When you purchase through link on our site , we may earn an affiliate commission . Here ’s how it work .
Scientists at Intel have built the existence ’s largest neuromorphic computer , or one designed and structured to mimicthe human brain . The company hop it will support future artificial intelligence ( AI ) research .
The machine , dub " Hala Point , " can perform AI work load 50 times faster and use 100 metre less energy than conventional computer science system that use cardinal processing unit ( CPUs ) and graphics processing unit ( GPUs ) , Intel representatives read in astatement . These figures are base on findings uploaded March 18 to the preprint serverIEEE Explore , which have not been peer - reviewed .
Hala Point will ab initio be deployed at Sandia National Laboratories in New Mexico , where scientists will use it to tackle problem in machine physics , cypher architecture and computer skill .
Related : China develops new twinkle - based chiplet that could power artificial general tidings — where AI is smarter than humans
Powered by 1,152 of Intel ’s newfangled Loihi 2 processors — a neuromorphic research check — this tumid - scale system be 1.15 billion unreal nerve cell and 128 billion contrived synapsis distributed over 140,544 processing core .
It can make 20 quadrillion operation per second — or 20 petaops . Neuromorphic computer litigate information otherwise from supercomputer , so it ’s laborious to compare them . But Trinity , the38th most powerfulsupercomputer in the world shoot a line approximately 20 petaFLOPS of power — where a FLOP is a floating - point cognitive operation per moment . Theworld ’s most powerful supercomputeris Frontier , which boasts a performance of 1.2 exaFLOPS , or 1,194 petaFLOPS .
How neuromorphic computing works
Neuromorphic computer science differ from formal computing because of its computer architecture , Prasanna Date , a estimator scientist with the Oak Ridge National Laboratory ( ORNL ) , wrote onResearchGate . These types of calculator practice neural networks to build the machine .
In classic computing , binary bits of 1s and 0s flow into ironware like the CPU , GPU or memory board before swear out calculations in succession and spitting out a binary output .
In neuromorphic computation , however , a " spike remark " — a set ofdiscrete electrical signal — is feed into the spiking neural internet ( SNNs ) , represented by the processors . Where software - based neural networks are a collection of machine learning algorithms arranged to mime the human brain , SNNs are a physical avatar of how that information is transmitted . It permit for parallel processing and spike outputs are measured following calculations .
Like the mastermind , Hala Point and the Loihi 2 processors use these SNNs , where dissimilar nodes are connected and information is process at different layers , standardised to neurons in the brain . The chips also integrate memory and computing power in one place . In conventional information processing system , process power and memory are separated ; this create a bottleneck as information must physically travel between these components . Both of these enable parallel processing and dilute power consumption .
Why neuromorphic computing could be an AI game-changer
former upshot also show that Hala Point accomplish a high vigour efficiency study for AI workload of 15 trillion operations per watt ( TOPS / W ) . Most ceremonious neural processing units ( NPUs ) and other AI systems reach well under10 TOPS / W.
Neuromorphic computation is still a arise field , with few other machines like Hala Point in deployment , if any . Researchers with the International Centre for Neuromorphic Systems ( ICNS ) at Western Sydney University in Australia , however , annunciate program to deploy a like machinein December 2023 .
— New brain - same transistor break down ' beyond motorcar encyclopaedism '
— New deoxyribonucleic acid - infused computing equipment chip can perform calculations and make future AI example far more effective
— World ’s enceinte computer chip shot WSE-3 will power monumental AI supercomputer 8 times faster than the current record - bearer
Their computing machine , called " DeepSouth , " emulates big connection of spiking neuron at 228 trillion synaptic operations per minute , the ICNS researcher say in the affirmation , which they say was equivalent to the rate of operations of the human brain .
Hala Point meanwhile is a " starting decimal point , " a research prototype that will finally fertilise into future system that could be deployed commercially , concord to Intel representatives .
These succeeding neuromorphic computers might even direct to large language models ( LLMs ) like ChatGPT learning ceaselessly from new data point , which would reduce the massive preparation burden inherent in current AI deployments .
scientist clear major barricade in mission to build hefty AI photonic chips
Scientists find how to use your body to process information in wearable devices
Could a planet really spring up a mastermind ?
