AI—such as predicting the mechanical properties of biopolymer gels—can
improve computational materials science.
In the final presentation, Michael Lee, editor of Science Robotics,
presented a workshop on the editorial process in the Science family
of journals. This included an overview of many of the journals and the
types of articles they publish,
plus some of the key features
that an accepted paper will
include, such as overcoming
technological limits, answering
longstanding questions,
altering the way scientists look
at a problem, and opening new
opportunities for R&D.Staying the course
In closing the forum, Xinlong
Zhao, vice president of Zhejiang
Lab, asserted that “intelligent
computing is an important
cornerstone for human–
cyber–physical integration.”
For that reason, he added, it is “critical to create a worldwide cooperative
ecosystem for intelligent computing and promote collaborative innovation,
in order to greatly boost the development of computing science and
technology.”
Research in this area creates many challenges, however. One that Zhao
noted is the need for “not only innovative computing architecture, but
also a whole set of supporting theoretical systems, which, from a global
perspective, have just begun to emerge.”
As intelligent computing develops, it will benefit society in many ways.
Citing one example, Zhao said, “The intelligent computing digital reactor
initiated by Zhejiang Lab will become a highly efficient innovation engine
and a means to serve scientific and technological ingenuity and social
development.”
The partnership between the Zhejiang Lab and Science/AAAS reflects
Zhao’s last point: “The Innovation Forum on Intelligent Computing will
become a significant catalyst for academic exchange.” He expressed keen
interest in continuing to shape the forum into “an annual flagship event in
computing science and technology, and a high-level, open international
platform for academic exchange.”Produced by the Science
Sponsored bySome of the most exciting advances in intelligent computing over the
past few decades have come from Jürgen Schmidhuber, director of the
Artificial Intelligence Initiative at King Abdullah University of Science and
Technology in Saudi Arabia. In his keynote address, he called 1990–
the “miracle year” when his team created the idea of deep learning, which
included some automated
methods of training a system.
As scientists learn more
about biological intelligence,
their discoveries can lead to
new developments in computing
technology, as demonstrated by
Yulia Sandamirskaya, leader of
the applications research team of
Intel’s neuromorphic computing
lab. In her talk, she observed that
human intelligence spends more
time controlling movements
than, for example, playing chess;
thus artificial neural networks
can improve by emulating the
processing behind human
movements.
Elaborating on the evolutionary development of AI, Yaochu Jin,
Alexander von Humboldt Professor for Artificial Intelligence at Bielefeld
University in Germany, discussed biological evolution, development, and
learning. Then, he described how these mechanisms might serve as
complementary tools for automating the design of autonomous AI-based
computing systems.
Looking further into automating software, Zhi Jin, a professor of
software engineering at Peking University in China, reviewed the early
top-down approaches to this process and the potential for a bottom-up
approach. She concluded that a hybrid of the two approaches might work
best, but that the preferred approach could vary by application.
Addressing numerical applications for high-performance computing,
Guangwen Yang, director of the National Supercomputing Center in Wuxi
and head of the Intelligent Supercomputing Center of the Zhejiang Lab,
mentioned astrophysics and life sciences, advanced manufacturing and
materials science, and more. Delving even deeper into such applications,
he commented on ways to make intelligent supercomputers.
Wei D. Lu, professor of electrical engineering and computer science at
the University of Michigan in Ann Arbor, described energetic bottlenecks
in computing. He explained how to improve the efficiency of computation
in various ways, including the use of memristors, which emulate the
processes of synapses in the nervous system.
Jincang Zhang, distinguished professor of condensed matter and
materials physics at Shanghai University in China, reviewed the
development of methods in computational materials science, ranging
from traditional to data-driven approaches. He showed how advances inOur goal is to bring [scientists]
together to confront the
huge challenges we face as a
society, and explore the limits
of science, thereby setting and
leading the trends of intelligent
computing in the future.
— Shiqiang Zhu
0121Product.indd 241 1/12/22 7:26 AM