BRN Discussion Ongoing

[Learning]

Good evening,

Could this be BRN?

CHIP DEVELOPMENTDevelopment of ultra-low power on-device learning edge AI chip​

08.12.2022 From ROHM (press release)


ROHM has developed an on-device learning AI chip (SoC with on-device learning AI accelerator) for edge computer endpoints in the IoT field. It utilizes artificial intelligence to predict failures (predictive failure detection) in electronic devices equipped with motors and sensors in real-time with ultra-low power consumption.

View attachment 33011

Combining the 20,000-gate ultra-compact AI accelerator with a high-performance CPU enables learning and inference with ultra-low power consumption of just a few tens of mW (1000× smaller than conventional AI chips capable of learning).

Going forward, ROHM plans to incorporate the AI accelerator used in this AI chip into various IC products for motors and sensors. Commercialization is scheduled to start in 2023, with mass production planned in 2024.

ROHM Develops Ultra-Low-Power On-Device Learning Edge AI Chip | ROHM Semiconductor - ROHM Co., Ltd.

www.rohm.com

Hi Serengeti,

Our more knowledgeable members would give you a more definitive answer.

However, reading this,

"Based on an ‘on-device learning algorithm’ developed by Professor Matsutani of Keio University,"

reading the above, I believe it's not Akida.

Learning 🏖

 

[chapman89]

Good evening,

Could this be BRN?

CHIP DEVELOPMENTDevelopment of ultra-low power on-device learning edge AI chip​

08.12.2022 From ROHM (press release)


ROHM has developed an on-device learning AI chip (SoC with on-device learning AI accelerator) for edge computer endpoints in the IoT field. It utilizes artificial intelligence to predict failures (predictive failure detection) in electronic devices equipped with motors and sensors in real-time with ultra-low power consumption.

View attachment 33011

Combining the 20,000-gate ultra-compact AI accelerator with a high-performance CPU enables learning and inference with ultra-low power consumption of just a few tens of mW (1000× smaller than conventional AI chips capable of learning).

Going forward, ROHM plans to incorporate the AI accelerator used in this AI chip into various IC products for motors and sensors. Commercialization is scheduled to start in 2023, with mass production planned in 2024.

ROHM Develops Ultra-Low-Power On-Device Learning Edge AI Chip | ROHM Semiconductor - ROHM Co., Ltd.

www.rohm.com

 

[equanimous]

Time in the market

 

[Fastback6666]

Prophesee’s Big Three: Sony, Qualcomm and Smartphones | TechSplicit

Finally, event-based image sensors leveraging neuromorphic engineering principles come to smartphones.

ojoyoshidareport.com

By teaming with Sony, the world’s largest CMOS image sensor company, and Qualcomm, which commands a 50-percent share of the mobile SoC market, Prophesee, a Paris-based startup, is finally finding a massive volume market for its unique event-based cameras in smartphones

So does this mean Brainchip will have massive volume in royalties flowing through from this? Sounds exciting....

 

[Diogenese]

I'm excited.

https://www.nviso.ai/en/extreme-edge-high-performance-computing-hpc

NVISO Neuro Models™ are purpose built for a new class of ultra-efficient AI processors designed for ultra-low deep learning on edge devices. Supporting a wide range of heterogenous computing platforms ranging from CPU, GPU, VPU, NPU, and neuromorphic computing they reduce the high barriers-to-entry into the embedded AI space through cost-effective standardized AI Apps which work optimally on edge devices for a range of common human behaviour use cases (low power, on-device, without requiring an internet connection). NVISO Neuro Models™ use low and mixed precision activations and weights data types (1 to 8-bit) combined with state-of-the-art unstructured sparsity to reduce memory bandwidth and power consumption. Using proprietary compact network architectures, they can be fully sequential suitable for ultra-low power mixed signal inference engines and fully interoperable with neuromorphic processors as well as existing digital accelerators.

Now who do we know who has a digital SNN capable of handling 1 to 8 bits ...
... in particular, 8-bit weights and activations?

 

[Fact Finder]

We all know Quantum Ventura they famously confirmed in a peer reviewed research paper that an AKIDA 1000 USB at $50.00 could match it with a Nvidia GPU at $30,000.

https://www.sciencedirect.com/science/article/pii/S1877050922017860/pdf?md5=c6abc5833a81c629954f7969fc65de04&pid=1-s2.0-S1877050922017860-main.pdf

Well this is what they are presently doing according to their website:

Our Current Federally-Funded Research Projects as the Prime Contractor:​

AI/ML/ Hyperspectral Imaging/ Neuromorphic/ Cybersecurity related topics:



DARPA: "AI Verification with provable guarantees" using advanced AI verification tools.

