Supercharged Data

While decades of major advances—from antibiotics and anesthesia to genome sequencing and immunotherapy—have transformed the practice of medicine, the field is now at the threshold of yet another seismic shift. Today, generative AI, a new approach to artificial intelligence, has the potential to reshape clinical practice, research, and education in ways that defy imagination.

It was 75 years ago that the computer scientist Alan Turing posed a profound and prescient question, “Can machines think?” in his seminal research paper Computing Machinery and Intelligence. A few years later, in 1956, a Dartmouth professor coined the term artificial intelligence to describe a burgeoning field of research.

Driven by algorithms, data, and computational power, AI technologies mimic processes of the human brain with breathtaking speed. Generative AI, which became widely adopted just over two years ago, differs from classical AI by creating new information or content based on large computing models trained by using machine learning techniques on text, images, videos, sounds, and software code often gathered from the internet. With such popular applications as ChatGPT, DALL-E, and Perplexity, generative AI has quickly become a driver of innovation and economic activity.

Generative AI harnesses clusters of high-performance computers in data centers; it uses new machine learning techniques to build vast models based on real-world scenarios to do everything from writing software code to forecasting the effects of climate change. Observers have said that society is experiencing a new industrial revolution—this one powered not by coal or oil but by data. Generative AI has the potential to supercharge data for human progress at an unprecedented rate.

Within the medical domain, generative AI shows great promise in relieving a highly stressed health care system. The technologies can help health care providers become more effective and efficient at a time when high-quality health care is expensive and inaccessible to many, and where there aren’t enough physicians to go around.

“I’m very optimistic that AI—including generative AI—will help address the big challenges faster and propose solutions,” says Lucila Ohno-Machado, MD, PhD, MBA, Waldemar von Zedtwitz Professor of Medicine and Biomedical Informatics and Data Science. She chairs the Department of Biomedical Informatics and Data Science (BIDS), which is spearheading the use of AI in research and clinical practice.

Yale School of Medicine (YSM) is known for being an early adopter of information technologies, and the creation of BIDS as a department in May 2024 signals the importance of AI to the school’s future. The department has already worked with the university to establish a computing environment that’s compliant with federal privacy and security standards, so that faculty members can develop AI models using real patient data while adhering to the highest ethical standards to protect confidentiality and trust.

On an even broader scale, Yale University has committed to spending $150 million over the next five years to support its AI ambitions. It is hiring faculty, adopting new AI applications, and, in a sustainability move, has joined other universities, including Harvard and the Massachusetts Institute of Technology, at the Massachusetts Green High Performance Computing Center (MGHPCC) in Holyoke, Mass. Yale’s project leaders expect to transition at least one megawatt of on-campus power to green hydroelectric sources, and eventually to operate dozens of racks of computing hardware there—many of which will be dedicated to YSM, says Wies Rafi, PhD, associate chief information officer of Yale’s Health Sciences Division.

Meanwhile, more than a dozen YSM faculty members assisted the Yale Task Force on Artificial Intelligence in publishing a 119-page report in June 2024 laying out a strategy to leverage the new technologies. In the report, YSM faculty members wrote that AI “is already driving a revolution in health care and is poised for mass adoption,” further predicting that “it is only a matter of time before nearly all health care workers engage with AI solutions on a daily basis.”

For decades, YSM has fostered the development and adoption of AI technologies. For example, Perry Miller, MD, PhD, professor emeritus of biomedical informatics and data science, developed PUFF, an AI program that interpreted lung function data and was used at Yale New Haven Hospital beginning in the 1980s.

Fast-forwarding to 2023, Rohan Khera, MD, MS, assistant professor of medicine (cardiovascular medicine) and of biostatistics (health informatics), has used novel deep learning techniques to develop a smartphone app that analyzes electrocardiogram images. The app enables cardiologists to screen patients for left ventricular systolic dysfunction and other conditions that, if undiscovered, can result in hospitalizations and premature deaths.

