In September 2017, Intel introduced Loihi, a “self-learning neuromorphic processor,” designed to accelerate and expand AI applications. Today the manufacturer spoke about the second generation of this microcircuit and the open-source Lava programming environment, which allows developing applications for neuromorphic processors.
“Loihi 2 and Lava have accumulated several years of research using Loihi. Our second-generation chip is faster and easier to write, which expands its use in power-constrained computing systems. We open source Lava to provide better software compatibility, testing and cross-platform interoperability in this area, as well as in order to accelerate the time for products to the level of commercial profitability, ”- Mike Davies, director of the Neuromorphic Computing Lab, described the new products. Intel.
It is believed that neuromorphic computing using specialized microcircuits, in which an attempt is made to reproduce the principles of the biological brain, will significantly increase the speed and energy efficiency of AI. We are talking about a wide range of tasks that arise during human-machine interaction, including computer vision, voice and gesture recognition, search, and optimization. Intel and its partners have already demonstrated some examples of such developments so far - robotic arms, neuromorphic artificial skin, and an odor recognition algorithm.
As for Loihi 2, improvements in the architecture enable it to work with new classes of neuromorphic algorithms and applications, providing a significant increase in processing speed - up to 10 times. The gain in resource density reaches 15 times. The Loihi 2 chip is manufactured using a pre-release Intel 4 process technology that uses EUV lithography.
Intel is currently offering the INRC Neuromorphic Research Cloud Neuromorphic Research Cloud Oheo Gulch with one Loihi 2 processor. The Lava Software Framework is available as a free download on GitHub.
0 Comments