COMP: Update on Petaflop Machine Progress

Doug Bailey (Doug.Bailey@ey.com)
Wed, 28 Oct 1998 9:04:34 -0500

While the objective of building a petaflop machine isn't new, here's some information I happened upon on the PR Newswire concerning progress towards the machine. I'm including a portion of the article since there is no URL:

3-D Transistor Key Enabling Technology to be Demonstrated

"COSTA MESA, Calif., Sept. 29 /PRNewswire/ -- Irvine Sensors
Corporation (Nasdaq: IRSN, Boston: ISC) announced today that it has received an approximate $738,000 research and development contract from the United States Army Space and Missile Defense Command to demonstrate the feasibility of an ultra high density interconnect to enhance its 3D Artificial Neural Network(TM) (3DANN(TM)) technology. The new interconnect is part of Irvine Sensors' planned progression of technologies intended to eventually lead to a "Silicon Brain," a recognition system conceived to emulate performance of the human central nervous system. The Army's interest in such technology is for potential application to ballistic missile defense systems. Irvine Sensors believes that human-level recognition technology could also have commercial applications in such fields as medical diagnostics, DNA sequencing, radiography, security systems, and speech recognition. Optical displays are also expected to benefit from the three-dimensional active transistor interconnect technology to be developed under the new contract."

"The problem being addressed by the Silicon Brain development can be
compared to that faced by a football quarterback who, when confronting multiple threats and opportunities simultaneously, must take instantaneous and decisive action. During missile defense, events unfold even faster, and the threats and opportunities are much greater in number. What you'd like is human-level intuitive responsiveness, but at higher speeds and typically under adverse, humanly-impossible conditions," said John C. Carson, Irvine Sensors' Sr. Vice President and Chief Technical Officer. "Although it will take several years before the technology can be made available for products, we believe the Silicon Brain will ultimately be able to solve that problem... attaining petaflop performance, or quadrillions of operations per second, using less than ten watts of power, in one-third of a cubic foot, or about the size of a shoebox. This will be the same volumetric efficiency as the human brain."

(Copyright(c) 1998, PR Newswire)