Within health care, AI has so far had its most significant impact on diagnostic radiology. Pattern recognition technology is used widely to prescreen medical images and spot diseased tissues, bone breaks, and other conditions. Now, researchers across the country experimenting with generative AI are creating highly detailed 3D models of organs and tissues that can be trained to simulate the effects of radiation therapy on a particular patient, allowing physicians to customize treatment plans.

Innovation is accelerating across the medical school with the emergence of generative AI. One project that could impact researchers and clinicians in practically every department is the Me-LLaMA chatbot. Working with associates, Hua Xu, PhD, Robert T. McCluskey Professor of Biomedical Informatics and Data Science, built a medicine-specific chatbot on top of Meta’s LLaMA large-language model, training and fine-tuning it on biomedical literature and clinical notes from electronic health records. One of the first applications of Me-LLaMA is to suggest diagnoses in complex clinical cases.

Yale New Haven Health System, which is staffed with YSM faculty members, is among those already using generative AI technologies to ease administrative burdens on clinicians. With the use of a new generative AI application called Abridge, AI agents can “listen in” on office visits and create summaries, enabling doctors and nurses to focus on human interactions. Other technologies suggest responses to emails from patients, thus saving human time and effort.

In an environment in which delays can be devastating, clinicians at Yale New Haven Health’s three emergency departments are utilizing Abridge to capture conversations with patients and automatically integrate that information into their existing electronic medical records—enabling a physician to better evaluate the patient’s condition. The next version of the application is expected to provide rapid diagnostic help. “We need these tools to help us stay afloat. They can rapidly scour information across multiple records and help us make sounder clinical decisions,” says Arjun Venkatesh, MD, MBA, MHS, professor and chair of Emergency Medicine.

For example, Venkatesh points to the complete blood count (CBC), a blood test that is frequently used in emergency departments. Today, physicians typically evaluate just a handful of perhaps 20 lab values produced by the test. Venkatesh anticipates that new generative AI applications will be able to quickly review all the lab values obtained, and identify relationships and implications that a harried ED physician simply does not have time or the mental bandwidth to evaluate.

On the research front, YSM faculty members expect generative AI to accelerate the development of new therapies, and to deepen our understanding of how diseases work and what it takes to cure or control them—delivering on the promise of truly personalized medicine. For instance, María Rodríguez Martínez, PhD, associate professor of biomedical informatics and data science, has been using AI models to understand immune system function in such complex diseases as cancer and autoimmune diseases.

Steven Reilly, PhD, assistant professor of genetics, is working with scientists from other research organizations to use generative AI to design synthetic DNA elements that switch on genes in certain cell types but not in others. The technology could help develop the next generation of genetic therapies, specifically targeting gene replacements or CRISPR sequences to affect only diseased tissues.

For many YSM students, generative AI is already part of their lexicon. In the first-year Professional Responsibility course last fall, about one-third of the students indicated that they already had experience with AI, according to David Rosenthal, MD, assistant professor of medicine (general medicine) and co-director of the course. The course now includes lectures on AI concepts and medical applications in addition to discussions of ethical issues.

Technology leaders expect advances in AI to pick up speed in the coming years, eventually producing machines that are sentient and autonomous. At that point, Rosenthal believes medical schools will need to rethink medical education. “We will have tools that outperform the best humans on knowledge tasks,” he says. “We need to prepare so physicians will be able to use these tools to help patients heal, to promote healthy behaviors, and to help patients live well.”

While AI’s potential has not yet been fully realized, YSM leaders say the future is bright. Yale is well positioned to conduct groundbreaking research at the intersections of biology, medicine, and AI, and to train the next generation of clinicians and biomedical researchers to use the technology responsibly and enhance it to promote better health for all. 

Editor’s note: Because OpenAI is rapidly evolving its GPTs, various models reflecting particular time frames are referenced in the special report articles.