The "learned faction" finds the next technology trend.

What's the next trend? This is a fascinating question. An organization called IARPA (FDI Advanced Research Projects) hopes to discover technologies that might be popular in the next 3-5 years by running data.

It has developed an application called fuse (Foresight and understanding from scientific exposition), which scans a large number of scholarly documents, patents, and other scientific documents to find the new technology of the Pearl Trail, For scientists to adjust project development priorities.

In addition to not using input keywords, fuse is really like a search engine, constantly digging up relevant Chinese and English databases, and periodically giving some emerging technology rankings according to their prospects for the next few years, with the ability to grow at 100% or 200% of those as significant technology. For example, "Quinone" (quinone) the organic compound is a word that fuse traced, which can be used to improve energy storage, while "perovskite" can be used to tin-based electric solar cells. "Quinone" is ranked higher here because the system knows that scientists have written more about its potential applications.

Fuse also evaluates the maturity of a technology. Patent documents on early technology often explain their contributions, and many scientists develop a new concept that compares their work with the results of existing research, and when a term becomes more prevalent, the number of comparisons is less. For the data of the well-known periodicals, fuse will also give higher weights to the algorithm.

According to Fuse's data report, IARPA's another project, forecasting science and Marvell, will then investigate tens of thousands of experts to assess the feasibility of a technology. Of course, the survey on this scale is also done online, and professionals are evaluating potential new technologies through a website called Scicast.

But the system is struggling to differentiate between outdated trends and long-term momentum, and the team is tweaking the algorithm to ignore the trend. They now judge, "if a technology trend lasts for 3 years, or if there is a 3-year momentum in a community journal that sends 20 papers a year, it's likely to create a vertical community rather than an over gas technology." ”