In the digital age, data is the new aether. Everything in modern life is surrounded and connected by data, and artificial intelligence has become the medium through which to tap into that invisible aether and enable it to guide and enrich our lives. The touch of AI can be seen in our curated social media, the responses to our queries of search engines, and soon AI will quite literally drive our lives forward.
For scientists, AI is rapidly becoming a key tool for data analysis, enabling the impossible. For example, pharmaceutical giant Merck is using deep-learning to significantly improve the identification of drug candidates, while for research organizations like CERN, AI has become a key tool for analyzing data from the Large Hadron Collider.
Hypothesis testing relies invariably on a foundation of both experimentation and literature discovery. Despite a revolution in the way scientific results are published, archived and shared, comparatively little change has come to the way scientists access this data. In much the same way that a previous generation of scientists might have searched a library catalogue, modern scientists search digital catalogues. Often armed only with the blunt instruments of keyword and date queries, researchers are searching larger literature databases than ever before. With archives expanding by up to 6,000 publications each day and more than 1.5 million per year, staying current is a task that, for scientists, has become next to impossible. Case in point — if a cancer researcher were to attempt to review each 2014 Plos One paper having a title containing the word “cancer,” they would need to review 1,417 papers.
- Read more: 7 Cool Things to Know about AI
The wealth of academic literature, growing by the hour, is as much an opportunity as it is an obstacle. Collectively, publishers have created an electronic history of scientific progress. Each journal article represents not only a milestone in scientific discovery, but also a collection of metadata. Academic papers are connected not only by subject, but also by author, technique and citation.
Those connections between papers trace the paths through which science has advanced, and reveal the fields, techniques and individuals who have been the sources of rapid innovation. That information, growing in real-time, traces not only the history of progress, but also reveals its path into the future. When artificial intelligence is applied to those digital archives, it becomes possible to see not only the current state of human progress, but where the most rapid progress will likely occur in coming years. The rise of CRISPR as a solution for genetic engineering, which began to receive widespread attention in 2013, was predicted by Meta’s Horizon Scanning technology in 2010. Imagine what you might have done differently had you known about CRISPR three years before anyone else.
As academics, publishers and funding agencies begin to recognize these opportunities for innovation, it becomes possible to direct their efforts to the maximum effect. With the aid of AI-driven bibliometric approaches, publishers are not only better-positioned to draw attention to key fields and share key results, but are also at a reduced risk of rejecting key papers. At the same time, AI-driven horizon scanning presents funding agencies with the best opportunity to maximize the impact of any budget, ensuring the most innovative projects are well-funded. Most importantly for working scientists, AI provides the only opportunity to turn information overload into a powerful engine of innovation, driving their research forward.
Sam Molyneux is the CEO and co-founder of Meta.