The potential of ZISC chips was a hot topic at a neurocomputing conference in Auckland last week.ZISC — zero instruction set computer – processors recognise patterns in the same way the human brain does, according to Remi Vespa (pictured) of US company Silicon Recognition, which makes them.
ZISC chips contain learning logic, eliminating the need for programming, Vespa says. They also promise unlimited parallel processing potential, thereby providing unprecedented speed.
Those characteristics suit it to two types of applications: those requiring “smart sensing”, when ZISC is used in combination with a sensor, enabling decisions to occur at the sensor level; and applications requiring massively parallel pattern recognition, when data is broadcast to a theoretically infinite number of processing elements, enabling instant pattern recognition.
Applications might include biometric recognition of faces, fingerprints, palm prints, iris prints and voiceprints.
Vespa was one of a number of speakers at the two-day AUT-hosted event. The emphasis was on the application and commercialisation of neurocomputing technology. The event was organised by the newly established Knowledge Engineering and Discovery Research Institute, KEDRI.