Partner: NC State University



Navy Air Warfare: "Certification of AI Systems - CORSI" using Advanced AI to certify AI/ML applications. (Phase 1 and Phase 2 SBIRs)

Partner: Lockheed Martin.



Missile Defense Agency: "Hypersonic Threat Detection" using bio-inspired processing, neuromorphic computing and Advanced AI. (Phase 1 STTR)

Partners: University of Florida and Lockheed Martin.



Navy Air Warfare: "Detection of UAVs and rogue drones using hyperspectral Imaging" (SBIR Phase 1).

Partners: Bodkin Imaging and Lockheed



Navy Air Warfare: "Vulnerability detection of source code using advanced AI/ML" - SBIR Phase 1



Homeland Security: "Opioid/contraband detection using hyperspectral imaging" - SBIR Phase 1



Department of Energy: "Cyber threat-detection using neuromorphic computing" - SBIR Phase 1

All sounding a bit ubiquitous if you ask me.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Diogenese]

I couldn't resist another peek at the Gen 2 Product Brief:

https://brainchip.com/wp-content/uploads/2023/03/BrainChip_second_generation_Platform_Brief.pdf

I can't wait!!!!!!!!!!!!!! ...

"Exceptional spatio-temporal capability: Patent-pending Temporal Event-based Neural Nets (TENNs) revolutionize time-series data applications"



"Efficient Vision Transformer acceleration: Vision Transformer encoder acceleration to provide radically better vision solutions"


Note that, when not operating in pure SNN mode, some processor participation is needed:
"Accelerates today’s networks: CNNs, DNNs, RNNs, Vision Transformers (ViT), and more, directly in hardware with minimal CPU intervention ...
Independent neural processor operation: Intelligent DMA minimizes or eliminates need for CPU in AI acceleration; minimizes system load"

Well I haven't looked at Renesas DRP-AI (dynamically reconfigurable processor -AI), but Akida can do it with your eyes closed.

Multi-Pass Processing Delivers Scalability, Future-Proofing:
Extremely scalable
• Runs larger networks on given set of nodes
• Reduces Silicon footprint and Power in SoC
Transparent to application developer and users
• Handled by runtime software
• Segments and processes network sequentially
Minimizes incremental latency
• Handles multiple layers concurrently
• Minimizes CPU intervention

That's pretty Tardus-like - if you need more nodes than there are on the SoC, we'll just keep using the ones we have until the jobs done. It's like when the team bus breaks down and you've only got a Mini ...


"Akida is model-, network-, and OS-agnostic"
So we can use custom models like nViso's as well as the standard models and out own in-house models.

 

[Deadpool]

We all know Quantum Ventura they famously confirmed in a peer reviewed research paper that an AKIDA 1000 USB at $50.00 could match it with a Nvidia GPU at $30,000.

https://www.sciencedirect.com/science/article/pii/S1877050922017860/pdf?md5=c6abc5833a81c629954f7969fc65de04&pid=1-s2.0-S1877050922017860-main.pdf

Well this is what they are presently doing according to their website:

Our Current Federally-Funded Research Projects as the Prime Contractor:​

AI/ML/ Hyperspectral Imaging/ Neuromorphic/ Cybersecurity related topics:



DARPA: "AI Verification with provable guarantees" using advanced AI verification tools.

Partner: NC State University



Navy Air Warfare: "Certification of AI Systems - CORSI" using Advanced AI to certify AI/ML applications. (Phase 1 and Phase 2 SBIRs)

Partner: Lockheed Martin.



Missile Defense Agency: "Hypersonic Threat Detection" using bio-inspired processing, neuromorphic computing and Advanced AI. (Phase 1 STTR)

Partners: University of Florida and Lockheed Martin.



Navy Air Warfare: "Detection of UAVs and rogue drones using hyperspectral Imaging" (SBIR Phase 1).

Partners: Bodkin Imaging and Lockheed



Navy Air Warfare: "Vulnerability detection of source code using advanced AI/ML" - SBIR Phase 1



Homeland Security: "Opioid/contraband detection using hyperspectral imaging" - SBIR Phase 1



Department of Energy: "Cyber threat-detection using neuromorphic computing" - SBIR Phase 1

All sounding a bit ubiquitous if you ask me.

My opinion only DYOR
FF

AKIDA BALLISTA

I think we're going to need a bigger boat.

