BRN Discussion Ongoing

[JB49]

Much of the model building was probably done in conjunction with Edge Impulse, who have developed a method of automatically classifying items added to a model, as I understand it (I may be wrong - (there's always a first time)) basically using one-shot learning from pre-existing models. This may relate to "based on our own TENNs family of models".

@TECH , you referred to Akida 2 silicon - remember when Anil announced the tapeout, and then this was quietly shelved in favour of some mysterious "other" entity doing the job?

Drum roll ... !!!

The curtains part to reveal ... ???

... any day now ...

Hey Dio,

Are you just suspecting that is the case, or was it confirmed somewhere that Akida 2 tapeout was pulled due to somebody else undertaking it?

 

[Diogenese]

Hey Dio,

Or just suspecting that is the case, or was it confirmed somewhere that Akida 2 tapeout was pulled due to somebody else undertaking it?

There was a statement to that effect at the time - then it all went dark ...

 

[JB49]

Interesting

 

[FJ-215]

Much of the model building was probably done in conjunction with Edge Impulse, who have developed a method of automatically classifying items added to a model, as I understand it (I may be wrong - (there's always a first time)) basically using one-shot learning from pre-existing models. This may relate to "based on our own TENNs family of models".

@TECH , you referred to Akida 2 silicon - remember when Anil announced the tapeout, and then this was quietly shelved in favour of some mysterious "other" entity doing the job?

Drum roll ... !!!

The curtains part to reveal ... ???

... any day now ...

Didn't we drop that idea so that BRN would not be seen as a competitor to any future buyers of the IP?

Why waste money on the FPGA version if there is someone else on the job? So to speak

 

[Pom down under]

No podcast, things seem extremely quite on any real news from BRN ( which is normal anyway ) so are we in a blackout period and going to get our surprise after 3 years of promises . So are we in for a real price sensitive announcement any time soon? And then a release of the podcast to go with maybe the announcement.


How lucky can we be?


View attachment 73761

Maybe I was right

 

[JB49]

Also interesting, another person who congratulated Anil and the team.

The Clink VoiceBuds are designed to handle calls with impressive clarity, courtesy of their six HD microphones and Environmental Noise Cancellation (ENC) technology. This particular feature peaked my excitement, and I was looking forward to seeing how well the earbuds perform during audio calls. During office meetings, voices came through crisp and clear, ensuring seamless communication even when colleagues were speaking softly. Outdoor calls posed no challenges either, as the wind noise cancellation effectively reduced disruptions during a breezy walk on the beach or while standing at a busy street corner.

Bluetooth Audio SoC with DSP and NPU – The earbuds utilize a Bluetooth Audio System-on-Chip (SoC) that includes Digital Signal Processing (DSP) and a Neural Processing Unit (NPU). The DSP enhances audio quality, while the NPU helps in improving noise reduction during calls through machine learning algorithms.

Denoising by TENNs?

https://www.livemint.com/technology/tech-reviews/clink-audio-voicebuds-review-surprising-performance-with-an-affordable-price-tag-11732699640442.html

Clink Audio Enters Indian Market with New TWS VoiceBuds

Clink Audio’s VoiceBuds are introduced as a new option in the TWS market, focusing on features like advanced noise cancellation, high-fidelity audio, and extended battery life. With their launch, Clink Audio aims to offer a competitive choice for consumers seeking quality audio products.Key...

clinkaudio.com

 

[Diogenese]

Didn't we drop that idea so that BRN would not be seen as a competitor to any future buyers of the IP?

Why waste money on the FPGA version if there is someone else on the job? So to speak

Tapeout does not relate to FPGAs. This was the real McCoy SoC. FPGAs have all the logic circuits pre-installed and the interconnexions can be adjusted in the field.

As with Akida 1, "engineering samples" help to demonstrate the full potential which ican only be viewed through a glass darkly using the software simulation in MetaTF.

 

[ArtGonzo]

Has anyone else read into the bio of Brainchip's largest holder - Geoff Cumming? Quite a guy!

Geoff Cumming is a Canadian-born international businessman and philanthropist known for his significant contributions to medical research and philanthropy. Here are some key points about him:

