BRN Discussion Ongoing

Jumpchooks

Regular
Everything old is new again:

View attachment 16769

Everything old is new again:

View attachment 16769
My father had at truck with a mechanical lever arm with hand turn and stop indicator.. To indicate turn right, raise the arm. To indicate raise the arm and pull the knob and the hand would turn up and represent a stop signal. In 1964 it was the only thing he would allow me to operate. ( I secretly learned how to turn on the headlights and flatten the battery, sorry dad ! )
 
  • Haha
  • Like
  • Love
Reactions: 11 users

Diogenese

Top 20
  • Haha
Reactions: 5 users

chef

Emerged
howdy folks, just a random question from a curious software guy. what kind of hardware adjustments need to be made for the LSTM supported chip? why is the existing chip not able to fullfill LSTM?
 
  • Like
Reactions: 2 users

Jumpchooks

Regular
Cheers Mate thanks
Some of us are a bit slower than others on the information uptake
but if I was to remove myself from the competitive race to neuromorphic supremacy
I would choose Akida to control my switching requirements.

Aside from tech advancement in vehicles, one problem is electronic fault and damage repair.
My beautiful 2016 3.0TDi Audi A5 Sline sports hatchback (with every current advanced feature such as active lane keeper etc)had to be written off by insurer recently after impacting a large rock to the undercarriage because of an intermittent electronic fault that no repairer was confident to guarantee.
Resulting in an unsustainable waist of a car being devalued to scrap for parts.:(
Same thing happened to a mate with a new Jaguar.:(
Same thing happens with white goods and consumer electronics where a faulty circuit means throwing an item away onto a bigger and bigger heap. 😟
I think products need improved error code analysis to detect the exact faulty parts for replacement and maybe Akida is that missing link.
I have similar experience with fancy High Tech Brand 4wds in remote Australia (Brand Anonymous). It is a massive problem that the development of high tech vehicles , tools and appliances have a virtual non-existent support base. If you are young and tech savvy and want a carrier in service industry, the world is your oyster.
 
  • Love
  • Like
Reactions: 2 users

Jumpchooks

Regular
  • Haha
Reactions: 1 users

Mugen74

Regular
But surely the ad would only show the back of your head because of professional etiquette.
And they would have to change your name too probably something generic like Rob I guess🤣
 
  • Haha
  • Like
Reactions: 6 users

Diogenese

Top 20
howdy folks, just a random question from a curious software guy. what kind of hardware adjustments need to be made for the LSTM supported chip? why is the existing chip not able to fullfill LSTM?
Hi chef,

I would guess that LSTM would require additional memory in the NPUs so motion can be predicted by extrapolation from recent snapshots. Possibly some sort of FIFO arrangement?

I don't know if this would involve some participation (software/firmware) from the MCU/CPU, which is not required when the SNN is just doing classification/on-chip learning, but it's possible that it would.

PS: Can't wait for the patent to be published.

PPS: I believe the company will keep its cards close to its chest until the patent is published.
 
Last edited:
  • Like
  • Fire
  • Love
Reactions: 31 users
  • Like
  • Fire
Reactions: 7 users

Diogenese

Top 20
Now here's a funny thing - a company called Digimarc makes reference to Akida in a few of is patent applications:

US2021390358A1 LASER MARKING OF MACHINE-READABLE CODES

[0268] Still another type of processor hardware is a neural network chip, e.g., the Intel Nervana NNP-T, NNP-I and Loihi chips, the Google Edge TPU chip, and the Brainchip Akida neuromorphic SOC.
[0269] Software instructions for implementing the detailed functionality on the selected hardware can be authored by artisans without undue experimentation from the descriptions provided herein, e.g., written in C, C++, Visual Basic, Java, Python, Tcl, Perl, Scheme, Ruby, Caffe, TensorFlow, etc., in conjunction with associated data
.

... but the general description in the patent specification does not betoken a deep understanding of Akida' operation.
 
