AI can help the platform understand which scientific problems its users are interested in.
This is relevant to:
- Search: Scientific problem-based manuscript search as a potentially more useful alternative to academic field or keyword based search
- Feed: Personalized user feed algorithms for manuscript and review discovery
AI can function as a creative assistant for academic research.
This would be a paid service made available to platform users. Broadly, the idea is to re-train or fine tune an existing LLM with platform data including manuscripts, code, reviews and supplementary research data. This would allow researchers to more easily understand what has been done before, to discover non obvious connections between seemingly unrelated ideas, and more generally to solve a wide range of practical problems efficiently.
Example:
A computational biologist is modeling immune selection. She is interested in testing the hypothesis that a person's HLA profile and specifically its degree of homozygosity has a causal influence on the likelihood of tumor clone immune escape after treatment with anti PD-L1 immune therapy. This is a complex hypothesis that can be addressed from many angles and with multiple data types. The rate of progress in this area of cancer research could be greatly improved if the AI assistant was aware of existing theoretical approaches that relate to the hypothesis, as well as published methods, data and code.