BRN Discussion Ongoing

[Fact Finder]

There are five other issues to consider when looking at the question of SynSense as a legitimate competitor to Brainchip:

1. SynSense has deep and pervasive links to China - https://www.northdata.com/?id=6093567565 ;

2. Prophesee is working in the NASA/DARPA space;

3. SynSense is a venture backed private company unconstrained by the likes of the ASX & SEC as to what it can announce and the claims it makes in such announcements;

4. BMW has strong and historical links with NASA;

5. BRAINCHIP is a qualified customer/ research partner of both NASA and DARPA.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Fullmoonfever]

A couple things found that were interesting but not necessarily that revealing but thought post anyway.

Though a little dated, Mar this year (only just found it), would have been nice to see the content & listen to Sean on this Management call.

Couldn't see any releases on the BRN website or any Ann's re the presso or call but maybe missed them or has been incorporated in bits of other pressos?

Curious as to some of the answers, expanded responses to the questions listed as saw some TSE posts around data centers & SiFive etc.

Also, the line about the call being just before NVDA's GTC conference was intriguing but could be nothing as well.

About - Wedbush Securities

www.wedbush.com

Also found the Tech Insights team are currently working on "BrainChip Akida AI Processor Digital Floorplan Analysis".

Probs one for our more Tech savvy holders to read through if becomes public.

Is in creation at the mo and they have previously published a couple of older reports.

Analysis search | TechInsights

Analysis search

www.techinsights.com

Snips:

Bit further on data centers, have seen that IBM have just released their LinuxOne Emperor 4 which is running the Telum chip with an AI accelerator?

Reading some specs, no mention of SNN but RNN / CNN.

IBM LinuxONE 4

IBM LinuxONE 4 is the latest iteration of IBM LinuxONE mainframes with on-chip AI inferencing and industry-first quantum-safe technologies.

www.ibm.com

On the 4C thoughts, it should be quite telling for mine as although it has been mentioned about things that have happened in the prevailing 9 months the half yearly figures said MegaChips was primary revenue obviously however, a chunk would be one off licencing fees so won't be recurring in these upcoming figures (unless new licences issued) so will be good to see how much more traction has been gained removing the one offs.

 

[TECH]

The SynSense/BMW arrangement is exploratory, ie, BMW are testing SynSense technology.

"This exploration with BMW in the area of neuromorphic technology will focus on SynSense’s dynamic visual intelligence SoC––Speck, which combines in one single chip SynSense’s low-power SNN vision processor with an event-based sensor. It performs both neuromorphic sensing and computing , and utilizes full asynchronous circuit design. It can be used to capture real-time visual information, recognize and detect objects, and perform other vision-based detection and interaction functions."

So that leaves the door open for better solutions.

Having Swiss origins, SynSense started out in 2017 with analog SNNs and were still going down that path in 2021:

CN113255905A Signal processing method for neurons in spiking neural network and network training method

View attachment 17907



The invention discloses a signal processing method for neurons in a spiking neural network and a network training method. Different from a single-pulse mechanism which is commonly used at present, the multi-pulse mechanism is designed. The neuron signal processing method comprises the following steps: a receiving step: at least one neuron receives at least one path of input pulse sequence; an accumulation step: performing weighted summation based on the at least one path of input pulse sequence to obtain a membrane voltage; and an activation step: when the membrane voltage exceeds a threshold value, determining the amplitude of the pulse excited by the neuron based on the ratio of the membrane voltage to the threshold value. In order to solve the problems of time consumption and low efficiency of a training algorithm caused by an increasing configuration parameter scale, efficient training of the spiking neural network is realized through technical means such as a multi-pulse mechanism, a periodic exponential function proxy gradient and addition of a neuron active inhibition degree as loss, and low power consumption of simulated hardware can still be maintained; and the technical effects of improving the precision and the convergence speed and the like are also brought.

I don't know if SynSense have solutions for the manufacturing variability which plagues analog SNNs.

Hi Dio,

That's a very well rounded, nicely balanced post.... at this stage of proceedings SynSense would need to have a digital SNN design out, if not I can recommend one to BMW.... but I'd strongly suggest that they are already well aware of such a company.

