Science

Rejuve AI Technical Design

Rejuve has developed its own AI platform, applying logical, evolutionary and neural net algorithms integrated within SingularityNET to meet the needs of longevity science. Using this platform, scientists construct a multi-resolutional, mechanistic, dynamic model of the body with which to infer longevity treatments. 

This model is now ready to predict the outcome of treatments in general populations and specific individuals. As ongoing work makes it more complete, it will become broad in scope and more accurate in its predictions.

The Rejuve platform comprises two distinct components, interlinked using customized AI methods. The first is the OpenCog AGI engine, incorporating probabilistic evolutionary learning (MOSES), probabilistic logical inference (PLN), stochastic pattern mining and other methods that act on Bio-Atomspace, a hypergraph store integrating biological knowledge from databases, ontologies and datasets.

The second is a deep neural embedding framework, which combines multiple neural architectures to create embedding vectors representing aspects of each body. This is connected to OpenCog via a biomedical-AI customized deployment of the DeepWalk graph embedding algorithm, which maps nodes in Bio-Atomspace into embedding vectors to be processed by deep neural networks.

AI platform biodata comes from Members as well as other biological and medical databases. Members contribute different types and amounts of data. One may contribute data from a wearable while others contribute nutritional interviews, blood test results or supplement data. Data from past studies and contributions help AI interpret smaller amounts of data from lighter users to make recommendations and leverage data for further studies.

Rejuve addresses many common issues prevalent in healthcare data, including how to deal with overlapping, sparse and multi-modal observational data. Health studies struggle to deal with issues in observational data which mask causal connections. Rejuve’s AI makes sense out of this ambiguity to reconstruct the whole out of individual causal relations.

Synergistic Data
Integration 

A powerful aspect of the Rejuve approach is to effectively integrate patterns found in multiple types of data. The volume of data and computational power needed to solve complex health problems is only recently available to researchers. AI is improving rapidly though most gains exists in silos applied to healthcare claims, biomarkers, genomics and imaging. These siloed applications are not enough to analyze complex systems like the human body.

Rejuve combines specific information and insights to develop a high-level understanding of the body and join these silos in a platform that allows for deeply synergistic data integration. Analysis tools fuse multiple data channels to facilitate interpretation of sparsely overlapping observational data. When a dataset contains patients with measurable variables A and B and patients with B and C, Rejuve can interpret variables A and C by using overlapping variable B.

Causation

Deep neural networks, today’s most popular data modeling method, learn a system in amazing detail but are poor when making recommendations. Neural networks struggle with suggesting the best treatment in the models they have made of the world, mainly because they are correlational rather than causal. Correlation is fine for general prediction, but when predicting treatments principles of causation matter. Rejuve’s AI prioritizes causation, leveraging tailored neural techniques combined with probabilistic causal logical inference methods.

Decentralization

Modelling and understanding the body is too difficult a task for any one entity or organization to perform. These constraints can be circumvented through crowdsourcing and Rejuve develops methods by which a society of researchers come to consensus on models of complexity while exploring multiple combinations of models and datasets. Metrics of consonance, formally measuring the sense models and data make together, facilitate automated complexification in a decentralized manner. The models submitted by researchers can be combined in an automated way or with researcher participation.

From Hong Kong, China, Europe, US, Russia, Brazil and the Middle East, Rejuve Team Members are passionate about AI, blockchain, longevity science and the opportunity to improve people’s lives. We bring together decades of experience in AI, medicine, bioinformatics and international business to provide a world-class team in pursuit of longevity.