Speaker 1: After sifting through hundreds of studies, reports, announcements, and books about artificial intelligence and generative AI's role in the near future of medicine and healthcare, I thought I would boil all those down for you into a list of 10 predictions I stand by. I hope that these will serve as a guiding light for you, so you would be in a better position to understand where we are heading when it comes to using this amazing breakthrough technology in medicine and healthcare. Let's jump right in. The future of large-language models is multi-modal. JGPT-like tools are amazing at providing us with the answers to the questions we ask through prompts or briefs, but these large-language models are usually unimodal, meaning they can only deal with text. Some of the newer versions, like JGPT-4, can now deal with images, but the next iterations will be able to deal with video, sound, and even full documents. You upload a PDF document, and you can ask questions about the content of it. So the large-language model's future is multi-modal, meaning these MLLMs will deal with all these modalities – text, image, video, sound, whatever kind of formats you want to upload, you will be able to get the answers you're looking for. Do you know why it's important? Because these MLLMs will be the ultimate interface between physicians using a myriad of different types of AI-based technologies. Data annotators will be celebrated. There are many physicians right now who are sitting in dark rooms manually annotating medical images, medical reports, electronic medical records. Without them, there is no AI revolution. Because AI models need a lot of data, especially annotated datasets, to improve. The more higher-quality data they have, the better predictions these models can create. So not only we should celebrate data annotators, but I think they will be cherished and rewarded accordingly. AI will not replace physicians. For many years, I'm sure you've heard the reports and investors in interviews saying that this or that medical specialty will be vanished, and that AI will replace medical professionals. That's simply not and cannot be the case. Not only patients need empathy from people they trust – and that's not technologies, but people, human beings, physicians. But there are so many issues and tasks and challenges in a medical professional's job that simply cannot be automated. Plus, the whole point of using AI in medicine is to contribute to the process of medical decision-making. So AI will not replace medical professionals. However, I still think that those medical professionals that use AI will replace those that do not. AI will take over repetitive and database tasks. These are the tasks that represent the low-hanging fruit for the whole healthcare industry. If a task is repetitive and or database, especially when combined, then that task should not be performed by a medical professional who has been training for years, if not decades. Their human judgment, their creativity, their intuition would not be needed to perform that task. But AI is amazing at doing the same. So all these repetitive and database tasks, especially thinking about administration here, could and should be automated. That's the holy grail of the whole AI revolution in medicine. AI will find unusual associations. There have been studies showing that artificial intelligence-based models and technologies can indeed find biomarkers or results in a study that no clinician, researcher, or physician has been thinking of. These are amazing things because they can find associations between two endpoints that we thought before to be unrelated, such as analyzing the phone recordings of patients and then coming up with a vocal biomarker predicting their risk for Alzheimer's disease. We have never done it before, but the AI started finding these unusual associations, especially in the case of unsupervised learning models. You will need a common language with AI to understand its progress. And I'm not talking about learning how to code. Coding is the language of developers that make AI possible. But you need a common language, such as a board game or chess or Go, something through which you would understand the constant progress AI has been going through. For me, it turned out to be chess. By dedicating thousands of hours to chess over the last few years, I think I've been able to better understand what an amazing progress AI has experienced in those last few years. The more I understand about the intricacies, the tiny nuances of the game, I think the better I see how far away the best chess algorithm is from the world's absolutely best player, Ivan Magnus Carlsen. So developing a common language with AI puts you into a better position to understand where it's heading. Prompt engineering is the number one skill in the generative AI era. Chances are that you have come across generative AI already by using a large-language model. And as large-language models like JGPT will become multimodal, dealing with image, sound, video, and any kind of data, you will have to interact with these AI-based technologies through the interface of a large-language model. Thus, the number one emerging skill for you now is prompt engineering, the ability to design the absolute best text-based briefs or prompts to get the best results, the best outcome you were looking for while using these large-language models. We need proper guidelines from medical associations about fighting health equity and bias. AI cannot be biased by themselves. They can be biased because of the database being biased that they were trained on. Therefore, this must be the conscious preparation and decision of developers to make sure that the database they use to train the AI on is not biased. A good example for that is around skin-checking applications. It turned out many studies have shown that people with darker skin had worse results from the AI-based algorithms, simply because the anatomical databases didn't contain enough photos and images of skin lesions of people with darker skin. The database was biased, so of course the AI started making biased decisions. But again, we can't fight against that. New regulations will be needed. Generative AI will have to get its own regulatory category. There is one more challenge here that regulators around the world must face now. It's about adaptive algorithms. Algorithms that will change with every decision they make. So imagine that there is a medical technology that regulators regulate and adopt, allow to be rolled out to the market, but that technology will be a million times different from the next day. Because that's how fast the algorithm can change. So dealing with adaptive algorithms and creating a new regulatory category for generative AI, especially medically-trained large-language models, will be the absolute priority for regulators for the coming years. Medicine and healthcare will struggle with deepfakes. Imagine reaching out to use a healthcare service virtually and having a remote consultation with a physician. But that physician is not standing or sitting behind a computer. There's actually no physician behind a computer. The entity you're having a video discussion with is a deepfake made by AI, and the discussions you are having with that entity is being fueled through a large-language model like ChatGPT. And to find out whether it's a deepfake or not, through a chat-based service, a chatbot, or through a video consultation, it's going to be a bigger and bigger challenge for patients, while healthcare providers will try to fill in the gaps left by doctor shortages worldwide. I hope you found these predictions useful, and as this is an extremely exciting yet dynamically changing world, I will keep on having a finger on the pulse of AI and generative AI. So if any new thing comes up, you can make sure that I will cover those in the upcoming videos. Cheers. If you liked this video, please subscribe below to get all the videos about the future of medicine, healthcare, and advanced technologies. Also, please check out medicalfuturist.thinkific.com to access our courses on digital health and AI's role in the future of healthcare. See you there.