AKD 2.0 really holds the key to displaying the difference between us and the mob again, especially in the Automotive Industry.

What makes us so special? now that's where our key executives have to really zero in (as they are doing) and sell the differences that set us apart from the mob, we all believe we have the best technology, the 3/5-year technological lead, BUT a potential new customer needs to fully understand WHY, before any real engagement gets off the ground.

If anyone can close out a deal, it will be Chris Stevens, ably backed by Ken and the team.

I'm quietly impressed with how Brainchip's price is holding up, it's great to see, maybe a little bit of positioning in anticipation of the 4C results.

Enjoy your daylight savings east coasters.

Tech.

 

[Dhm]

Some quick quotes and a note or two from the interview:

"I'm even more convinced today. The tech is real, it's transformational - nothing even comes close."

"The technology is extremely flexible"

"We have the best-kept secret in the world."

"Tech companies and investors are beating a path to our door right now." - Tony

When asked about when we will know about products containing Akida in the market Sean responded that he does not foresee a time and most
customers are pretty serious about keeping it confidential. He then added: "of course, I know (trying to stop himself from smiling), but most people could not." He later mentioned that the best way to get an idea of this will be to keep an eye on financials.

He spoke about how it is not a short sales cycle but how customers are carefully selected with the criteria of being those who will get products to market. He spoke about the delays in chip manufacturing which are twice what they were and how companies have typical design cycles for products.
He also mentioned that many customers want to have their own IP (though incorporating BRN's IP into something larger - that is how I understood it).
These concepts reinforce for me why patience is required.

I asked about what % of EAP customers continue to engage. Sean responded "virtually all of them".

Overall, nothing really new but I thought it was a fantastic interview. I was very impressed with Sean and am delighted to have him at the helm.
I'm sure others who listened will share more, but above is what stood out to me.
Back to my charts.
GLTA

Edit: He also mentioned they are in discussions with a major communications company, a use case he did not expect, but that companies who want to have a foot in the future of AI are engaging.

I'm resurrecting a post but more importantly a brilliant podcast by Sean with Strawman in late March 2022. It is important because it reminds us of what a great company we are backing.



at 29:20 potential (for Brainchip) of revenue in the billions of (US) dollars.
30:40 companies are beating a path to our door
34:10 we have the best kept secret in the world

As @Fact Finder has stated a number of times in the past, because we are an ASX company we have a stringent set of obligations that basically state that any claim an office bearer makes has to be realistic and achievable. So what Sean and Tony state are realistic assumptions and would be jumped on by the ASX if otherwise.

 

[Diogenese]

Hi Dio,

That's a very well rounded, nicely balanced post.... at this stage of proceedings SynSense would need to have a digital SNN design out, if not I can recommend one to BMW.... but I'd strongly suggest that they are already well aware of such a company.

AKD 2.0 really holds the key to displaying the difference between us and the mob again, especially in the Automotive Industry.

What makes us so special? now that's where our key executives have to really zero in (as they are doing) and sell the differences that set us apart from the mob, we all believe we have the best technology, the 3/5-year technological lead, BUT a potential new customer needs to fully understand WHY, before any real engagement gets off the ground.

If anymore can close out a deal, it will be Chris Stevens, ably backed by Ken and the team.

I'm quietly impressed with how Brainchip's price is holding up, it's great to see, maybe a little bit of positioning in anticipation of the 4C results.

Enjoy your daylight savings east coasters.

Tech.

A substantial influence on the BMW/SynSense deal seems to have been the ability to combine an event sensor (Prophesee) with a SNN (SynSense) on a single chip.

Prophesee and SynSense have agreed to do co-promotion/marketing, so does SynSense have a conflict of interest?

I wouldn't think that Prophesee would have signed up to an exclusivity clause, so my guess is that they are free to promote Akida in conjunction with their event camera.

Interesting that the SynSense/BMW cooperation was not kept under NDA at the start of the cooperation?

 

[Fact Finder]

A substantial influence on the BMW/SynSense deal seems to have been the ability to combine an event sensor (Prophesee) with a SNN (SynSense) on a single chip.

