Will Shapiro, vice president of data science at Flatiron Health, discusses opportunities to implement artificial intelligence in cancer care.
There is a huge opportunity for artificial intelligence (AI) and machine learning to revolutionize oncology care, says Will Shapiro, vice president of data science at Flatiron Health.
These findings were presented at the Association of Cancer Care Center's 50th Annual Meeting & Cancer Center Business Summit (ACCC AMCCBS).
Transcript
Can you elaborate on your AMCCBS presentation about building language models and their impact on cancer care?
I kind of want to give an overview of a lot of the terms that I think get thrown around pretty liberally these days—deep learning, natural language processing, artificial intelligence, and machine learning—to talk about how they relate to each other and give people a bit of a grounding on what they actually mean, because there's so much hype, especially around generative AI and artificial intelligence. I wanted to try to give people an intuitive understanding of what some of those terms mean. And then also, where the opportunity is for artificial intelligence and machine learning to truly revolutionize oncology and cancer care, which is something I'm incredibly excited about.
What specific aspects of AI or business intelligence (BI) did your presentation cover?
At Flatiron I kind of work across both [AI and BI]. My team comprises product data science, data reporting, business intelligence, workflow, automation, machine learning, and artificial intelligence. But [at AMCCBS], I really focused more on machine learning in artificial intelligence, in particular, and applications in oncology. One of the grounding things that I started with is the fact that 1 in 10 doctors are already using ChatGPT in their practice, and that's a significant fraction.
So, I think, just kind of understanding a bit more about how GPT works. I kind of broke down the acronym GPT, which is something that a lot of people don't necessarily know about, so just talked about how “G” is for generative and what is generative AI. It's something that people actually use every day without really even thinking about it. Autocomplete is an example of generative AI. And then pretrained models is what the “P” stands for. Pretrained models are really exciting, because they can do a lot of different things. They can answer a question, they can flag content that might be problematic, and they can write the next Harry Potter novel.
And then of course, they can also be used for things in oncology. I think one of the big appeals of this new generation of models is they can do so many different things. But with that generality comes tradeoffs, too. My team has found that often purpose-built models that just do one task, like read through a chart and figure out whether a patient has metastatic disease or not, often outperform these general pretrained models. So, that’s one of the things that I touched on. And then finally just to wrap it up, the “T” in GPT is the transformer architecture, which is something that was developed at Google in 2017, and I talked a bit about the transformer architecture.
Frameworks for Advancing Health Equity: Urban Health Outreach
May 9th 2024In the series debut episode of "Frameworks for Advancing Health Equity," Mary Sligh, CRNP, and Chelsea Chappars, of Allegheny Health Network, explain how the Urban Health Outreach program aims to improve health equity for individuals experiencing homelessness.
Listen
Alopecia Areata Risk Associated With Atopy, Allergies
May 13th 2024A genetic study suggests people with allergies and eczema are more likely to develop alopecia areata, an autoimmune disease that causes hair loss. Researchers believe this link may be due to similar immune system responses.
Read More
Tackling Health Inequality: The Power of Education and Experience
April 30th 2024To help celebrate and recognize National Minority Health Month, we are bringing you a special month-long podcast series with our Strategic Alliance Partner, UPMC Health Plan. Welcome to our final episode of this limited series and our conversation with Janine Jelks-Seale, MSPPM, director of health equity at UPMC Health Plan.
Listen
Metabolic Syndrome, Obesity Contribute to Breast Cancer Mortality in Postmenopausal Women
May 13th 2024A large study finds that obesity and metabolic syndrome raise breast cancer mortality risk, but through different mechanisms. Metabolic syndrome is linked to a specific type of breast cancer, whereas obesity increases risk across all breast cancer subtypes.
Read More