I stumbled upon this article explaining neuromorphic AI:
"Combined with the high costs and complexity of the technology, the accuracy issue leads many to prefer traditional software."
"Neuromorphic computing software has not caught up with the hardware."
"Neuromorphic computers aren't available to nonexperts. Software developers have not yet created application programming interfaces, programming models and languages to make neuromorphic computers more widely available."
So I felt compelled to write to the guy:
Dear Nick Barney,
It´s always nice to see somebody explaining neuromorphic computing and I think it´s very relevant. Conventional AI models are hitting a brick wall of exponential increasing needs of processing power, that neuromorphic computing can solve. Also considering our environment, neuromorphic computing is a much more future proof solution.
I believe that you´re not aware of a fully working and commercially available neuromorphic chip called Akida from Brainchip. It has a working IDE and can accept existing DNN models. There´s even an Akida 2 on the horizon.
It will be sold primarily as IP licenses for a few dollars for a complete chip and can be licensed as IP blocks for individual parts of the design, but can also be purchased as a a PCIe board, a Shuttle computer or a Raspberry Pi kit:
Specifications ARM Cortex-M4 32-bit 300MHz (Subblock of Akida) RAM: 256M x 16 bytes LPDDR4 SDRAM @ 2400MT/s FLASH: Quad SPI 128Mb NOR 12.5MHz Onboard Akida core current monitor Operates under Linux on arch or x 86-64 architectures Installation and Bringup GPIO: 2 LED's Interfaces: 5GT/s PCI...
shop.brainchipinc.com
Specifications SoC: Broadcom BCM2711C0 quad-core ARM Cortex-A72 (ARMv8-A) 64-bit @ 1.5GHz GPU: Broadcom VideoCore VI RAM: 1GB, 2GB, 4GB, or 8GB LPDDR4 SDRAM Networking: GB Ethernet and Optional WI-FI Includes Akida on PCle Board installed in the development kit Includes Meta TF Software...
shop.brainchipinc.com
Specifications Storage – SATA / M.2 1x 2.5" bay for SATA hard disk or SSD, max. 9.5 mm 1x M.2-2280M slot (supports PCIe x4 NVMe or SATA) 1x M.2-2230AE for an optional WLAN card Connectors HDMI 2.0a D-Sub VGA 4x USB 3.2 Gen1 4x USB 2.0 1x USB 2.0 internal USB stick 2x audio (line out, mic) Intel...
shop.brainchipinc.com
Further amongst the very compelling advantages of Akida is the one-shot-learning on device, I´m not aware of any other system that can do this.
Brainchip contracts with MegaChips and Rensas, further they are recently incorporated into the ARM ecosystem. It´s already integrated into several prototypes, one of them is the new Mercedes Vision EQXX, but you can study the included chart.
So neuromorphic computing is in production, available, powerful and being adopted by the market place.
Best regards,
Frederik Grøn Schack