Prophesee and SynSense have agreed to do co-promotion/marketing, so does SynSense have a conflict of interest?

I wouldn't think that Prophesee would have signed up to an exclusivity clause, so my guess is that they are free to promote Akida in conjunction with their event camera.

Interesting that the SynSense/BMW cooperation was not kept under NDA at the start of the cooperation?

Hi @Diogenese
Your post reminded me of the fact that under the SynSense Prophesee announcement while they agreed as you say to co promotion and marketing it was SynSense which carried the burden of making their SNN work on chip with Prophesee’s event based sensor.

An OK if you really want to have a go then if you make it work we are happy to share any profits and pay for any marketing from Prophesee’s point of view. No risk and no responsibility.

Whereas the Brainchip Prophesee announcement speaks of a true partnership with Prophesee carrying the product solution ball to its customers.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Deleted member 118]

Anyone care to do a little bit more research on this company

Please verify you are human.

www.einnews.com

Today, Neurobotx announced a paid partnership with leading aerospace company Urban Aeronautics, through their Neurobotx EVTOL Partnership Program.

Not only will their (...) demo showcase the capabilities of the CityHawk (...), all the data generated through Metapilot will be used for future training in our AI simulator.— Haran Ben-Eliahou, VP of Business Development at Urban AeronauticsNEW YORK, USA, October 2, 2022 /EINPresswire.com/ -- Today, Neurobotx announced a paid partnership with leading aerospace company Urban Aeronautics, through their Neurobotx EVTOL Partnership Program.

Neurobotx is a Boeing-backed company led by neuroscientist Dr. Diana Deca, aimed at building the cognitive metaverse for airtaxi and EVTOL companies. The company collaborates with different aviation companies to combine state-of-the-art simulation, haptics, digital marketing, PR, business development, synthetic data and deep learning to help expedite and optimize go-to-market and investment opportunities for their EVTOL clients. Through this partnership, neurobotx and Urban Aeronautics are collaborating to visualize and simulate CityHawk for new and existing use cases, industry stakeholders and VC partners.

Urban Aeronautics’ highly advanced CityHawk eVTOL is the first hydrogen-powered aircraft vehicle to be both wingless and silent. Its compact exterior and proprietary fully-enclosed Fancraft™ rotor system gives it a minimal footprint and near-silent presence, making it an aircraft that fits the city as it currently exists. Through their partnership with Neurobotx, a virtual CityHawk model will become accessible to millions of gamers, aviation enthusiasts, investors, clients, and regulators within their Metapilot platform, allowing gamers and/or future pilots the ability to fly the CityHawk eVTOL through a simulation around a city. This brings the CityHawk’s future of urban flight one step closer to reality.

“We are excited to have the Neurobotx team behind us as they provide a full stack of services that gives Urban Aeronautics the opportunity to bring the CityHawk eVTOL to millions of users and generate valuable IP data,” says Haran Ben-Eliahou, VP of Business Development at Urban Aeronautics. “Not only will their customized demo showcase the capabilities of the CityHawk as a flying ambulance and air taxi, all the data generated through Metapilot will be used for future training in our AI simulator.”

"Dr. Deca adds on the neurobotx - Urban Aeronautics commercial partnership: "City Hawk is one of the latest EVTOL models we've added to our Metapilot platform, and we are very selective with the models we bring into our partnership program. Metapilot is designed to become a fully immersive XR experience, with a neuromorphic backend, based on my own research looking at the brain’s response to VR. By being the first silent and wingless airtaxi, CityHawk now has an extraordinary advantage, by also having their in house AI simulator, and now a large dataset of training data along with a major partner for business development and regulatory approval in neurobotx. Because we like the model and the commitment to AI innovation, we will leverage our network among our partners such as Boeing and NASA - National Aeronautics and Space Administration to speed up the mass adoption and scaling of Urban Aeronautics. "
For more potential partnerships, collaborations and discussions you may reach out to

 

[Fullmoonfever]

A substantial influence on the BMW/SynSense deal seems to have been the ability to combine an event sensor (Prophesee) with a SNN (SynSense) on a single chip.

Prophesee and SynSense have agreed to do co-promotion/marketing, so does SynSense have a conflict of interest?

