Learn more about our AI
Our AI, developed by our team of research scientists, engineers, and healthcare professionals, is a suite of AI tools designed around a doctor’s brain to provide accessible healthcare for millions. A number of these tools are modular, so they can be standalone and used in isolation, or combined to suit different requirements.
We’re working with partners, from governments and foundations, to businesses and pharmaceuticals, to tech companies and telcos — tailoring our platform to meet their specific needs.
What our AI does
We’ve designed our AI tools to empower people with knowledge of their health, with the goal of relieving pressure on clinicians.
Our AI can efficiently read, comprehend, and learn from anonymized, aggregated, and medical datasets—when patients give consent for us to use their health information.
It can use data in order to decide on, and provide information about, the likely causes for people’s symptoms. It then suggests possible next steps, including treatment information. It can let you know about general risks for certain conditions, when comparing user-inputted information and generally available data.
How it works
Our AI revolves around four main parts - the knowledge base, the comprehensive health record, the probabilistic graphical model.
Central to our AI is a form of digital encyclopedia of medicine (our knowledge base) that contains the definitions, characteristics and relationships of certain diseases, symptoms and treatments. It contextualizes this information with a graphical representation that shows the relationships between the medical components.
The comprehensive health record
This holds all of the available information about our individual users, when they give consent for us to use their health information. The record includes their medical history and consented data gathered through interacting with Babylon. It helps us make connections between users and different types of conditions, and their likely progressions over time when compared against generally available data.
Probabilistic graphical model
This uses the knowledge from our digital encyclopedia, combined with all the data to test different models about certain illnesses. It helps identify conditions which may match the information entered. A similar approach is also used to predict disease risks over the next five years when compared against generally available data.
Simulations are used to estimate ‘what-if’ scenarios, to predict what happens if people continue their routines for diet, exercise, sleep, and stress. It helps users understand the impact of their actions and helps us develop optimized care plans for them.
We’re part of the community
We contribute to the AI community by publishing papers, speaking at conferences and open-sourcing some of our work for the benefit of all.
Estimating Mutual Information Between Dense Word Embeddings
Vitalii Zhelezniak, Aleksandar Savkov, April Shen, Nils Hammerla
Hybrid Reasoning Over Large Knowledge Bases Using On-The-Fly Knowledge Extraction
Stoilos, Giorgos and Juric, Damir and Wartak, Szymon and Schulz, Claudia and Khodadadi, Mohammad
Can Embeddings Adequately Represent Medical Terminology? New Large-Scale Medical Term Similarity Datasets
Claudia Schulz, Damir Juric