Building Foundation Models in Biology

Abstract

Over the last three years, Foundation models such as Dall-E and ChatGPT have taken the world by storm and have ushered in the "AI Boom". The next challenge is to build such models in Biology. This article examines the way text-based foundation models were built and the ways in which the approach has to be tweaked to build Foundation models in Biology. More specifically, it looks at the three components of the scaling laws - Data, Architecture and Compute and how they can be adapted to build foundation models in Biology.
These Foundation models can then be used for a variety of downstream tasks such as Identification and if possible, prevention of conditions, treatment planning and nutrition planning.