I wouldn't think that Prophesee would have signed up to an exclusivity clause, so my guess is that they are free to promote Akida in conjunction with their event camera.

Interesting that the SynSense/BMW cooperation was not kept under NDA at the start of the cooperation?

Speaking of Synsense.

Have you ever looked into Xylo / Rockpool AI?

Recent paper re Xylo and when I went to Rockpool...section further down re Synsense.

Sub-mW Neuromorphic SNN audio processing applications with Rockpool and Xylo

08/27/22 - Spiking Neural Networks (SNNs) provide an efficient computational mechanism for temporal signal processing, especially when couple...

deepai.org

Sub-mW Neuromorphic SNN audio processing applications with Rockpool and Xylo​

08/27/2022
∙
by Hannah Bos, et al.
∙

6
∙

share
Spiking Neural Networks (SNNs) provide an efficient computational mechanism for temporal signal processing, especially when coupled with low-power SNN inference ASICs. SNNs have been historically difficult to configure, lacking a general method for finding solutions for arbitrary tasks. In recent years, gradient-descent optimization methods have been applied to SNNs with increasing ease. SNNs and SNN inference processors therefore offer a good platform for commercial low-power signal processing in energy constrained environments without cloud dependencies. However, to date these methods have not been accessible to ML engineers in industry, requiring graduate-level training to successfully configure a single SNN application. Here we demonstrate a convenient high-level pipeline to design, train and deploy arbitrary temporal signal processing applications to sub-mW SNN inference hardware. We apply a new straightforward SNN architecture designed for temporal signal processing, using a pyramid of synaptic time constants to extract signal features at a range of temporal scales. We demonstrate this architecture on an ambient audio classification task, deployed to the Xylo SNN inference processor in streaming mode. Our application achieves high accuracy (98 low power (<100μW inference power). Our approach makes training and deploying SNN applications available to ML engineers with general NN backgrounds, without requiring specific prior experience with spiking NNs. We intend for our approach to make Neuromorphic hardware and SNNs an attractive choice for commercial low-power and edge signal processing applications.

Welcome to Rockpool — Rockpool 2.8 documentation

rockpool.ai

Rockpool is a Python package for working with dynamical neural network architectures, particularly for designing event-driven networks for Neuromorphic computing hardware. Rockpool provides a convenient interface for designing, training and evaluating recurrent networks, which can operate both with continuous-time dynamics and event-driven dynamics.

Rockpool is an open-source project managed by SynSense.

 

[Fullmoonfever]

Just saw that Samsung had a patent filed back in 2020 granted this year in June.

Obviously have to cite others and thankfully PVDM patents also included.

Be nice if had knocked on BRN door at some point seeing Samsung obviously want SNN....we could save them a lot of time, $ and effort in getting to mkt

US Patent for Neuromorphic device and neuromorphic system including the same Patent (Patent # 11,362,868 issued June 14, 2022) - Justia Patents Search

A neuromorphic device includes a neuron block, a spike transmission circuit and a spike reception circuit. The neuron block includes a plurality of neurons connected by a plurality of synapses to perform generation and operation of spikes. The spike transmission circuit generates a non-binary...

patents.justia.com

Neuromorphic device and neuromorphic system including the same​

Jun 25, 2020 - Samsung Electronics
A neuromorphic device includes a neuron block, a spike transmission circuit and a spike reception circuit. The neuron block includes a plurality of neurons connected by a plurality of synapses to perform generation and operation of spikes. The spike transmission circuit generates a non-binary transmission signal based on a plurality of transmission spike signals output from the neuron block and transmits the non-binary transmission signal to a transfer channel, where the non-binary transmission signal includes information on transmission spikes included in the plurality of transmission spike signals. The spike reception circuit receives a non-binary reception signal from the transfer channel and generates a plurality of reception spike signals including reception spikes based on the non-binary reception signal to provide the plurality of reception spike signals to the neuron block, where the non-binary reception signal includes information on the reception spikes.