 

[Diogenese]

I think we're going to need a bigger boat.

I think we're gonna need a flotilla.

 

[Diogenese]

So does this mean Brainchip will have massive volume in royalties flowing through from this? Sounds exciting....

We haven't seen any evidence that we are involved in this first round of Sony/Prophesee, ...

... but

... does anyone else make mobile phones?

 

[Fact Finder]

I couldn't resist another peek at the Gen 2 Product Brief:

https://brainchip.com/wp-content/uploads/2023/03/BrainChip_second_generation_Platform_Brief.pdf



View attachment 33016

View attachment 33018



I can't wait!!!!!!!!!!!!!! ...

"Exceptional spatio-temporal capability: Patent-pending Temporal Event-based Neural Nets (TENNs) revolutionize time-series data applications"



"Efficient Vision Transformer acceleration: Vision Transformer encoder acceleration to provide radically better vision solutions"


Note that, when not operating in pure SNN mode, some processor participation is needed:
"Accelerates today’s networks: CNNs, DNNs, RNNs, Vision Transformers (ViT), and more, directly in hardware with minimal CPU intervention ...
Independent neural processor operation: Intelligent DMA minimizes or eliminates need for CPU in AI acceleration; minimizes system load"

View attachment 33019


Well I haven't looked at Renesas DRP-AI (dynamically reconfigurable processor -AI), but Akida can do it with your eyes closed.

Multi-Pass Processing Delivers Scalability, Future-Proofing:
Extremely scalable
• Runs larger networks on given set of nodes
• Reduces Silicon footprint and Power in SoC
Transparent to application developer and users
• Handled by runtime software
• Segments and processes network sequentially
Minimizes incremental latency
• Handles multiple layers concurrently
• Minimizes CPU intervention

That's pretty Tardus-like - if you need more nodes than there are on the SoC, we'll just keep using the ones we have until the jobs done. It's like when the team bus breaks down and you've only got a Mini ...


"Akida is model-, network-, and OS-agnostic"
So we can use custom models like nViso's as well as the standard models and out own in-house models.

It reminds me of this Charlie Chaplin skit - just keep going around till you knock it out of the park:

 

[Deadpool]

I think we're gonna need a flotilla.

This just keeps getting better by the day.

 

[alwaysgreen]

I'm keeping a low profile on here at the moment but I will say one thing.

I did have a conversation with Tony Dawe in my hiatus and his direct quote to me asking about companies purchasing licenses through third parties (such as Megachips) was this:

"It seems highly unlikely that a customer of one of our licensees would keep its use of neuromorphic AI a secret for very long.

Such a company would almost certainly use the employment of Akida neuromorphic IP as part of its marketing pitch to attract customers by differentiating its product from its competitors".

So hopefully, the floodgates are released soon and we hear of a number of products utilising Akida. Also, whoever has already purchased through Megachips, please shout it from the rooftops! (edit: likely it is Mercedes as stated by DB below)

 

[DingoBorat]

Hi Tothemoon
I highly doubt that AKIDA is in these new Mercedes models as their has not been a licensed agreement with Mercedes Benz signed yet.
Wasn't it stated that the new MB system would be released in 2024 and that imo will be our big year.

"has not been a licensed agreement with Mercedes Benz signed yet"

You can't say that as a a fact TasTroy, only as an assumption, on your part.

For my part, I'd bet money, that one of the licence agreements secured by MegaChips (which we've received funds for) was with Mercedes.

 

[Foxdog]

I'm keeping a low profile on here at the moment but I will say one thing.

I did have a conversation with Tony Dawe in my hiatus and his direct quote to me asking about companies purchasing licenses through third parties (such as Megachips) was this:

"It seems highly unlikely that a customer of one of our licensees would keep its use of neuromorphic AI a secret for very long.

Such a company would almost certainly use the employment of Akida neuromorphic IP as part of its marketing pitch to attract customers by differentiating its product from its competitors".

So hopefully, the floodgates are released soon and we hear of a number of products utilising Akida.

Ok so the silence on AKIDA being utilised might not be solely because of NDA's but simply because no one has yet employed AKIDA in their products (in a commercial sense anyway). Or perhaps they are really close to letting the cat out of the bag after prolonged R&D. I wonder if our management team can shed any light on licensee numbers, progress etc. Surely they are updated through the licensees.

 

[alwaysgreen]

"has not been a licensed agreement with Mercedes Benz signed yet"

You can't say that as a a fact TasTroy, only as an assumption, on your part.

