In the annual AI index report published by the Human-Centered Artificial Intelligence Institute (HAI), experts detailed the trends in artificial intelligence over the past year and how it is transforming society. This year, the report features an expanded chapter on science and medicine developed by the team at Raise Health, a collaboration between Stanford Medicine and High.
Russ Altman, MD, PhD, Professor of Bioengineering in Genetics and Biomedical Data Science, led the development of science and medicine chapters highlighting AI milestones, including advances in protein and molecular design, support for clinical care, and automated disease detection.
Altman discussed the top takeout of the year, how AI trends shape the future of biomedicine and medicine, and what he considers the most promising growth field.
The science and medicine section of this report has expanded significantly over the past few years. How does this reflect what you’re seeing regarding the AI explosion?
Last year’s chapter did an incredible job highlighting some important examples of AI in science and medicine. There is something called a system review when you meet a patient. Complete detailed examinations are performed to avoid evaluating the skin, eyes, heart, lungs, etc. I think this year’s chapter will go from a biopsy to a small place here or there and then to a review of the system.
Everyone in medical school knows that AI is everywhere. You may be having lunch with someone who has never hit you as an AI researcher, and you will find that they are leading efforts to build a big language model in their clinical practice. I’m hosting a podcast, The Future of Everything, and I see the same trend there. I interview faculty members from all over the university. The most common thing they say to me just before I press the record is, “Ask me how AI is revolutionizing my work.” It really revolutionizes university life and scholarships, and it’s pretty exciting.
The Science and Medicine chapters touch on a range of trends, from how AI influences clinical care to ethical considerations surrounding AI use and development. What was your three biggest takeaway?
The first is creating a basic model. A basic model is basically a statistical model that describes a very large data set. About 10-15 years ago, we all talked about “big data.” People were collecting huge amounts of data, but they didn’t always know what to do with it. Scientists use data pointing to clean and tidy learning to become cherry picking. With the basic model, you can see all the cherries, not just ripe, perfect and eye-level. The basic model takes all the data from a large dataset and allows it to be predicted and predicted towards a rich statistical model. That’s why much of science sees such rapid advances. Because scientists essentially have models that allow them to speak to their data, ask questions and get answers.
Second, AI research directly contributed to two Nobel Prizes. It’s a real bet on the ground. Yes, there is hype and there are questions about whether AI is good or bad for society. For science, that’s good and I can’t think of a better short answer validation than “two Nobel Prizes awarded for AI technology in the same year.” That’s the headline I say to my mothers and children: “Yes, this is true.”
Third, the ability to improve all parts of clinical care using large language models is enormous. Many of my clinical colleagues who are on trench every day are interested in integrating AI into their daily workstream, such as using LLM to help write notes, listening to and watching surgeries, and receive quality summary of what happened in the operating room. Large language models can reduce what is called “pajama time.” This is the time that doctors spend after the clinic closes and keeps up with all the paperwork. It can have a major negative impact on quality of life.
Where do you think there are possibilities for AI next year?
The ability to deliver messages at different educational levels of large language models, or with nuances of different cultural backgrounds, is a major untapped opportunity. Language models allow information to be distilled to help patients understand their illness and treatment plans. AI can suggest ways to effectively communicate or provide a variety of perspectives that doctors may not have thought of.
For example, someone might dislike taking medication. The chatbot tells the doctor, “You can hear your voice. You understand that you’re not a pill person. There are five medications that can be prescribed, but they’re only given two because they’re the most important.” They are optimistic that improved communication and clarity will lead to improved patient understanding of illness, and thus improved doctor-patient treatment alliances.
This story was first reported by Stanford Medicine.
