March 11, 2026 / biostate.AI
Why ChatGPT/Claude Can’t Understand Biology
Dear Biostate AI investors, followers, and collaborators,
AI models like ChatGPT, Gemini, and Claude are trained on human language. The "language" of biology is very different from human language. You can’t simply feed gene expression data into a language model and expect it to understand disease.
When we measure gene expression from a single blood sample, we’re observing the activity levels of more than 20,000 genes at once. The interactions between genes are complex, nonlinear, and highly dependent on biological context.
There’s also a deeper difference in what the AI is trying to model. Many AI drug discovery companies focus on designing static molecules: drug candidates that, once synthesized, do not change. But patients are not static systems. Human biology changes constantly as disease progresses, treatments are given, and the body responds.
Understanding those changing biological states requires a different kind of AI.
That’s why we built N-Act, our general-purpose biomedical AI, from the ground up. N-Act is designed specifically to learn from molecular data and identify patterns in gene expression that relate to health, disease, and patient outcomes.
Explore AI VisualizationLanguage models still play an important role. We designed a separate system called K-Dense, built on top of models like Gemini and Claude, to help structure and summarize biomedical knowledge written in human language.
But when it comes to understanding the molecular basis of disease, N-Act is purpose-built for that task.
Thank you for being part of this journey. Simply reply to this email if you would like to talk.
Best regards,
The Biostate AI Team