BRN Discussion Ongoing

[Diogenese]

BrainChip Akida Cloud

Benchmark Akida’s Low-Power AI in the Cloud No hardware required. Test, benchmark, and validate Akida’s sparsity-driven AI in minutes. Request Your Akida Cloud Trial Why Akida Cloud? Fast, Hardware-Free Evaluation Akida Cloud provides direct access to BrainChip’s Akida 2 platform...

brainchip.com

View attachment 89325
View attachment 89326

Hi TTM,


Re my post on Monday, these are the models which BRN provides:

• #103,092

Ready to use:

AkidaNet Object Detection/Recognition

AkidaNet Regression (Facial Age prediction)

AkidaNet/KWS

AkidaNet/Point Cloud Classification

AkidaNet/TENN Gesture Recognition (DVS)

AkidaNet/TENN Eye Tracking (90% activation sparsity)



ADVANCED TENNS MODELS – Available to run on Akida 2 FPGA Developer Platform – By Request

AkidaNet/TENN Audio Denoising

AkidaNet/TENN Automatic Speech Recognition (ASR)

AkidaNet/TENN LLM

AkidaNet/TENN LLM with RAG

In addition, other open source and customer proprietary models can be used. The change to 8-bit in Akida 2 makes it compatible with a whole new range of models.

The really exciting thing is that this makes TENNs immediately available for customers to test. Of course, because it's on-line, the real latency benefits will be masked, but maybe the system can differentiate the on-line delay.

ASR is a major selling point, as is eye tracking, gesture recognition, ... in fact they virtually all have major market potential.

For the customer/3rd party open source models, this will give the customers the opportunity to fine tune and optimize their models before committing to silicon. So, for example, they may determine the number of layers, the number of nodes per layer, the weight allocation/synapse connexions, ... so they will be able to hit the ground running when they get their "personalized" Akida 2 chips.

 

[Space Cadet]

Found this quite interesting. If we were to be included in the Golden Dome project, we would have to fit in here somewhere. They d9 mention neuromorphic sensors.

https://brianlantz.substack.com/p/golden-dome-a-manhattan-project?utm_campaign=post&utm_medium=web

SC

 

[IloveLamp]

Using generative AI to create synthetic training data for Edge AI models is a rapidly advancing technology. | Edge Impulse (a Qualcomm company)

Using generative AI to create synthetic training data for Edge AI models is a rapidly advancing technology. Have a look at this state-of-the-art experimentation with NVIDIA AI Cosmos Predict 2 to make industrial and manufacturing imagery that can be labeled and ingested into Edge Impulse. Nice...

www.linkedin.com