SOCR Foundational Generative Artificial Intelligence Models (GAIMs)

The SOCR foundational generative artificial intelligence models (GAIMs) utilize transfer learning, transformers, and deep neural networks to build reliable, scalable, and trustworthy AI predictors and classifiers supporting timely and responsible decision-making. In general, GAIMs include different models of varying complexities with 1-100 billion parameters. The DSPA GAIM Appendix in the SOCR DSPA Book provide additional information, details, and experimental examples.

Core SOCR AI Applications

SOCR Synthetic Brain Generator

The Synthetic Brain Generator leverages custom in-house models to produce realistic synthetic imaging data, including 3D brain volumes and 2D MRI slices, for developing predictive models like tumor detection, while allowing users to generate various image types based on different parameters.

Go to app ➔

SOCR AI Bot

The SOCR AI Bot leverages SOCR/DSPA computational libraries and Generative AI interfaces, including OpenAI's GPT-4o, to translate natural language commands into R code, generate synthetic text and images, analyze data from uploaded files, and provide results as downloadable reports.

Go to app ➔

SOCR GitHub GAIM

The SOCR GAIM GitHub site includes software, code, documentation, and other resources on the SOCR AI Bot, GR Virtual-Hospital, and many other SOCR Applications.

Go to GitHub GAIM ➔

Virtual Hospital

The GrayRain Virtual Hospital utilizes SOCR DataSifter technology for synthetic patient generation, generation of heterogeneous simulated biomedical/health datasets, statistical data obfuscation of existing clinical data that balances the risk of patient de-identification and preservation of information content (data utility/value), and data augmentation to append existing data archives and increase the number of cases and the richness of the data features.

Go to VirtualHospital ➔