Neuromorphic Computing
Today's personal-assistant services, such as Microsoft's Cortana, Google's OK Google and Apple's Siri, are limited because they must call out to the cloud for more powerful computers to answer or anticipate queries. According to a Canadian start-up, neuromorphic computing may hold the answer to these issues.
A new family of circuit building blocks for processing analog signals using fully digital components, can now be incorporated to provide the long sought after benefits of neuromorphic computing according to the firm Circuit Seed.
Neuromorphic computing is a concept developed in the late 1980’s by Carver Mead describing the use of very-large-scale integration (VLSI) systems containing electronic analog circuits to mimic neuro-biological architectures present in the nervous system. Greg Waite, CEO of InventionShare, announced recently that Circuit Seeds’s family of inventions will make Mead’s vision viable by processing analog signals in digital components, thereby gaining the benefits of advanced digital processes that analog does not have.
"What we need now is for hardware to catch up to software and emulate human processing through analog in digital circuitry."
Neuromorphic chips are so named because they are modeled on the brain and nervous system and are designed to process sensory data such as images and sound, and to respond to changes in that data in ways not specifically pre-programmed for. Neuromorphic processors also hold the promise of being much more energy efficient than other methods.In order to accelerate the progress in artificial intelligence and lead to machines that are able to understand and interact with the world in humanlike ways, many point to the need for neuromorphic computation models.
New medical sensors and devices could track vital signs and response to treatments over time, learning to adjust dosages or even catch problems early. Or you could have your smartphone learn to anticipate what you want next, such as background on someone you’re about to meet and, when you meet that person, interpreting their body language and verbal responses as a second opinion of how that meeting went.
Other examples could be more alert or physical action oriented, such as reacting to dozens of variables and recommending the best course of action as in emergency medical surgery, or taking control over your car if it detects human failure or an impaired driver state.
Related articles
Continuing to improve the performance of such processors requires that manufacturers include more applications and provide ever faster transistors, silicon memory caches, and data pathways onto a chip - but the sheer heat generated by all those components is limiting how fast chips can be operated, especially in power hungry mobile devices. This has slowed the development of new devices to effectively process images, sound, and other sensory information and tasks such as face recognition and robot or vehicle navigation. Beyond that challenge is the cognitive processing required.Speaking from his Ottawa office, Waite said Circuit Seed is significant and a game changer for companies interested in the neuromorphic chip space. Circuit Seed’s new foundation inventions can reduce the size, heat, and power consumption required in the chip design. Circuit Seed circuits can also increase sensitivity and accuracy, reduce the number of circuits needed, and run at digital processing speeds.
Waite said that today’s personal-assistant services, such as Microsoft’s Cortana, Google’s OK Google and Apple’s Siri, are limited because they must call out to the cloud for more powerful computers to answer or anticipate queries. Companies are running up against walls.
“What we have now is semiconductor companies looking for the key to the future to support manufacturers’ ever increasing demands for chips that are smaller, lighter, faster – products that consume less power, give off less heat and that are easier, quicker and less expensive to design, manufacture and test. They want products with fewer parts that offer higher reliability and lower total product costs and that take up less real estate. What we need now is for hardware to catch up to software and emulate human processing through analog in digital circuitry.”
According to the company's press release, Circuit Seed inventions are founded on a mixture of decades of experience and creativity, supported by a robust portfolio of IP. Circuit Seed introduces a dramatic shift in the existing way of designing analog circuits, effectively eliminating many of the known shortcomings.
Benefits include: unprecedented precision without precision parts; ultra-linear operation over an extreme dynamic range (>106); low noise, ultrahigh speed (analog at logic speed) and extreme low voltage (down to ~0.1V power supply); process parameter independence; compact footprint (amplifiers about the size of a couple of NAND gates); low power (~100uW for multi-GHz-range operation down to 10pW for KHz-range operation); and portable designs between ultra-deep sub-µm CMOS technology process nodes. This paradigm shift has eliminated the use of current mirrors or differential matched pairs with traditional bulky analog transistors while being able to recover from sleep mode at logic speed.
So far, simulation and silicon test results have demonstrated a dramatic reduction in power, better performance and higher accuracy, all with small integrated circuit footprints. The circuit designs are much simpler, reducing product costs with faster development, so less testing is required and there are fewer parts to assemble. These circuits are far more reliable in the newest CMOS processes, making Circuit Seed a real breakthrough for companies in the neuromorphic chip space.
InventionShare is now looking for strategic partners who want to incorporate Circuit Seed technology into their neuromorphic research projects or for more near term product offerings, with the benefit of leveraging market share, profitability through competitive feature set and manufacturing advantages.
0 comments:
Post a Comment