  • Like
  • Fire
  • Love
Reactions: 23 users

Diogenese

Top 20
howdy folks, just a random question from a curious software guy. what kind of hardware adjustments need to be made for the LSTM supported chip? why is the existing chip not able to fullfill LSTM?
Well now that you've got me thinking about NPU architectures, MemRistors/ReRAM are architecturally more similar to wetware compared to Akida, but Akida's N-of-M coding is functionally closer to the operation of grey matter, and this feature contributes more to the improved performance than MemRistor's architectural verity.
 
  • Like
  • Fire
  • Love
Reactions: 14 users

Cardpro

Regular
I think most companies are not ready for / don't need brainchip's advanced technology... we should really aim for the industry that actually requires the latest technology - mobile phone manufacturers, drones, military, car manufacturers, robotic, space, etc.

Oh wait, we are probably somewhat engaged to most of industries (if not all) mentioned above already IMO :cool::cool::cool::cool::cool::cool:
 
  • Like
  • Love
  • Haha
Reactions: 15 users

Pepsin

Regular
Nice momentum for the BRN-website!
https://www.similarweb.com/de/website/brainchip.com/#ranking
1663416323226.png

1663416348615.png


Do you have any idea why BRN is getting so much attention by the US when it´s only traded as ADR with a lower volume (I didn´t check the volume) there? The interest-increase shown on the right states that interest in Germany and USA grew and in relation declined in Australia.
May it be because of media articles published? For Germany there were some articles (posted by Sirod, i guess). Didn´t see anything similar for the US. Any other explanation? Custumer uptake wouldn´t affect the statistics, right?

Greetings from Germany!
 
  • Like
  • Fire
  • Love
Reactions: 39 users

Diogenese

Top 20
Nice momentum for the BRN-website!
https://www.similarweb.com/de/website/brainchip.com/#ranking
View attachment 16838
View attachment 16839

Do you have any idea why BRN is getting so much attention by the US when it´s only traded as ADR with a lower volume (I didn´t check the volume) there? The interest-increase shown on the right states that interest in Germany and USA grew and in relation declined in Australia.
May it be because of media articles published? For Germany there were some articles (posted by Sirod, i guess). Didn´t see anything similar for the US. Any other explanation? Custumer uptake wouldn´t affect the statistics, right?

Greetings from Germany!
Well we did appoint some publicity gurus. Don't have the date for this, but it dates back to LdN. Still, it would have been difficult to get traction until there was some product out there (Mercedes?)

https://brainchip.com/brainchip-acc...ic-system-on-chip-to-market-brainchip-031019/
SAN FRANCISCO–(BUSINESS WIRE)–
BrainChip Holdings Ltd. (ASX: BRN), a leading provider of ultra-low power, high performance edge AI technology, today announced that it has retained the services of technical industry analyst firm ActualTech Media and global public relations and marketing experts JPR Communications. Both firms will help position the company and its revolutionary new Akida spiking neural network (SNN) processor technology that is designed to provide a complete ultra-low power, high performance AI Edge Network.

...

Businesses are becoming aware of the potential impact that AI enabled technologies will have on their ongoing success,” said Scott D. Lowe, CEO and co-founder of ActualTech Media. “The challenge is that these organizations may not see the complete picture of just how game changing AI is going to be and how they can successfully deploy such solutions to their benefit. We look forward to working with BrainChip to help IT decision makers not only understand why they need to seriously consider deploying AI enabled technologies from the core to the edge but why BrainChip’s approach with Akida is the one that will yield the most productive results.”

JPR Communications is an experienced technology public relations and consulting agency with deep knowledge of the IT market and its influencers, as well as the demonstrated expertise to develop and implement highly effective communications programs that will ensure that BrainChip’s message is delivered across multiple offline and online platforms to reach their intended audiences. The company’s proven methodology will ensure that Akida is introduced to the market and exposed to buyers in a compelling, distinctive way that generates maximum impact.