• Early Life and Education: Born into a medical family in Kingston, Canada, Geoff Cumming pursued education in economics at the University of Calgary, The London School of Economics (LSE), and the University of British Columbia.
• Career:
  • He worked in tax and international economics within the Government of Alberta before moving to the Global Energy Group of the Royal Bank of Canada.
  • Cumming led the bank's venture capital group in Western Canada and later joined Peters & Co Limited as a partner.
  • He held positions like President and CEO of Gardiner Oil and Gas and its parent company, Gardiner Group Capital of Toronto, and founded international investment firms Emerald Capital and Karori Capital.
  • In the field of healthcare and biotechnology, he has held senior roles such as Managing Director of Roche Diagnostic Systems (Oceania), where he turned around a loss-making business, and CEO of companies like Biosceptre International and Anteo Diagnostics.
• Philanthropy:
  • Cumming made one of the largest donations in Canadian history by gifting over C$100 million to the University of Calgary for medical research, matched by the Government of Alberta.
  • He donated $250 million to the University of Melbourne for the establishment of the Cumming Global Centre for Pandemic Therapeutics in 2022, which is noted as the largest philanthropic donation to medical research in Australia's history.
  • His philanthropy also extends to supporting various international charitable organizations focused on public health, sanitation, environmental issues, and global over-population.
• Recognition:
  • He has been honored as a Fellow of the University of Melbourne in recognition of his philanthropy.
  • Cumming is active in global philanthropy, including establishing the Ryman Prize for advancements in elderly care quality of life.
  • He's been involved with numerous companies as a director or chairman, particularly in solar and energy technologies.
• Personal Life: Geoff Cumming lives in Melbourne, Australia, with his wife, Anna. They have two children.

His contributions and involvement in both business and philanthropy have significantly impacted medical research and health globally, particularly with initiatives aimed at preparing for future pandemics and improving human conditions.

If he's the top holder, we're in good company

 

[yogi]

"An impressive number of beyond SOTA models for edge inference ... based on our own TENNs family of models."

I read this as "We have a whole lot of customers lined up and working with us to adapt their, probably CNN, models to run on Akida2/TENNs."

"I think we might be able to schedule a funeral for CNN models for temporal processing."

Ubiquitous!

Yeah I like line too about funeral for CNN, So he or Brainchip is pumped, What does that funeral for CNN means now @Diogenese

 

[supersonic001]

BrainChip, the world’s first commercial producer of ultra-low power, fully digital, event-based, neuromorphic AI, today announced that it was awarded a development contract for $1.8M from Air Force… | BrainChip | 11 comments

BrainChip, the world’s first commercial producer of ultra-low power, fully digital, event-based, neuromorphic AI, today announced that it was awarded a development contract for $1.8M from Air Force Research Laboratory (AFRL) on neuromorphic radar signaling processing technologies. Read more...

www.linkedin.com

ONE More time for clarity and celebration … ! One step at a time !!!

 

[Smoothsailing]

Maybe I was right

Why wasn’t this price sensitive I wonder ?.

 

[supersonic001]

BrainChip, the world’s first commercial producer of ultra-low power, fully digital, event-based, neuromorphic AI, today announced that it was awarded a development contract for $1.8M from Air Force… | BrainChip | 11 comments

BrainChip, the world’s first commercial producer of ultra-low power, fully digital, event-based, neuromorphic AI, today announced that it was awarded a development contract for $1.8M from Air Force Research Laboratory (AFRL) on neuromorphic radar signaling processing technologies. Read more...

www.linkedin.com

ONE More time for clarity and celebration … ! One step at a time !!!

View attachment 74021

Also liking the jump in followers …. Last time i looked a few weeks back we were 16,600ish, now 17,300’s….. nice traction team….. !

 

[Diogenese]

Yeah I like line too about funeral for CNN, So he or Brainchip is pumped, What does that funeral for CNN means now @Diogenese

He specifically mentions "CNN models".

CNN models are used in Software NNs and in competing NN accelerator hardware.

CNN models are usually associated with MAC (multiply/accumulate) processing in full digital, 16 bit or 32 bit per byte processors, although the pressure to reduce energy and latency has seen 8 bits adopted as standard recently. While some zero multiplication can be skipped to reduce the number of multipliction operations, basically all the input bytes are processed.

MAC operations are carried out in a matrix array (rows and columns) where all the multiplication results are added column by column.

Akida on the other hand, only processes bytes (1, 2, or 4 bit events or spikes) which change.

In addition, Akida uses N-of-M coding, processing only the largest N of M input bytes as these N bytes carry the most relevant information, meaning that M-N events are dropped from the calculation.

Further more, data in Akida models is 1, 2, or 4 bit bytes and can also use N-of-M coding, greatly reducing the size of each model, so, in fact, you get a double-whammy of compression.

Thus Akida 1 is much more power efficient as it performs far fewer operations in less time to process the same input information as conventional CNNs.

Now TENNs beings additional efficiencies, which I haven't got to the bottom of yet (It took me a long time to get some sort of grasp of N-of-M coding). This makes TENNs more power efficient and further reduces latency.

So I think Tony is saying that CNNs are yesterdays tech, so make way for tomorrow.

 

[Tothemoon24]

HAW research cooperation with Mercedes-Benz AG is revolutionizing autonomous driving: Karlsruhe University of Applied Sciences (HKA)/University of Applied Sciences is further developing complex camera technologies in neuromorphic computing. New "event cameras" are at the heart of the project #EVSC the #HKA, which provide a dramatic improvement in temporal resolution. Unlike conventional cameras, event cameras dynamically perceive changes in your field of vision instead of taking an image at fixed intervals. This allows them to supply the autonomous driving system with new information within a few microseconds, while conventional image sensors are temporarily "blind".

