IBM's Watson Computer System Plays Jeopardy! in a Practice Round

Shown residing in IBM’s lab in the large photo above (credit: Clockready/Wikimedia Commons) and in its Jeopardy! TV appearance (credit: IBM), Watson has unique analytic powers that are now being applied to an array of real-world problems, including materials discovery.

According to IBM, the wired world creates 2.5 quintillion (2.5 × 1018) bytes of data each and every day. How to harness that amount of data—turn it into available information and eventually into usable knowledge—has become a real problem and spawned a new branch of information science loosely termed “big data.”

In the materials science community, the Materials Genome Initiative is an ongoing attempt to use big data concepts for materials discovery by making the vast amounts of materials property data already available accessible to all scientists. Announced by the White House in June 2011, the MGI has since been the topic of multiple CTT reports (see, for example, here and here).

A few months before the MGI announcement, in February 2011, a computer system developed by IBM and capable of understanding and answering questions posed in natural language pretty much crushed two human competitors on the TV quiz show Jeopardy! And its vanquished carbon-based foes weren’t just any contestants—they were the most successful Jeopardy! champions of all time.

Since then, IBM has worked to harness the analytical firepower of Watson—so named in honor of long-time chairman and CEO Thomas J.—to solve various types of real-world problems. In January, the company donated a version of the system to Rensselaer Polytechnic Institute. Just last week, IBM announced plans to apply Watson in a new lab aimed at unifying “data, expertise, and novel analytics to speed discoveries including retail, medicine, and finance,” according to the news release.

Add materials discovery to that list. According to an EE Times interview with IBM director of strategy and program development Jeff Welser, materials science is high on the priorities list at the company’s Accelerated Discovery Lab.

“We are in the process of expanding into the discovery space of materials science,” Welser says in the interview. “For instance, what if I need to find a new insulator, or a new conducting material, or even a transistor? We want to discover what new combinations of materials might be interesting to try, and whether there are ideas out there for new types of materials that have not been tried yet.”

“Superconducting materials are a good example, since they are a complicated combination of alloys, so there are lots of different elements that could be tried,” Welser continues. “What we can do in the Accelerated Discovery Lab is look through literature about materials used for different purposes and figure out whether that material has a property that might make it worth looking at for superconducting applications.”

Sifting through mountains of data to unearth the nuggets that are most valuable to a specific investigation is going to require a new skill set. Recognizing that fact, Georgia Tech announced earlier this month that it will launch a program aimed at training a new breed of data-savvy, entrepreneurial scientists.

According to Richard Fujimoto of Tech’s School of Computational Science and Engineering, the new graduate training program will teach students to “employ advances in ‘big data’ and information technology to significantly reduce the timelines now required for new materials to be created and incorporated into commercial products.”

Called FLAMEL (From Learning, Analytics, and Materials to Entrepreneurship and Leadership) with a nod to 14th-century French alchemist Nicolas Flamel, the new program is based on materials informatics, which, according to the release, “aims to develop new approaches to materials design and manufacturing using data analytics combined with modeling and simulation.” The five-year program is supported by the National Science Foundation. Georgia Tech says NSF funding will allow training of 24 doctoral students, but the school expects FLAMEL to “create educational opportunities that will impact hundreds of students in the years ahead.”

Author

Jim Destfani

CTT Categories

  • Education
  • Modeling & Simulation