Last year, the U. S. Defense Advanced Research Projects Agency (DARPA), which funds a range of research in military technology, launched a five-year $ 1.5 billion program. More commonly known as the Electronics Resurrection Initiative (ERI), the program is designed to support breakthrough efforts in chip technology. The agency has just unveiled the first list of research groups selected to explore untested but potentially promising approaches that could revolutionize chip design and manufacturing in the United States.
In recent years, hardware development has receded into the background compared to software development, and this worries the US military for several reasons.
Topping the list of concerns is Moore's Law, which states that the number of transistors installed on a chip doubles approximately every two years. At the moment, there are all signs that modern electronics is reaching its physical limits, and in the near future the process of its development will stop.
There are also concerns about the rising cost of designing integrated circuits and an increase in the number of foreign manufacturers (China intends to produce chips that will add AI systems to any gadget!).
ERI's budget provides for a fourfold increase in DARPA's regular annual hardware costs. The initial projects reflect the three main pillars of the Initiative: chip design, architecture, new materials, and integration.
One project aims to drastically reduce the time it takes to create new chip designs by reducing design time from a few years or months to a few days by supplementing the development automation process with machine learning and other tools so that even relatively inexperienced users can create high-quality designs in short terms.
Moving beyond Moore's Law will require radically new materials and new ways of integrating computing power and memory. The architecture of typical modern chips involves the constant movement of data between the memory components that store them and the processors that process them, which literally sucks energy and creates one of the biggest obstacles to increasing processing power.
Another ERI project will look at ways in which new integration schemes can eliminate, or at least significantly reduce, the need for data migration. The ultimate goal is to efficiently inject computing power directly into memory, which can lead to dramatic performance gains.
Promotional video:
DARPA primarily plans to create such hardware and software that can be reconfigured in real time to handle general tasks or specialized applications such as specific applications for artificial intelligence. Today, multiple chips are required to solve such problems, which increases the complexity and cost of an AI system.
Some of DARPA's efforts are encroaching on areas that are already being intensively developed by industry. An example is a project to develop a 3D system on a chip, which aims to further prolong Moore's Law using new materials such as carbon nanotubes and smarter ways to place active components and separate electronic circuits.
Erica Fuchs, a Carnegie Mellon University professor who is an expert on public policy related to new technologies, believes that the general approach of the US government to supporting innovation in electronics is "an order of magnitude lower" than necessary to solve problems. To somehow help close the gap, it is necessary to at least double government investment, as large corporations are extremely reluctant to spend on research in the field of promising technologies.
Serg kite