What Is The Role Of Generative AI In The Life Sciences Industry?

THE USES OF GENERATIVE AI IN LIFE SCIENCES ARE MULTIPLYING RAPIDLY and crossing into a number of functional areas within biopharma companies, with the recent excitement around ChatGPT creating an acceleration of interest around this class of technology. The use of text prompts to generate images highlights a very powerful capability this technology has created: the power to leverage latent spaces to find non-obvious relationships between disparate classes of data. Now imagine, instead of text and images, we can navigate a latent space with biological, chemical, preclinical, and clinical concepts, creating a powerful multidimensional bridge between discovery, translational, and health domains. However, one must be mindful of provenance, or the knowledge base upon which a system is drawing. Large language models are very good at synthesizing and creating very compelling rationales based upon their source data, but don’t assume this source data is completely accurate, unbiased, or complete. To navigate these concerns, you can fine-tune and train models on the information you curate and in which you have confidence. And the low-hanging fruit of improved search, code assessment, and analysis is already at hand. Personally, this strikes me as a “crossing-the-Rubicon” evolution in AI. Powerful technologies can reshape industries, but we must be very careful to test the hypotheses and trace the assertion provenance from this new chapter of AI’s evolution.

JOHN REYNDERS is chief data sciences officer at Neumora Therapeutics.