Saahil: I’ve always wanted to pursue medicine; when I was a kid, I would say that I wanted to be a doctor when I grew up. I loved that medicine combines intellectual rigor with human connection. During clinical rotations in medical school, I’ve been motivated by the privilege of developing my own relationships with patients and being part of their lives at critical moments. At the same time, my research keeps me motivated, showing me how innovation is driving medicine forward.

Chanseo: My journey toward becoming a physician reflects a dissonance of the tension between two powerful influences in my life: my training in science and “Big Data”-driven research, and my personal experience of how illness ripples beyond just the patient, but to those around them. On one hand, my academic path has been grounded in translating biology and chemistry into novel technologies. On the other, my grandmother’s struggle with Alzheimer’s and its rippling effects on my family connected the science to the human phenomenon. Bridging these two perspectives, scientific innovation and human experience, has shaped my commitment to medicine.

Diego: I was a philosophical teen, so I aimed my sights at consciousness early on, which I thought was the hardest unsolved problem. That led me to an undergrad in neuro, but epigenetics really widened my interests out towards tissues and systems: questions on how living matter senses, computes, heals… and how it fails too, giving us things like fibrosis, cancer, and even death. I realized that solving the hardest problems might exceed one lifetime, so the most pressing issue seemed to be time. And, I learned that not every patient gets to make peace with death; I thought if we could postpone our tissues failing, or even in our wilder fantasies, make death optional, then many people would choose more time with their loved ones, more time to be healthy, and more time to live out bigger projects and dreams. I want to build tools that give people more of that sort of optionality. The problems in medicine are hard, but the possibilities and how cool the science is? That keeps me incredibly optimistic and endlessly motivated.

Mekala: Honestly, I had no clear idea what I wanted to do going into undergrad, and getting to where I am now was a very gradual process. I chose finance because it’s what my sister had studied, and I signed up for premed courses because they interested me. I took those interests and worked backwards, figuring it might be neat to work at the intersection of them. Even still, it took a few years of working to go “all in” on medicine rather than pursuing investing or policy. What finally pulled me in was the day-to-day work of caring for patients—something you only fully appreciate after doing a non-clinical job.

Saahil: AI has the potential to make diagnosis not only faster and more accurate but also more personalized. In oncology, the number of therapeutic options continues to grow, which creates both opportunity and complexity. I see AI as an essential tool for synthesizing vast amounts of imaging, molecular, and clinical data, helping physicians identify patterns that aren’t visible to the human eye and guiding them toward the most effective treatment strategies. Rather than replacing clinical judgment, AI can serve as an intelligent partner, allowing doctors to focus on the human side of care.

Chanseo: Even after years of the “AI bubble,” artificial intelligence still boils down to extremely large neural networks and transformer architectures. Large language models and their derivatives, whether small language models, AI agents, or specialized systems, are essentially extensions of these foundations, optimized for domain-specific tasks, cost-efficiency, autonomous systems, and performance. Because LLMs are fundamentally trained to “predict the next word” across vast datasets, they are remarkably well-suited to identifying patterns across the broad spectrum of disease phenotypes. This makes them especially powerful in fields like oncology and imaging, where prior deep learning advances like CNNs extract interpretable features for LLMs to generate clinical insights, especially in the context of patient data from EHRs.

While these models excel at statistical learning, current medical applications still lack true reasoning and the uniquely human capacity for intuition and insight. These qualities remain central to medicine, not only in patient care but also in driving research and innovation. For now, AI serves as a transformative augmentation: an indispensable tool that enhances precision care and expands physicians’ capabilities, but not a replacement for the clinician’s judgment in diagnosis and treatment. Finally, LLMs are not the central, “end-all-be-all” artificial intelligence the public makes it to be. Investments must also be heavily diverted to broadening and deepening the science in other sectors of machine intelligence, including frontiers in computer vision.

