Intel unveiled the world’s largest neuromorphic system, Hala Point, on April 18, 2024. This technological marvel boasts of 1.15 billion artificial neurons, roughly equivalent to the processing power of an owl brain or a capuchin monkey’s cortex. Initially deployed at Sandia National Laboratories, Hala Point paves the way for groundbreaking research in brain-inspired artificial intelligence.
Here’s what the Director of Intel Labs has to say on this breakthrough innovation:
“The computing cost of today’s AI models is rising at unsustainable rates. The industry needs fundamentally new approaches capable of scaling. For that reason, we developed Hala Point, which combines deep learning efficiency with novel brain-inspired learning and optimization capabilities. We hope that research with Hala Point will advance the efficiency and adaptability of large-scale AI technology.”
– Mike Davies, Director (Neuromorphic Computing Lab, Intel Labs)
Hala Point is a part of Intel’s broader effort to make AI more sustainable. It highlights the potential of neuromorphic computing for reducing the energy consumption of data-intensive operations. This powerful system has been used for brain-scale computing research at Sandia National Laboratories, serving as a testbed for future AI models. It significantly boosts neuron capacity and performance in comparison to its predecessor, Pohoiki Springs.
Loihi 2 neuromorphic processors form the basis for Hala Point. These brain-inspired chips use a special type of network where neurons communicate directly with one another instead of communicating through memory and hence significantly reduce power consumption.
The patent, WO2024076823A1 describes a method of communication between devices which appear relevant to Hala Point’s architecture. Specifically, the way the patent discloses the information transfer and mapping between devices could be similar to how Loihi 2 processors communicate.
To conclude, Intel’s Hala Point leverages Loihi 2 neuromorphic chips, represents a pioneering approach in neuromorphic computing, aiming to enhance AI and IoT efficiency. Hala Point distinguishes itself from other ongoing similar projects such as BrainScaleS and HP’s Machine Project with advanced on-chip learning capabilities claiming to solve optimization problems using 100 times less energy at speeds which are 50 times faster than traditional GPU and CPU architectures, setting new standard in neuromorphic technology applications.
Here are some interesting articles you can explore to learn more about Intel’s patents. Click here to access them now.
Need to know anything else? We got you covered!