Hi, everyone. Great to be here. My name is Dr. Masoud Nafey, and I'm here to talk to you a little bit about the most common AI models or large language models. These are platforms that you can use either for free or use the pro versions for roughly $20 a month. That’s pretty much the standard fee for all these models.
With the pro versions, you will get faster processing power, the latest and greatest models, and more functionality. You'll have more features available to you, and you'll be able to use the model’s computational power more. You’ll be able to process more data, and you won't get that error message saying you ran out of usage. So all of these that I'm going to speak about have a free version you can use, but the pro versions with those additional features are about $20 a month.
We'll start with a very popular model created by OpenAI, which is ChatGPT. Overall, I'll say this is the best assistant, if you will. Universally and overall, it's one of those models that can do a lot of things pretty well. Lately, they released an image management or image generation tool, and the image generation tool is really great for patient education graphics, marketing visuals for your website or social, presentation slides, logos. It’s wonderful at that. Honestly, before they came out with this image generation tool, Google Gemini was whooping everybody's butt in this class, so it's fantastic that OpenAI came out with this, and it's honestly really powerful.
OpenAI is also really good with uploaded PDFs and screenshots. You can upload patient information, like OCT images and those types of things. Of course, make sure that PHI is blacked-out. You don't want to share any of that with any of these large language models.
It has great voice interaction, so you can speak to it instead of type. And of course, it's really good at creating workflows for you too. An example would be, "Help me create a script explaining myopia management to the parent of a child." It's really good at creating that in a way that can resonate really well with the parent. So patient communication, workflow support, and all around productivity is what ChatGPT is the best at, and lately image generation is one of those categories they're doing really well in.
As far as its limitations, I would say it can hallucinate a little more than some of the other models and sound overly confident, both generally and clinically, which is why they created ChatGPT Health, and I expect we’re going to see that start to take over. Apparently 250 million people search health conditions on a weekly basis on ChatGPT, which is incredible.
So the takeaway is, if you adopt only one AI platform today, ChatGPT is likely the easiest and most versatile starting point.
Read more in my full column, "A Practical Guide to Choosing an AI Partner" from the May/June issue of Optometric Management.