Patent History
Patent number: 11362868
Type: Grant
Filed: Jun 25, 2020
Date of Patent: Jun 14, 2022
Patent Publication Number: 20210160109
Assignee: Samsung Electronics Co., Ltd. (Gyeonggi-do)
Inventors: Changkyu Seol (Osan-si), Hongrak Son (Anyang-si), Geunyeong Yu (Seongnam-si), Pilsang Yoon (Hwaseong-si), Jaehun Jang (Hwaseong-si)
Primary Examiner: David B Lugo
Application Number: 16/911,801
Classifications
Current U.S. Class: Forward Error Correction By Tree Code (e.g., Convolutional) (714/786)
International Classification: H04L 27/04 (20060101); H04L 1/00 (20060101); H04L 27/06 (20060101); G06N 3/063 (20060101); G06N 3/04 (20060101);

10157629

December 18, 2018

van der Made et al.

20180225562

August 9, 2018

van der Made

20200143229

May 7, 2020

Van Der Made

 

[Foxdog]

I'm really very sorry, but I still haven't gotten around to reading everything here, so once again I don't know if you guys have already discussed the new Mercedes EQE SUV, which will be presented on October 16th

Do you think that our chip is integrated here? Does anyone know more about this?

View attachment 17865

It’s nearly time! The digital world premiere of our EQE SUV will be streamed here on LinkedIn (Mercedes-Benz AG) on October 16, and we’re very excited about it.

The new EQE SUV is the multi-purpose variant of the EQE business saloon and what we call a true EV all-rounder. It’s spacious, highly versatile and equipped with all sorts of innovative features that make life more comfortable and convenient.

In fact, you could call it a double premiere, as we will also be unveiling the performance-oriented AMG version. It has its own unmistakable AMG flavour – powerful and dynamic with a distinctive exterior and interior design.

But I’m not going to say anymore right now. For the full picture, tune in here https://lnkd.in/etvFvSjZ at 8.00 pm CEST on October 16 for the live event.

OMG noooooo, I just realised we've backed the wrong company - they can't even put the steering wheel on the proper side

 

[Fact Finder]

OMG noooooo, I just realised we've backed the wrong company - they can't even put the steering wheel on the proper side

Relax. That steering wheel is for nervous passengers and driving instructors. The driver does not have a steering wheel in Mercedes Level 5 autonomy.

 

[Foxdog]

Relax. That steering wheel is for nervous passengers and driving instructors. The driver does not have a steering wheel in Mercedes Level 5 autonomy.

 

[Foxdog]

Oh, of course, how foolish of me

 

[JK200SX]

FWIW, this article came out today. It focuses on Intel/Loihi, and it mentions Brainchip toward the end:

"Lava is an open source framework with permissive licenses, so the expectation is that other neuromorphic chip makers—which include IBM, Qualcomm, and BrainChip—will port Lava to their own frameworks. It’s not proprietary, although Intel is the biggest contributor to it, Davies said."

Anyone care to comment?

Software, not hardware, will drive quantum and neuromorphic computing

aumag.net

 

[Fact Finder]

FWIW, this article came out today. It focuses on Intel/Loihi, and it mentions Brainchip toward the end:

"Lava is an open source framework with permissive licenses, so the expectation is that other neuromorphic chip makers—which include IBM, Qualcomm, and BrainChip—will port Lava to their own frameworks. It’s not proprietary, although Intel is the biggest contributor to it, Davies said."

Anyone care to comment?

Software, not hardware, will drive quantum and neuromorphic computing

aumag.net

The absence of quotation marks is a little troubling so if these are Mike Davies exact words to have the head of neuromorphic computing at Intel line up Brainchip with IBM and Qualcomm concludes the Chapter in the Brainchip book titled “Has Brainchip Achieved Market Traction.”

Now coming back to Lava I raised a question with Brainchip when it was first announced as to whether it was compatible with AKIDA technology and while it would be it was not thought to advance Brainchip as Brainchip had been and is continuing to pump out all necessary or expected software at a more than sufficient rate to meet market demand.

I would personally see Lava after this discussion as an attempt by Intel to remain neuromorphically
relevant.

My opinion only DYOR
FF


AKIDA BALLISTA

 

[jk6199]

Just saw the Frankfurt price down 7.49%.

Then I saw the volume, 10 shares!

