Camels being horses designed by committee, I'm not sure a consortium is the best way to go. Who would be the cat-herder-in-chief?
The company has decided that its main business model will be in the licensing of Akida IP for customers to incorporate in their chips, rather than selling chips as a product.
As we know, this is a much longer time to market than simply making chips.
However, along the way, the company has gained an understanding that each customer has unique requirements, so a one-size-fits-all chip is not appropriate.
Akida IP now comes in 3 customizable flavours, e, s, p which can be incorporated into customer SoC including a processor.
https://brainchip.com/akida-generations/
e has up to 2 nodes (4*NPEs), and can be used for ultra-low power applications such as:
- Vibration Detection
- Anomaly Detection
- Keyword Spotting
- Sensor Fusion
- Low-Res Presence Detection
- Gesture Detection
at up to 200 MHz.
s has up to 8 nodes, and can be used for application such as:
- Advanced Keyword Spotting
- Sensor Fusion
- Low-Res Presence Detection
- Gesture Detection & Recognition
- Object Classification
- Biometric Recognition
- Advanced Speech Recognition
- Object Detection & Semantic Segmentation
at up to 500MHz.
- all without CPU intervention after initial configuration.
p has up to 256 nodes (1024 NPEs), and can be used for applications such as:
- Gesture Detection
- Object Classification
- Advanced Speech Recognition
- Object Detection & Semantic Segmentation
- Advanced Sequence Prediction
- Video Object Detection & Tracking
- Vision Transformer Networks
Akida can be adapted for each different application by "programming" it with a model library of examples representative of the intended application.
The Akida neural network (NN) includes nodes of 4 Neural Processing Engines (NPE) interconnected by a switchable communication matrix so different arrangements of the NPEs can be configured to suit each different application. Usually the configuration will consist of three or more layers of NPEs, each with a selectable number of connexions with the NPEs in the following layer.
The configuration is "programmable" and not fixed. A processor (MCU, MPU, CPU, GPU) is used to set up the configuration but is not used for the basic classification/inference tasks, although it can also be used to utilize the output from the Akida silicon. The Akida IP will normally be manufactured as part of the same SoC as the processor.
Akida IP can be incorporated with any "standard" processor (processor agnostic).
We are in the "ecosystem" of ARM and IFS.
Renesas and MegaChips have Akida IP licences.
Prophesee and Valeo have products for which Akida is a natural fit.
VVDN is producing an Akida Edge Box.
Unigen and BrainChip are developing an Edge Server.
BRN is working with Lorser and separately Ipsilon on software defined radio
BRN is working with emotion3D and nViso on DMS.
In space, there's NASA, ANT61, EdgX.
The list goes on ...
Cybersecurity - Quantum Venture, DUTH.
RISC-V - SiFive.
Socionext - DMS, in-cabin radar ...
If we had any more irons in the fire, we'd need a blast furnace to keep them hot.
This was the BRN pre-CES announcement a year ago:
https://brainchip.com/brainchip-joi...rate-capabilities-of-akida-powered-solutions/
BrainChip Joins Technology Partners During CES to Demonstrate Capabilities of Akida-Powered Solutions
Laguna Hills, Calif. – December 29, 2022 –
BrainChip Holdings Ltd (ASX: BRN, OTCQX: BRCHF, ADR: BCHPY),
the world’s first commercial producer of ultra-low power, fully digital, event-based, neuromorphic AI IP, today announced that it will be joining partners Prophesee, Socionext, and VVDN January 5-8 at CES to showcase compelling solutions on constrained edge devices, featuring its Akida™ processors. Akida processors simplify development by supporting today’s mainstream network models and workflows while being future-proofed for next-generation edge AI solutions.
BrainChip joins Prophesee at The Venetian Hotel as part of a technology showcase highlighting Prophesee’s Metavision® platform. BrainChip partners with Prophesee to optimize computer vision AI performance and efficiency to deliver next-generation platforms for OEMs interested in integrating event-based vision systems with high levels of AI performance coupled with ultra-low power technologies.
BrainChip joins Socionext in the Vehicle Tech and Advanced Mobility Zone, located in the Las Vegas Convention Center, North Hall in Booth 10654. Socionext will be showcasing its leading-edge technologies and solutions in automotive imaging, AI, and smart sensing to help customers develop feature rich custom SoCs that enable product differentiation with an added competitive edge. Socionext has played an important role in the implementation of BrainChip’s AI technology in SoCs and the companies are working together to satisfy the need for AI in edge computing. As part of this showcase, BrainChip will be highlighting its functionality with NVISO’s human-behavior software designed for the automotive in-cabin environment.
VVDN Technologies has more than a decade of experience in product engineering and manufacturing electronic solutions. VVDN offers a value proposition by being a one-stop shop for customers by offering complete hardware, software, mechanical, testing, prototyping, certification, and manufacturing services. The company’s focus is to help customers with their next technology innovation in the space of 5G, Wireless, AI/ML on Vision-based solutions, IoT, Cloud and Apps. To enable the same, BrainChip will be highlighting its industry leading technology at the VVDN Technologies Booth 10870 at CES developing opportunities for next generation products.
“We are excited to join our partners in Las Vegas during CES to showcase how combining BrainChip’s Akida with their technologies delivers leading edge AI solutions,” said Rob Telson, BrainChip Vice President of Ecosystem and Partnerships. “We look forward to demonstrating these capabilities in person.”
BrainChip’s neural processor Al IP is an event-based technology that is inherently lower power when compared to conventional neural network accelerators. BrainChip IP supports incremental learning and high-speed inference in a wide variety of use cases, such as convolutional neural networks with high throughput and unsurpassed performance in micro watt to milli-watt power budgets. Among the markets that BrainChip’s Essential AI technology will impact are the next generation of smart cars, smart homes of today and tomorrow, and industrial IoT.
We've heard from VVDN and Socionext ... what's keeping Prophesee?