300 likes | 434 Views
Silicon Recognition, Inc. Neuro-Computing Colloquium & Workshop October 31, 2002. About the Founder. Guy PAILLET - Founder and Chairman Extensive background in artificial vision Worked on parallel processing with Carlo Rubbia, Nobel Prize winner
E N D
Silicon Recognition, Inc. Neuro-Computing Colloquium & Workshop October 31, 2002
About the Founder Guy PAILLET - Founder and Chairman • Extensive background in artificial vision • Worked on parallel processing with Carlo Rubbia, Nobel Prize winner • Worked on solid state physics with Leon Cooper, Nobel Prize winner Silicon Recognition, 2002 - Company Confidential
Silicon Recognition, Inc. • Founded in 1997 by Guy Paillet, Chairman • Worldwide Headquarters in Northern California • ZISC Original Design by IBM • Patented Technology Mission: Become the World Leader in Silicon-based Pattern Recognition Technology Silicon Recognition, 2002 - Company Confidential
*Marco Landi, former President EMEA & ASPAC, Texas Instruments • Unlimited Parallel Processing Scalability • No Operating System • No Computer Language • No Differentiation Memory / Processor • Continual Learning Capability “The Next DSP!”* The first chip that mimics the brain!
Board of Directors Guy Paillet – Chairman Marco Landi – Former President of TI ASPAC and EMEA, and COO of Apple Len Perham – Former President & CEO of IDT Frank Lee – Co-founder of IDT & Galventech Remi Vespa – Former President & CEO of ICD, Application development arm of France Telecom Silicon Recognition, 2002 - Company Confidential
PEL #1 ZISC Memory Category Distance evaluation Search & Sort unit AIF PEL #2 Category Memory Category Distance evaluation Learn Vector Recognize AIF PEL #n Distance Memory Category Distance evaluation AIF Status New PEL Memory Distance evaluation Category yes Unknown AIF Learning/Recognition Silicon Recognition, 2002 - Company Confidential
ZISC complements Classical Approaches ZISC Imprecise, corrupted Artificial Intelligence Data Conventional, Algorithms exact Process easy to describe Ill-defined, not linear Pentium, DSP Silicon Recognition, 2002 - Company Confidential
Products Massively Parallel Data Mining Machines Embedded Smart sensors Silicon Recognition, 2002 - Company Confidential
ZISC Chips First Generation 1 Technology 36 neurons 144 pins Second Generation 0.25 Technology 78 neurons 100 pins Silicon Recognition, 2002 - Company Confidential
Boards First Generation ISA PCI Second Generation: MUREN PCI Embedded Silicon Recognition, 2002 - Company Confidential
Massively Parallel Engines 34,000,000 Vectors Processed per second Silicon Recognition, 2002 - Company Confidential
Major Market Targets • Machine Vision (HALS) • Massively Parallel Engines Silicon Recognition, 2002 - Company Confidential
Selected Users Silicon Recognition, 2002 - Company Confidential
Data Capture Feature Extraction HALS MUREN ZISC Smart Sensing (part 1) Learning & Recognition Silicon Recognition, 2002 - Company Confidential
Smart Sensing (cont) ZISC MUREN HALS Silicon Recognition, 2002 - Company Confidential
Smart Sensing (cont) • Wireless Telecommunications • Video Compression • Authentication • Noise reduction • Automotive • Smart airbags • Obstacle detection • Smart mirrors • Driver alertness • Toy and gaming • Smart dolls • Xbox, PlayStation, etc. Silicon Recognition, 2002 - Company Confidential
Example of ApplicationPerson Tracking Courtesy of Panasonic Silicon Recognition, 2002 - Company Confidential
Example of ApplicationMotion Detection Courtesy of General Vision Silicon Recognition, 2002 - Company Confidential
Example of ApplicationFacial Expression Recognition Silicon Recognition, 2002 - Company Confidential
Example of ApplicationCar Tracking Courtesy of Wescam Silicon Recognition, 2002 - Company Confidential
Example of Application“Check the Checker” Areas Not Checked Yet! Eye Calibration Silicon Recognition, 2002 - Company Confidential
Example of ApplicationTarget Tracking Courtesy of French / German Institute of Defense Silicon Recognition, 2002 - Company Confidential
Example of ApplicationWirelessTelecommunications • Noise reduction at the source and destination • User Identification: • Eye, voice, and fingerprint • Enabling 3G-based transactions • Smart compression Silicon Recognition, 2002 - Company Confidential
Massively Parallel Engines (part 1)Comparison of processing time with increasing number of records Serial Processing @750Mhz ZISC Parallel Processing @8Mhz Processing Time Number of records Silicon Recognition, 2002 - Company Confidential
Massively Parallel Engines (part 2) • Up to 85% of corporate data are unstructured ones • Content-Management a $ 10 B market by 2004 (Meta Group) • Applications: • Video Indexing • Medical • Genomics • Security Silicon Recognition, 2002 - Company Confidential
1 3 2 ZISC 2’ Accessing Unstructured Data John Smith Index Speed Enables Hypothesis Generation Data
Massively Parallel Engines (part 3)Decision Making System 10 billion Number of cells per DIMS 1 billion 1,000,000 100,000 2002 2003 2005 Time Silicon Recognition, 2002 - Company Confidential
Silicon Recognition in New Zealand • ZISC enables a new industry to be created around smart sensing and parallel computing • Fast moving, smart entrepreneurs can bring to the market a new series of designs, based on ZISC • Large corporations will eventually use these designs Silicon Recognition, 2002 - Company Confidential
Silicon Recognition in New Zealand • Excellence Center • AUT • TfE • Patenting • SR feeds Entrepreneurs with requests from large corporations • SR helps sell the prototype to large corporations (US/EMEA/Japan) Silicon Recognition, 2002 - Company Confidential
May Auckland Remain Its Home! Silicon Recognition, 2002 - Company Confidential