Diego: AI will probably fuse radiology, pathology, and -omics into some sort of patient-specific risk and response models. More compressed, usable information means earlier detection, less overtreatment, and smarter trials. I think future treatments won’t just be “one drug, one target,” but something more like coordinated cocktails and perturbations that shift whole tissues into where they need to be, optimized by continuous data tracking. There’s sooooo much ambient data as well, it makes me think the invasiveness of diagnostics and treatments will decrease dramatically over the coming decades.

Mekala: AI is going to transform how we approach disease. It will accelerate research, shorten timelines for innovation, and enable earlier diagnosis. With faster drug development and more precise genomic tools, I think we’ll finally be able to deliver personalized treatment at scale.

Saahil: I began medical school in Fall 2022, just as ChatGPT was released, and it has already completely transformed how medical students learn. Rather than relying solely on memorization, AI can now summarize information, generate practice questions, and simulate clinical reasoning. I would redesign the curriculum to include training in AI, data literacy, and digital tools, while emphasizing the limitations of AI, including issues with generalizability and hallucinations. In my role as Associate Course Director of the Clinical Reasoning course, I am working with Drs. Thilan Wijesekera and Jaideep Talwalkar to integrate these concepts into Yale’s curriculum.

Chanseo: One of the most striking gaps in medical education is how little formal, centralized training exists at the intersection of medicine, data, and technology (of any kind, not just AI). Physicians are expected to navigate electronic health records, use new digital health tools, and interpret data-driven decision support systems, yet few curricula offer structured training in any IT or data vertical. Furthermore, medical students and residents are dropped head-first into varying levels of technical adoption, from pre-2000 MS-DOS legacy systems at Veterans Affairs to impossible integrations of digital health tools. We have created a disconnect — we are preparing students for 20th-century medicine while practicing in a rapidly evolving 21st-century healthcare system.

If I could redesign the curriculum, I would integrate three key elements. First, data literacy and AI fluency: not to make every physician a programmer, but to help them critically evaluate algorithms, understand bias, and responsible tool use. Second, interdisciplinary collaboration: creating opportunities for medical students to work directly with engineers, computer scientists, and policy experts on real-world healthcare problems. Finally, human-centered care and communication: ensuring that while technology grows in importance, students also train deeply in empathy, cultural humility, and the relational aspects of care that no algorithm can replace.

Diego: I think there’s not enough retrieval practice in classrooms. Not enough personalization too, but that’s harder to solve. We know that group teaching is about two standards deviations worse on average that personalized tutoring. The dream would be that every student gets a tailored tutor, who somehow encourages them to drill with a spaced repetition system and perfect scaffolding. AI’s will help with that, but in 2025, we’re still a ways away from the chatbots doing this all for us. That doesn’t stop students from trying to jerry-rig this, though (I try to get LLMs to quiz me using voice mode while I walk between buildings). I’d probably look to “Math Academy” as the best example of a personalized training program that scaffolds students well. For now, I’d just encourage the professors to constantly be drilling the students live on things they know, showing them examples when they get things wrong, and interleaving spaced repetition of topics, rather than doing long lectures.

Mekala: Many med schools talk about AI as a priority, but very few actually teach students how to use it. It’s not enough to describe what AI is—we need to show how it fits into our daily work. Even something as simple as using AI to study more efficiently could be transformative. Once you learn to wield the hammer, you start seeing all the nails.

Saahil: I see prevention as central to medicine. Whether through counseling patients, advocating for healthier systems, or using data to identify risks early, physicians can help shift the focus from reacting to disease to preventing it.

Chanseo: Our current healthcare system is still largely oriented toward reactive rather than preventive medicine. Yet in an era of big data, digital twins, and advanced population science, we know that most diseases are not random events. Many have identifiable, intervenable causes, or at the very least, early warning signs that can be detected long before clinical symptoms appear. As both a future physician and a data scientist, I see my role as bridging these two worlds: using insights from data and risk modeling to identify vulnerabilities earlier, while also educating and empowering patients to take preventive steps. By shifting the focus upstream, we can reduce disease burden, ease strain on physicians, and create a healthcare system that emphasizes maintaining health rather than simply responding to illness.