Someone’s rolling the dice

 

[cassip]

It is no usual trading day today. It is a public holiday (German Unity Day).

 

[Diogenese]

FWIW, this article came out today. It focuses on Intel/Loihi, and it mentions Brainchip toward the end:

"Lava is an open source framework with permissive licenses, so the expectation is that other neuromorphic chip makers—which include IBM, Qualcomm, and BrainChip—will port Lava to their own frameworks. It’s not proprietary, although Intel is the biggest contributor to it, Davies said."

Anyone care to comment?

Software, not hardware, will drive quantum and neuromorphic computing

aumag.net

The absence of quotation marks is a little troubling so if these are Mike Davies exact words to have the head of neuromorphic computing at Intel line up Brainchip with IBM and Qualcomm concludes the Chapter in the Brainchip book titled “Has Brainchip Achieved Market Traction.”

Now coming back to Lava I raised a question with Brainchip when it was first announced as to whether it was compatible with AKIDA technology and while it would be it was not thought to advance Brainchip as Brainchip had been and is continuing to pump out all necessary or expected software at a more than sufficient rate to meet market demand.

I would personally see Lava after this discussion as an attempt by Intel to remain neuromorphically
relevant.

My opinion only DYOR
FF


AKIDA BALLISTA

I'm a little bemused.

The only software that is involved in Akida relates to the configuration of the NN layout (asigning nodes to layers) and in the NN models (image libraries with which Akida compares the incoming images). The ARM Cortex microprocessor performs these preliminary operations.

The functioning of Akida does not require any software program. Akida operates automatically and asynchronously to compare the input images with the model library images. The input image is automatically fed through the nodes of the layers and compared with the stored model library images to establish the most likely match (eg, Rob Telson's recent Corona v Corona Lite test where a couple of training shots were required to establish the difference).

As @Fact Finder pointed out, BrainChip has teams of programmers developing the optimal configurations and model libraries, and optimizing CNN2SNN, and who knows what else LSTM, cortical, ...?

So Lava's Magma enables simulation of neuromorphic hardware on CPU/GPU. Brainchip has MetaTF to do this. Brainchip also has CNN2SNN conversion to handle the legacy CNN applications, so current CNN programmers can readily adapt their existing applications to run on Akida.

New SNN applications for Akida can be developed in a programming language (eg, TensorFlow, Python) with which current programmers are familiar, a feature which Intel now seeks to emulate.

Lava supports software tools for off-line training using backpropagation. Akida does on-chip training in hardware.

https://www.intel.com/content/www/u...rphic-computing-loihi-2-technology-brief.html

.

 

[TasTroy77]

I'm really very sorry, but I still haven't gotten around to reading everything here, so once again I don't know if you guys have already discussed the new Mercedes EQE SUV, which will be presented on October 16th

Do you think that our chip is integrated here? Does anyone know more about this?

View attachment 17865

It’s nearly time! The digital world premiere of our EQE SUV will be streamed here on LinkedIn (Mercedes-Benz AG) on October 16, and we’re very excited about it.

The new EQE SUV is the multi-purpose variant of the EQE business saloon and what we call a true EV all-rounder. It’s spacious, highly versatile and equipped with all sorts of innovative features that make life more comfortable and convenient.

In fact, you could call it a double premiere, as we will also be unveiling the performance-oriented AMG version. It has its own unmistakable AMG flavour – powerful and dynamic with a distinctive exterior and interior design.

But I’m not going to say anymore right now. For the full picture, tune in here https://lnkd.in/etvFvSjZ at 8.00 pm CEST on October 16 for the live event.

I would be wonderful but I highly doubt that AKIDA IP has been integrated so fast 8 months after the EQxx was released as a concept car. Late 2023 - 24 would be a more likely scenario in my opinion.

 

[Diogenese]

I would be wonderful but I highly doubt that AKIDA IP has been integrated so fast 8 months after the EQxx was released as a concept car. Late 2023 - 24 would be a more likely scenario in my opinion.

Hi TT,

The issue with your hypothesis is that the wide screen was also premiered in the EQXX.

Would it be more difficult to include the brains with the display?

@Sirod69 My half-full glass view would plump for the proper electronics behind the screen.