IBM's Watson makes science elementary and useful

The recent victory of a supercomputer called 'Watson' against the human champions of quiz show Jeopardy! marked a milestone for the elite mainland scientists of International Business Machines Corp.

IBM's China Research Lab, led by senior technical staff member Pan Yue, is credited with developing key software over the past three years that improved Watson's learning capability and accurate processing of answers to questions posed in natural human language.

Watson is a high-performance computer the size of 10 refrigerators named after IBM founder Thomas Watson. In three telecast rounds of Jeopardy! from February 14 to 16, it beat Ken Jennings and Brad Rutter, the two most distinguished contestants to win in the popular United States television game show.

'The Watson project demonstrated the world-class capabilities of IBM China Research Lab, and its contribution to the advancement of science and technology worldwide,' Pan said.

Three other mainland researchers were part of the Watson project. Qiu Zhaoming helped Watson formulate 'a confidence level' for its answers. Zhang Lei worked on using so-called 'open-linked data as a source of knowledge and strategies for multi-layer machine-learning'. And Ni Yuan integrated open-linked data as a source for forming Watson's text answers.

'They have contributed pieces of software that can accelerate the 'thinking' behind Watson,' John Kelly, an IBM senior vice-president and director of its global research operations, said.

IBM's other research laboratories worldwide were also enlisted to work on the project over a four-year period. Chinese scientists were 'very significant contributors', Kelly said.

According to IBM, Watson's deep understanding of language allows it to process and answer complex questions that include puns, irony and riddles common in natural language. That compares with the commonly used internet search engines, which typically provide links to websites where answers may be found.

Watson is a supercomputer built with 10 racks of IBM's commercially available Power 750 servers running the Linux operating system. It uses 15 terabytes of random-access memory and 2,880 Power7 processors. A terabyte is 1,000 gigabytes.

Kelly said Watson was 'about 1,000 times more powerful than Deep Blue', the computer developed by IBM that won a six-game chess match against world champion Garry Kasparov in May 1997. He said the software behind Watson, known as 'DeepQA', would soon be applied commercially across many industries, especially in health care.

IBM's Beijing-based research team, which was formed in 1995, is now focused on applying knowledge management and artificial intelligence technologies in health-care applications, such as chronic disease management.

Pan described chronic diseases, including diabetes and heart diseases, as the most significant cause of death on the mainland. He said health care on the mainland, particularly in remote cities and villages, was often inaccessible and expensive. 'To address those challenges, there is an urgent need for the transformation of the health-care system.'

The central government has pledged 850 billion yuan (HK$1.01 trillion) for health-care reform that promises to make medical services safer and more affordable for the country's 1.3 billion citizens this year.

The IBM China Research Lab was involved in introducing in 2009 an information sharing and analytics systems at the Guangdong Hospital of Traditional Chinese Medicine. The system was designed to share electronic medical records that incorporated traditional Chinese and modern Western medicine.

The system integrated health records that combine Eastern and Western medicine into one standard system and applied sophisticated analytics. This provided a way for health-care practitioners to more deeply understand which treatment plans and techniques from each approach work best for specific diseases and medical conditions.

Last year, IBM researchers and the Peking University People's Hospital built a so-called evidence-based patient-centric care (ePC3) system to enable co-operation and resource-sharing among medical services providers. The ePC3 system was designed to provide better and ultimately more affordable clinical decision support at the point of care.

IBM was the first multinational enterprise on the mainland to establish joint-venture plants and research and development institutes to develop local talents, and introduce world-class professional services to enterprises and government entities.

At present, IBM has 29 branches across the mainland and businesses in 320 cities. It has 10 joint ventures and wholly owned companies that are involved in computer hardware manufacturing, software development, information-technology services and equipment leasing.

IBM chief financial officer Mark Loughridge last month reported that the company's mainland sales grew 25 per cent year on year in the quarter to December, when its global revenue reached a record US$29 billion.

Watson was not the only high-performance computing system to win the week it beat the human champions of the Jeopardy! show.

The other system is called World Community Grid, a virtual supercomputer that helps scientists solve humanitarian challenges by tapping the unused processing power of personal computers around the world.

Besides receiving US$500,000 in prize money from the show, scientists who use World Community Grid have earned unprecedented support worldwide because of Watson's victory. The day after the tournament's conclusion, the grid saw a 700 per cent spike in the number of people who normally volunteer their computers' spare power for the effort, according to IBM.

The grid, an initiative of the IBM International Foundation, was designed to tackle projects such as new treatments for HIV/AIDS, cancer research, and affordable water purification.

It works by pooling the unused power of 1.7 million personal computers from 535,000 volunteers in more than 80 countries. It then makes this computational power available for scientists.

Scientists at Tsinghua University, for example, use the grid for an international initiative the Beijing institution leads, called 'Computing for Clean Water'.

That project, now under way at the university's Centre for Novel Multidisciplinary Mechanics, aims to develop ways to filter polluted water. It will also convert salt water into drinkable water.