• Menu
  • Skip to main content
  • Skip to primary sidebar

All Tech News

Latest Technology News

AllTech.News

World’s fastest optical neuromorphic processor

You are here: Home / Computers and Smartphones / World’s fastest optical neuromorphic processor

An global group of researchers led by Swinburne University of Technology has demonstrated the world’s quickest and most potent optical neuromorphic processor for artificial intelligence (AI), which operates quicker than 10 trillion operations per 2nd (TeraOPs/s) and is capable of processing extremely-massive scale facts.

Released in the journal Nature, this breakthrough represents an tremendous leap ahead for neural networks and neuromorphic processing in common.

Artificial neural networks, a key type of AI, can ‘learn’ and carry out elaborate operations with vast programs to laptop eyesight, all-natural language processing, facial recognition, speech translation, enjoying strategy game titles, medical analysis and numerous other places. Impressed by the biological framework of the brain’s visible cortex method, artificial neural networks extract crucial options of raw data to forecast properties and behaviour with unprecedented precision and simplicity.

Led by Swinburne’s Professor David Moss, Dr Xingyuan (Mike) Xu (Swinburne, Monash University) and Distinguished Professor Arnan Mitchell from RMIT College, the workforce obtained an excellent feat in optical neural networks: radically accelerating their computing velocity and processing energy.

The team shown an optical neuromorphic processor functioning far more than 1000 situations speedier than any earlier processor, with the system also processing record-sized extremely-big scale illustrations or photos — plenty of to realize complete facial impression recognition, a thing that other optical processors have been not able to complete.

“This breakthrough was attained with ‘optical micro-combs’, as was our earth-document internet information speed noted in May 2020,” states Professor Moss, Director of Swinburne’s Optical Sciences Centre and not too long ago named 1 of Australia’s best study leaders in physics and mathematics in the area of optics and photonics by The Australian.

Although condition-of-the-art electronic processors these types of as the Google TPU can operate beyond 100 TeraOPs/s, this is performed with tens of 1000’s of parallel processors. In distinction, the optical system shown by the team takes advantage of a single processor and was obtained employing a new approach of concurrently interleaving the information in time, wavelength and spatial proportions via an built-in micro-comb supply.

Micro-combs are fairly new gadgets that act like a rainbow built up of hundreds of substantial-high-quality infrared lasers on a single chip. They are a lot a lot quicker, lesser, lighter and less costly than any other optical source.

“In the 10 years given that I co-invented them, built-in micro-comb chips have develop into enormously critical and it is actually exciting to see them enabling these enormous innovations in info communication and processing. Micro-combs offer tremendous assure for us to satisfy the world’s insatiable require for info,” Professor Moss claims.

“This processor can serve as a universal ultrahigh bandwidth entrance finish for any neuromorphic hardware — optical or digital primarily based — bringing significant-data device finding out for authentic-time ultrahigh bandwidth data within get to,” states co-guide writer of the review, Dr Xu, Swinburne alum and postdoctoral fellow with the Electrical and Computer system Units Engineering Office at Monash College.

“We are presently obtaining a sneak-peak of how the processors of the foreseeable future will seem. It can be really showing us how radically we can scale the electric power of our processors through the innovative use of microcombs,” Dr Xu explains.

RMIT’s Professor Mitchell adds, “This technology is applicable to all sorts of processing and communications — it will have a huge affect. Long phrase we hope to realise absolutely built-in devices on a chip, greatly reducing value and power consumption.”

“Convolutional neural networks have been central to the synthetic intelligence revolution, but current silicon technology increasingly provides a bottleneck in processing speed and power performance,” says critical supporter of the exploration crew, Professor Damien Hicks, from Swinburne and the Walter and Elizabeth Hall Institute.

He provides, “This breakthrough displays how a new optical technology makes this kind of networks more quickly and extra productive and is a profound demonstration of the gains of cross-disciplinary thinking, in possessing the inspiration and braveness to choose an idea from one particular field and utilizing it to address a fundamental difficulty in an additional.”


Some parts of this article are sourced from:
sciencedaily.com

Previous Post: « Threatpost Poll: Weigh in on Ransomware Security
Next Post: Masks no obstacle for new NEC facial recognition system »

Reader Interactions

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

Primary Sidebar

Recent Posts

  • U.S. Sanctions North Korean Andariel Hacker Behind Fraudulent IT Worker Scheme
  • How To Automate Ticket Creation, Device Identification and Threat Triage With Tines
  • Chinese Hacker Xu Zewei Arrested for Ties to Silk Typhoon Group and U.S. Cyber Attacks
  • Microsoft Patches 130 Vulnerabilities, Including Critical Flaws in SPNEGO and SQL Server
  • Hackers Use Leaked Shellter Tool License to Spread Lumma Stealer and SectopRAT Malware

Copyright © 2025 ยท AllTech.News, All Rights Reserved.