Machine Learning Applied to Scientific Discovery – Meta & The Chan Zuckerberg Initiative

Shanghai 1

Meta is a team of data engineers and bench scientists with an aim to unlock the world’s scientific and technical insights using artificial intelligence (AI). Specifically, it aims to use AI to machine-read, as well as understand the contents of the near 4,000 papers a day that are now published globally, in an effort to create structured connections that then find their appropriate audiences.

As part of the Chan Zuckerberg Initiative – a non-profit organisation founded by Facebook’s Mark Zuckerberg – Meta’s toolset is open access and free, meaning that researchers and others can utilise and develop them.

Direct indexing partnerships with hundreds of the world’s leading publishers means that more than 37-million scientific entities can be mapped, inter-connected and ranked, creating a “knowledge graph” that forms the base dataset powering a variety of applications.

These include Meta Science, an AI-enabled literature discovery engine comprising 65-million pages representing virtually every article, person and entity in biomedicine/life sciences.

Formulated from machine-read data, this allows the engine to read a new paper, and automatically compare it to every published paper (across multiple dimensions) since 1665.
A second application, Bibliometic Intelligence, filters and pre-ranks incoming manuscripts so that journal editors can gauge the appropriateness of a paper for their journal, the likely impact factor, and which candidates might be appropriate for peer review. Finally, Horizon Scanning is a predictive intelligence engine that looks at emerging “hop topics” in science, harnessing semantic patterns and plotting likely trends for the next three years.

Scientific discovery In the Machine Age: New tools for competitive advantage.