Our scientific infrastructure — the tools and methods for conducting scientific research — is rusting, so we are calling for an artificial intelligence-driven, Manhattan Project-sized effort to revive it. Just as our roads and bridges are overdue for investment, so is the infrastructure for scientific research; that is, the body of scientific thought and the tools for searching through it. That’s because the sheer quantity of research papers being published daily makes it impossible to keep up with the state of the art in any scientific discipline, which is necessary for scientific progress.
Online scientific search services, such as those provided by the multinational tech companies, try to solve some of the problem, but that’s like having an automobile manufacturer owning the roads — a conflict waiting to happen.
{mosads}Scientists need the infrastructure for scientific search to aid their research, and they need it to offer relevancy and ways to separate the wheat from the chaff; the useful from the noise, via AI-enabled algorithms. With AI, such an infrastructure would be able to identify the exact study a scientist needs from the tens of thousands on a topic. It could extract the finding, chart or study one is looking for, and suggest other relevant and cited articles. It could do analysis over several studies or papers, comparing results among them. And it could rate citation quality so that scientists are finding the more credible papers to draw from.
Such an infrastructure would be best offered up as a public or nonprofit service, so that commercial and profit considerations do not skew the offering, or worse, close it down altogether. If online search services ended, what recourse would any scientist have? Government is the most logical, but not only, possible provider of investment in open-source scholarly search.
Just as the Manhattan Project brought together the best minds in physics in a time of great danger, we need a similar large, visionary effort with regard to helping scientists have the AI-powered tools to work together and build on each others’ research to solve the most pressing issues in health, climate and energy. The vision is this: In concert with nonprofits, government can provide scientists from all disciplines the access and tools — including artificial intelligence tools — to find those studies that can propel their work the furthest and the fastest.
Why is search so important to the scientific process? Well, what if a paper from eight years ago happened to note a side effect of a drug that may have therapeutic value for another disease? What if a finding from a different discipline might help propel the research of a scientist in another? Scientific search, powered by AI, offers scientists the gift of clearing away the clutter so they can pay attention to what is important.
The foundation of this “Manhattan Project for scientific research” exists today. The nonprofit Allen Institute for Artificial Intelligence has a free scholarly search service, called Semantic Scholar, for scientists in computer science and biomedical fields. It employs methods from data mining, natural-language processing and computer vision to create powerful search and discovery experiences.
But Semantic Scholar, funded by a single nonprofit, is in its infancy. It addresses only those research papers in computer science and biomedical fields, meaning scientists in only a few disciplines can reap the rewards of easy and smart access to others’ work.
What’s needed instead is an investment in scientific search that is as strategic as the Manhattan Project was for our nuclear capability. This investment, separate from any existing effort, would offer many downstream benefits. It would keep the scientific search capability in the public’s hands, for the common good, and for all scientists to draw from. It would be foundational in pushing the state of the art of science to solve the thorniest problems facing our nation. It would ensure doctors are up to date on the latest treatments, and physicists up to date on the latest in material science. Over all, it would help maintain our scientific leadership on the world’s stage.
Infrastructure investment in science is an investment in jobs, in health, in economic growth and environmental solutions. In today’s resource-constrained environment, taxpayer dollars have to be invested where they will do the most good. And it is our belief is a public/private partnership to solve the data tsunami problem is one of the most important investments. The future of humanity depends upon it.
Oren Etzioni is CEO of the Allen Institute for Artificial Intelligence in Seattle. a world-renowned researcher in the field of AI and computer science, he is a professor at the University of Washington.