With the project cooperation, Karlsruhe University of Applied Sciences is consistently pursuing
another step in helping to shape the future of autonomous
because Prof. Dr. Jan Bauer is sure that the use of
event cameras will significantly expand the ability of autonomous vehicles to capture the environment faster and more accurately. "And with the EVSC project, we can significantly improve the integration capability of this new camera technology into the vehicle," says the HKA researcher.

 

[schuey]

This announcement is, in my view, even more significant than when we mentioned NASA or when Mercedes referenced us a few years ago. As usual, the market seems slow to fully appreciate today's suprise annoucement..

1. This announcement represents a tangible contract backed by funding, signaling a direct entry point with immediate financial impact.

2. The collaboration with AFRL can potential to lead to broader engagements with the DoD. However, realizing this opportunity depends on executing the partnership effectively.

3. Lastly, the first move has been made, we should now *fingers crossed* see more now following.

This is a pivotal step forward, both financially and strategically for Brainchip, bring on 2025

Normally takes 24 hours to hit everyone

 

[Diogenese]

View attachment 74035

HAW research cooperation with Mercedes-Benz AG is revolutionizing autonomous driving: Karlsruhe University of Applied Sciences (HKA)/University of Applied Sciences is further developing complex camera technologies in neuromorphic computing. New "event cameras" are at the heart of the project #EVSC the #HKA, which provide a dramatic improvement in temporal resolution. Unlike conventional cameras, event cameras dynamically perceive changes in your field of vision instead of taking an image at fixed intervals. This allows them to supply the autonomous driving system with new information within a few microseconds, while conventional image sensors are temporarily "blind".

With the project cooperation, Karlsruhe University of Applied Sciences is consistently pursuing
another step in helping to shape the future of autonomous
because Prof. Dr. Jan Bauer is sure that the use of
event cameras will significantly expand the ability of autonomous vehicles to capture the environment faster and more accurately. "And with the EVSC project, we can significantly improve the integration capability of this new camera technology into the vehicle," says the HKA researcher.

Event cameras are great when the camera is stationary, but there's a whole lot of complexity when the camera is moving. In-cabin use may be ok except for the windows, but for the exterior you'd need some clever algorithms to account for camera motion.

 

[7für7]

Event cameras are great when the camera is stationary, but there's a whole lot of complexity when the camera is moving. In-cabin use may be ok except for the windows, but for the exterior you'd need some clever algorithms to account for camera motion.

You mean clever by akidalicous?

 

[Falling Knife]

Another interpretation of today's announcement:

The contract terms are as follows:

1. Total contract value: US$1.8 million to be paid by AFRL over the 12-month term of the agreement
2. Brainchip's share: The company will receive US$800,000 to develop and optimize algorithms during the 12-month contract term
3. Subcontractor involvement: Brainchip will collaborate with a subcontractor (likely an aerospace and defense company) to complete the contract


Therefore, the total payment to Brainchip from this contract is US$800,000, not US$2.6 million. The remaining portion of the US$1.8 million contract will likely go to the subcontractor and other project-related expenses.

The beauty of this ASX ann is that it will bring eyes to our company.
DoD do not select products or companies without due diligence. BRN tech is the future as stated by our CTO today with the death knell of CNN being predicted. Distinction or extinction nears for those who adopt or choose not to.

The systems, frameworks and collaborators involved in this neuromorphic disruptive tech are almost ready to show their hands.
All the work BRN team have been engage in to this point has been for this purpose....
The switch from CNN to TENNs + Akida is in our future.


A turning point in our companies future I suspect.

 

[JB49]

Just a quick reminder to everyone of 4 key points Sean made in the Stocks Down Under video. It was a very impromptu interview - Almost as if Sean was the one who reached out to them, so that he could include a message to shareholder to keep patient as there are good things just around the corner!

Point 1 below in bold is now complete. Hopefully he intended "things" to be plural with more to come soon!

Very keen to see where point 2 and 4 takes us!

1. We're executing exceptionally well on our roadmap plan and exceptionally well on our commercial elements aswell. So how do I feel about it, I feel very bullish and optimistic about it. I'm not backing off my statements that 25 will be a good year for us, and I anticipate some things even in the next couple of months that are all positive for Brainchip
2. We've had several request from customers for an exceptionally low configuration with a small footprint and high performance... Pico remains the ideal solution for them and interest remains incredibly high in Pico since its been introduced.
3. We've begun shipping edge boxes directly, as have our partners, with interest in audio, home security and IOT
4. The interest were seeing right now... a lot of strong interest in audio and quite frankly vision aswell, and anything around spatio-temporal and streaming data at the edge. We're seeing a lot of activity there and use cases developing there. Some use cases around IoT that are very power sensitive

 

[7für7]

Maybe I was right

Finally someone with a working
Any other suggestions? Your success rate seems reliable.