Diego: I think a lot of prevention often feels like the burden is being thrown back on patients. Getting 10k steps, eating healthy, sleeping 8 hours, etc. Going in for screenings and checkups. I think as medicine evolves, we’re hopefully going to take a more passive role in the lives of patients, using ambient data to detect risk factors earlier, and pinging patients with easier and easier fixes. I think if we can take an example from GLP1 drugs, we can probably find ways to make it easier for people to self-actualize the ways they want and help make the healthy choice the default. Even with things as hard to crack as say social media addiction. Refined therapeutics will probably erase all the stigma of “better living through chemistry,” so I think physicians will get better at finding elegant solutions without damaging side effects, and we’ll find ways to redefine prevention into something like context engineering for people.

Mekala: I’m going to go against the grain here and say prevention is always going to be an uphill battle. Just like the healthcare system, people are resource-constrained. They aren’t going to devote effort to things until they become definitive problems. I think the most pragmatic way to bolster prevention is to invent better predictive tools, allowing providers to put very specific risks on patients’ radars early on so they can act accordingly.

Saahil: AI should enhance, not replace, the human bond. Doctors will still be the interpreters, advocates, and listeners, ensuring that technology serves patients’ values rather than overshadowing them.

Chanseo: AI and Big Data tools are essentially another partner in the room with the doctor and the patient. It provides the scale, pattern recognition, and predictive insight, while the physician offers judgment, empathy, and narrative understanding. AI should expand the time physicians can spend listening, counseling, and building trust. If used well, it can restore some of what has been lost in modern medicine: the ability to focus less on the screen, and more on the human being across from us.

Diego: I used to think doctors would always be there to help translate things to patients, but honestly, at a certain point, the chatbots will probably do a lot of that for us too. I think our role is probably going to shift to that of guiding the value system of medicine, and of helping to promote trust in the treatments. Until you can trust AI’s with your life, there’s going to be a team of professional humans vetting the pipeline and all the treatments. A lot might become more automated, but people in the loop will help it become less of a black box, and patients will feel better if the transition is stamped with a seal of approval that the medical institutions are certifying all these new tools and techniques. Hopefully, visits will be shorter, but I’m also thinking that trust, transparency, and outcomes will get stronger too as we get better at implementing.

Mekala: While people often view AI as a “dehumanizing” force, it might be quite the opposite. With AI and big data, patient care will become more tailored to the individual and, thus, personal. Patients with the same issue may be given different treatment plans and providers will need to explain the rationale. Additionally, AI will greatly reduce administrative burden—I hope that this will allow providers to look away from their screens and devote their attention to the person in front of them, as once used to be the case.

Saahil: Invest equally in technology and people. The tools are advancing quickly, but the future of medicine depends on training clinicians who can use them wisely and keep compassion at the center of care.

Chanseo: Prepare for a future where medicine is inseparable from data and machine intelligence. The next generation of care will be defined by technology, yet these tools will only succeed if they are integrated thoughtfully, equitably, and with human-centered design. Technology has rarely been impeded by fear and gatekeeping of its potential. This has only led to irresponsible practices and poor adoption, a process that takes exponentially more time to reverse. Instead, I’d ask our leaders to centralize and expand education at an even more rapid rate, equipping physicians with tools that allow them to be more present, more precise, and ultimately more human in the care they deliver.

Diego: Learn the new tools and learn fast. We’ll help more patients that way. There used to be a time when we could get stuck in our ways and our elders were fine being out of touch. Change is coming faster, and there’s no way to just opt out. I think it’ll be a great wonder to see the new therapeutics and tools used for good, I do think a rising tide raises all boats. But still, make sure you’re building yourself a sturdy boat with a good foundation.

Mekala: One quote I heard at Oxford that really stuck was, “AI won’t replace doctors, but doctors who use AI will replace doctors who don’t.” Fighting progress is never wise. When Excel first appeared, people in finance thought it was the end of the world; instead, it became a tool to make their work better. AI will do the same in medicine—it will raise expectations of care, shift some duties, and enhance our role rather than erase it.