For my part, I'd bet money, that one of the licence agreements secured by MegaChips (which we've received funds for) was with Mercedes.

I would also agree with that. I believe Merc has signed through Megachips but that is purely speculation (I'd prefer to call it an educated assumption).

Ok so the silence on AKIDA being utilised might not be solely because of NDA's but simply because no one has yet employed AKIDA in their products (in a commercial sense anyway). Or perhaps they are really close to letting the cat out of the bag after prolonged R&D. I wonder if our management team can shed any light on licensee numbers, progress etc. Surely they are updated through the licensees.

It would be great if they would provide that information but I don't see it being forthcoming.

 

[TasTroy77]

"has not been a licensed agreement with Mercedes Benz signed yet"

You can't say that as a a fact TasTroy, only as an assumption, on your part.

For my part, I'd bet money, that one of the licence agreements secured by MegaChips (which we've received funds for) was with Mercedes.

Hi DingoBorat
No I can't but for Mercedes Benz to be branded on the Brainchip website one would think that they are a stand alone not through Megachips or Renesas.
Man I would love to be totally wrong believe me.

 

[DingoBorat]

Hi DingoBorat
No I can't but for Mercedes Benz to be branded on the Brainchip website one would think that they are a stand alone not through Megachips or Renesas.
Man I would love to be totally wrong believe me.

It's not about being stand alone, it is likely Mercedes needs MegaChips's help, to integrate the BrainChip IP into their A.I. solutions (remember, that they have also just helped BrainChip, with the backend design of AKIDA1500).

Remember from the Strategic Partnership agreement with MegaChips.

"By partnering with BrainChip, MegaChips is able to quickly and easily maintain its industry innovator status by supplying solutions and applications that leverage the Akida revolutionary technology in markets such as automotive, IoT, cameras, gaming and industrial robotics"

Now sure, Mercedes could licence the IP directly from us, or they could licence it through the party designing their ASICS (assuming it's MegaChips).

Same same really, but makes more sense, just to do it through MegaChips.

"Hey there MC, we've just licenced BrainChip's IP, which you already have a licence to use, can you design our chips, using said IP?"

The above doesn't sound like a likely scenario..
Remembering, that regardless, Mercedes has to pay the licencing fee, for our IP and any other IP, being used.

We know Mercedes is not shy about mentioning us, so it's not for reason of keeping the association under wraps, as it "may" be with other companies.


@Foxdog's and @alwaysgreen's wish for a disclosure of the number of licence agreements, through MegaChips (or Renesas, but they only licenced 2 nodes) is a Great idea and someone should ask it at the AGM, or even through Company channels.

I don't see any reason at all, why such a figure, couldn't be revealed?..

 

[M_C]

Possibility here imo

"The eight-core Snapdragon 8 Gen 3 will supposedly use derivatives of new, unannounced Arm core designs and, more notably, shift the configuration of the performance clusters"

New Snapdragon 8 Gen 3 intelligence suggests more powerful cores are coming

We're on the road to Qualcomm's Snapdragon 8 Gen 3 , but we've hit our first 'up' and 'down' along the way

www.androidpolice.com

 

[jtardif999]

I am thrilled to share my latest creation with you! I've developed an AI-powered Q&A application that can be used on any website, and I've chosen the Brainchip Inc website for the purposes of this demo.

You can access it through this link: https://tw.zeebot.org/web-crawler/brainchip/.

I know the demo might be a little slow, but please bear with me! This is just a temporary issue and is a result of the early demo/prototype nature of the app. I wanted to share it with you as soon as possible because I'm so excited about what it can do.

As early testers, your feedback is incredibly valuable to me. Please don't hesitate to share your screenshots, comments, and suggestions with me. It will help me make the app better and ensure that it meets your needs as users of my Q&A type application.

My app crawls any website in full to provide you with accurate and up-to-date information using advanced language AI.

There's no setup required for any company that would like to implement it on their website, or as a live chat, or email responder, or anything else you can imagine! It's like having an extremely knowledge personal assistant, sales, or customer care agent who never sleeps.

It’s only limited by what has been published on the website.

So what are you waiting for? Give my app a try and join me in the excitement of the future!

Best regards,

zeeb0t

Wow pretty good first attempt at a mini ChatGPT with a domain of just this website or is the Brainchipinc website? Anyway, it stumbled on my first question - I asked it ‘What is JAST?’ and it said (paraphrasing) that this website contains no information about JAST? A little disappointing but that’s why you want the feedback I’m guessing.