As large language models (LLMs) continue to gain traction across scientific industries, many researchers are exploring how these tools can improve efficiency without compromising scientific rigor.
Ahead of MetrumRG’s PAGE workshop, Beyond the Hype: Practical Applications of LLMs in Pharmacometrics, Seth Green shares both his excitement and his concerns about the growing role of AI in science.
In specific cases, I've already seen them deliver on the promise of "less time on drudgery, more time for thinking." I find this really exciting. I have gotten to meet so many brilliant scientists, and I'm always blown away by how much of their time and brainspace is taken up by things like formatting tables and wrestling with data manipulation syntax. Recently, one of our senior scientists was using an LLM to help put together a simulation script. They were thrilled by how smoothly it went, and commented, "often by the time I've finished coding a script like this, my brain is too tired to actually think about the scientific question I was trying to answer." I'm excited that frustration like that could become a thing of the past.
I'm concerned with how easy LLMs make it to "do bad work, faster." I worry that AI hype in the media, and an understandable desire for greater efficiency, are driving unrealistic expectations for "productivity" in science. This is compounded by the fact that LLMs are very good at generating content that "looks right," but doesn't contain any real nuggets of value. My concern is that scientists will feel pressure to churn through more work, but will not feel they have the time to do thoughtful work. As a result, valuable insights could be missed, mistakes or wrong assumptions could slip through, and talented people could see their intellectual and professional growth be stunted.
MetrumRG’s workshop: Beyond the Hype: Practical Applications of LLMs in Pharmacometrics will take place during the PAGE Meeting in Dubrovnik, Croatia.
Hosted by Seth Green, M.S. and Sam Callisto, Ph.D., the session will explore practical, human-in-the-loop approaches for integrating LLMs into pharmacometric workflows while maintaining scientific rigor, regulatory compliance, and data governance standards.
Learn more and register today.