Automating Scientific Discovery

Many believe that the biggest crisis facing science today is “too much information.” Last year, in the biomedical field alone, more than a million papers were published, and to date, the total number of peer-reviewed biomedical studies is more than 26 million. At the same time, the quality of studies has declined. Publishers’ restrictions and author bias can make it difficult to extract reliable data, making it virtually impossible for scientists to keep abreast of all the latest developments, much less separate out the most relevant ones.

To combat this information overload, a new system has been developed that uses artificial intelligence to comb through the 4,000 scientific papers published every day and deliver insights regarding which ones have the greatest potential impact on future development. A combination of neural networks, algorithms and machine-learning-based predictive intelligence goes beyond data mining to analyze the information and generate new hypotheses based on sound research.

Most importantly, as part of the Chan Zuckerberg Initiative, the system will be available to researchers at no cost.