probably one where we're talking about what we're most familiar with in terms of chat, you know, it's natural language processing and this focus on understanding and analyzing text is kind of like the next word auto correct, but on steroids times a ton. So this is chat GPT and this is NLP in action.
So an ophthalmology and healthcare in general, it's sifting through thousands of notes, records, research papers to find insights. It's about helping you document your notes, scribing, doing patient charting and identifying patients in clinical trials. Yeah, so Scott, that was the three types of AI and looking forward to talking about the topic next week. All right, well, we promise we're going to try to keep this in a short amount of time.
So this was really good, Rayana. I really enjoyed it. I think that, you know, I think that a offer is going to offer as tremendous potential to really kind of change the way research works and I think machine learning as you addressed in the types of AI. That's going to be really pretty essential for implementing these AI and clinical trials.
So this will be a fun thing for us to keep tabs on and address some of the news that comes out about how AI has changed in research. Sounds good. Next week we'll be talking about what's an LN and some of the basics of AI and diabetic Renapathy. Stay tuned and thank you for listening.
You've been listening to Real Talk, an AI and I care. Your weekly podcast to keep you informed about AI technologies revolutionizing I care.