“We are excited to be working with BrainChip to launch its revolutionary Akida product into the marketplace, helping to secure the industry and media exposure that will help positively position it as the de facto solution for those wanting better efficiency, higher performance and lower power requirements out of their AI devices,” said Mark Smith, Executive Vice President and Partner at JPR Communications. “We have a long history of launching the most disruptive technology companies in the world and look forward to sharing with analysts, editors, influencers, and decision makers on why BrainChip’s approach with spiking neural networks will deliver the best results and drive new use cases never before thought possible
.”

So they have been beavering away behind the scenes for a couple of years.

Chris Stevens didn't officially come on board until 6 September. Maybe he was captivated by the subliminal publicity campaign.

Late edition:

https://brainchip.com/brainchip-fea...as-emerging-ai-ml-and-data-science-ecocast-2/

BrainChip Featured in ActualTech Media’s Emerging AI/ML and Data Science EcoCast​

SAN FRANCISCO, Feb. 25, 2020 (GLOBE NEWSWIRE) — BrainChip Holdings Ltd. (ASX: BRN), a leading provider of ultra-low power, high performance edge AI technology, will present to IT professionals in the upcoming EcoCast webinar “Supporting Emerging AI, ML, and Data Science Initiatives,” February 27, 2020. The EcoCast, a live in-depth demonstration and discussion, addresses business and technology initiatives in emerging AI, ML, and data science.

The presentation by BrainChip COO Roger Levinson will include BrainChip’s Akida™ that brings highly efficient AI processing capabilities to edge devices such as motion sensors, medical devices, automotive systems, or drones. Joining BrainChip are presenters from data science platform Anaconda and Run:AI, a provider of virtualization layer for deep learning training models
.
 
  • Like
  • Fire
  • Love
Reactions: 34 users

Dirac

Member
  • Like
  • Sad
Reactions: 3 users

Diogenese

Top 20
Why would you go there?
What has it to do with BRN?
Why would you get exercised about that?

Sometimes when you come in at the tail end of a discussion you miss the finer nuances:

 
  • Like
  • Fire
  • Haha
Reactions: 14 users

Sirod69

bavarian girl ;-)
Have found this on facebook, but I found no link to it

1663419638468.png
 
  • Like
  • Fire
Reactions: 11 users

Diogenese

Top 20
  • Haha
  • Like
Reactions: 10 users

Dirac

Member
  • Like
Reactions: 4 users

chef

Emerged
Well now that you've got me thinking about NPU architectures, MemRistors/ReRAM are architecturally more similar to wetware compared to Akida, but Akida's N-of-M coding is functionally closer to the operation of grey matter, and this feature contributes more to the improved performance than MemRistor's architectural verity.
well thank you for your response! I had some similar ideas. what is meant by n-of-m coding? do you mean the tensor? I guess they need at least some additional memory and units (sigmoid & MAC) in each NPU to not fall back to CPU and ensure performance and parallelisation. thus I guess I understood the problem why they have to build a completly new chip. thanks!
 
  • Like
Reactions: 5 users

Diogenese

Top 20
well thank you for your response! I had some similar ideas. what is meant by n-of-m coding? do you mean the tensor? I guess they need at least some additional memory and units (sigmoid & MAC) in each NPU to not fall back to CPU and ensure performance and parallelisation. thus I guess I understood the problem why they have to build a completly new chip. thanks!
Hi chef,

N-of-M coding is based on the realization by Simon Thorpe's group that the strongest signals are recognized earlier by the retinal photoreceptors of the eye, and by the pixels of a photodetector, and thus generate spikes earlier than the weaker light signals. This means that most of the visual information is carried in the spikes which arrive earlier, and that the later-arriving spikes can be ignored with little loss of accuracy.

the following slides are from ST's presentation

1663423854861.png



but I don't have the link.

1663423679330.png



1663423782885.png


The 4-out-of-16 is just an example as there can be many more than 16 synaptical connexions to a neuron.

Akida does not do MAC.


Late edition:

This patent application, soon to be granted in US, is useful to understand the Akida NPU:

WO2020092691A1 AN IMPROVED SPIKING NEURAL NETWORK

1663424986487.png
 
Last edited:
  • Like
  • Fire
  • Love
Reactions: 26 users
Top Bottom