How do we look at healthcare with AGI?
Humans have evolved much more than any other living being. We also have reshaped the earth more than anyone (except nature itself).
Dinosaurs lived for more than 165 million years, went extinct due to ecological changes on Earth, and had much less impact on Earth. Humans have only been here for 300,000 years. Dinosaurs were more powerful, and so were many other animals. Chimpanzees, Dolphins, and Elephants have problem-solving skills and are somewhat comparable to humans in terms of intelligence. But no one else has had a comparable impact on Earth.
So, humans are not as powerful as some other animals. We are more intelligent, which means we understand the things around us much more than anyone. We have built more things than anyone on any timescale known and have a much higher ecological impact on Earth.
The first thing that is common among all animals is survival. The second is the convenience of living. Both are correlated. The idea is to make survival more accessible; hence, life becomes more convenient, and we can look for more entertainment. We can think of basic survival as a balanced food chain for animals or for humans; we can think of having enough fundamental needs like good food, good physical and mental health, a place to live, or having enough money. We can think of the convenience of living or entertainment in terms of eating what we want, living where we want, having a healthy life, knowing what we want, etc.
The current global average life expectancy is around 72 years, which was around 30-40 years a thousand years ago. Especially in the 20th and 21st centuries, it increased at a rate of a few years every decade. Better medical science, nutrition, sanitation, vaccinations, focus on fitness, and fewer wars have played a huge role.
AGI in healthcare will be critical for our better survival and also for our convenience of living. Just imagine knowing about 100x more diseases than we know now, discovering vaccines and medicines much more quickly at 1000x lower cost, knowing our genetics better, eating what we like, and still being fine. Getting the right treatment within 30 minutes compared to what takes months right now.
It feels like it will be a better place to live. We will have a higher life expectancy, more freedom, and an easier life for everyone.
But this will be possible if the AGI that we will use in the real world is safe and accurate. And it has to be much more accurate than it is currently with real doctors.
We will have to build a powerful AGI that is super-aligned with humans and the well-being of all living beings. We will have to build strong safety models like healthcare LLMs trained on safety data at scale. We will have to unlock more modalities to interact with this AGI, like making better devices than current X-ray, MRI, ECG, Pathology, Ultrasound, and other machines that can interact with people more easily and in real-time. We can think of wearable smartwatches with sensors and CGM devices as starters. But we will have to go beyond these. We need better hardware systems that can give a complete view of our health in more detail and real-time. Real-time X-ray or MRI-level data with a wearable. Real-time blood tests with a CGM-like device.
It will take time. Even with a superintelligent system around, it will take a few years to make such a system. It will take time for the government to approve such a system. The Drug and Cosmetic Act of 1940 is 80 years old and is yet to be updated in India as of December 2023. The laws around online pharmacies in India are also unclear and unsafe. We have seen customer support people in big online pharmacies generating fake prescriptions just to sell more prescription medicines. We have also seen unicorn surgery tech companies forcing people to get needless surgeries.
Such powerful technologies should be super aligned, rightly regulated, deeply discussed, and thought of.
The current state of AI can read medical documents in the form of PDFs, images, and well-written handwritten prescriptions. It currently cannot read radiology images well, as far as I have tested on a few cases. It can do some basic reasoning and works really well on diet plans, nutrition support, information on medicines, and basic medical issues. These capabilities themselves can add a lot of value to people’s lives. It can solve many things internally or as an assistant to a doctor, pharmacist, or patient directly in many cases.
Ultimately, it always comes down to building a great product that is more powerful, faster, and cheaper than the current systems and adds great value to your users. I believe it will be a matter of building better safety models in healthcare. You can think of a safety model for healthcare as any system that forces our AGI/AI to work 100% accurately in the defined environment on a set of use cases.
It will be solved with better-fine-tuned LLMs working as safety models, UI-UX, better hardware systems, better regulations, and much more.
In the end, it's about enabling every human to live a better life, which is the core idea of improving healthcare.