BRN Discussion Ongoing

[VictorG]

Renesas & Socionext have both been partnered separately with Hailo for years yet both keep on keeping on with Brainchip and AKIDA.

Hailo is an accelerator and uses more power than AKIDA and the last time I looked was 2021 and it was running at multiple watts. In 2020 Socionext, Foxconn and Hailo had a product out cannot remember now exactly what it was but not in any way competing with AKIDA 1000 and now with AKIDA 2000 and it’s performance levels well it’s a look out Hailo AKIDA is coming through.

My opinion only DYOR
FF


AKIDA BALLISTA

You are right again FF,
I note the Hailo-15 max FPS is 700 where as Akida smashes through 1000 FPS.

 

[HopalongPetrovski]

"E,S and P versions"
Extrasensory perception: perception or communication outside of normal sensory capability, as in telepathy and clairvoyance
Norse clairvoyant shapeshifter goddess

One IP Platform, Multiple Configurable Products​

The 2nd generation IP platform will support a very wide range of market verticals and will be delivered in three classes of product.​

Akida-E: Extremely energy-efficient, for always on operation very close to, or at sensors.
Akida-S: Integration into MCU or other general purpose platforms that are used in broad variety of sensor-related applications
Akida-P: Mid-range to higher end configurations with optional vision transformers for ground breaking and yet efficient performance.

 

[Fact Finder]

You are right again FF,
I note the Hailo-15 max FPS is 700 where as Akida smashes through 1000 FPS.

As usual Hailo also does not provide actual power consumption numbers.

When I first dug into Hailo could find nothing so asked Peter van der Made and he just dismissed them saying something about how much power they used compared to AKIDA. I was not completely satisfied and kept digging and on a Hailo customers site on their specifications page they had the Hailo spec doc and it’s a while back but what I think I remember was 3 to 7 watts for Hailo 8.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[robsmark]

I think this proves that without deals being signed with guarantees of near-term revenue, or proven revenue from the existing two contracts, we aren’t going anywhere.

Shorts will continue to devalue this company until the company provides the market with further proof of adoption.

Non-material announcements don’t get much better than yesterdays, but unfortunately it didn’t generate the financial traction required to reverse the trend.

It is crucial that some of these EAP customers show some commitment to Brainchip by signing a deal. I see no reason why this shouldn’t be the case either given how the company has already announced that changes make to this latest revision were directly requested by the EAP customers.

More time is required I guess.

And there we have it - back to pre announcement SP.

How the market doesn’t think this is worth money is beyond the joke. WBT now worth $1.4b and we’re sitting at $900m.

 

[Rise from the ashes]

As usual Hailo also does not provide actual power consumption numbers.

When I first dug into Hailo could find nothing so asked Peter van der Made and he just dismissed them saying something about how much power they used compared to AKIDA. I was not completely satisfied and kept digging and on a Hailo customers site on their specifications page they had the Hailo spec doc and it’s a while back but what I think I remember was 3 to 7 watts for Hailo 8.

My opinion only DYOR
FF

AKIDA BALLISTA

Is that the correct data?

 

[Diogenese]

Sure shooting' t'was.
Good with a gun and good with a rope.
And that white horse.....Dang!

Are you impressed with the capability of Akida 2 Diogenese?
Surprised? Delighted?
As our resident Tech expert would love to have your considered opinion, if you're willing?
Also your views on what would be the killer applications for the E,S, and P versions?
And anything else you'd like to volunteer, please.

Hi Hoppy,

I'm as much in the dark as everyone else.

One application for E (up to 4 nodes) would be on-sensor SoCs, but I would think that, for Prophesee, we'd need at least the S (up to 8 nodes) or possibly the P (up to 128 nodes), although the full P would be for some really heavy lifting, possibly running several NN model libraries in parallel.

The SiFive thing is, in my opinion, very significant as we now have direct compatibility with their 8-bit X280 Intelligence processor, which, as Fmf has intimated, may be moon-bound, as well as saving Intel's bacon.

SiFive have said Akida 2E is compatible with their Efficiency MCU and P & S are compatible with X280 Intelligence so clearly they have had hands-on experience with the Akida Gen 2 simulation software (or possibly a deep cover FPGA?).

Semantic segmentation is moving well beyond wake-word recognition into the realm of speech recognition/NLP.

Similarly we are looking at object tracking in silicon, rather than relaying that task to CPU software. So tumbleweeds are in our sights, as are Colt 38 bullets.

Semantic segmentation is moving well beyond wake-word recognition into the realm of speech recognition/NLP.

We are importing into silicon some further time-and power consuming tasks beyond classification/inference/ML which were previously performed by the CPU.

@Diogenese

Thoughts on this latest NASA SBIR paper and the relevant to BRN section highlighted towards end of post.

In your thinking, does our recent discussions around 22nm FDSOI, the X280 NASA use and recent Akida platform upgrade have any applicability to the NASA comments re critical gaps possibly being filled?

TIA

Couldn't attach full paper as extension not compatible and just on moby at the mo.

National Aeronautics and Space Administration
Small Business
Innovation Research (SBIR)
Phase I
Fiscal Year 2023 Solicitation



Scope Title: Neuromorphic Software for Cognition and Learning for Space Missions

Scope Description:
This scope seeks integrated neuromorphic software systems that together achieve a space mission capability. Such capabilities include but are not limited to:
• Cognitive communications for constellations of spacecraft.
• Spacecraft health and maintenance from anomaly detection through diagnosis; prognosis; and fault detection, isolation, and recovery (FDIR).
• Visual odometry, path planning, and navigation for autonomous rovers.
• Science data processing from sensor denoising, through sensor fusion and super resolution, and finally output the generation of science information products such as planetary digital elevation maps.
In this scope, it is expected that a provider will pipeline together a number of neural nets from different sources to achieve a space capability. The first challenge is to achieve the pipelining in a manner that achieves high overall throughput and is energy efficient. The second challenge is to put together a demonstration breadboard integrated hardware/software system that achieves the throughput incorporating neuromorphic or neural net accelerators perhaps in combination with conventional processors such as CPUs, GPUs, and FPGAs. Systems on a chip (SOC), could be another demonstration hardware platform. In either case, the neural cores should do the heavy computational lifting, and the CPUs, GPUs, and FPGAs should play a supportive role. The total power requirements shall be commensurate with the space domain, for example, 10 W maximum for systems expected to operate on CubeSats 24/7 and even less wattage for lunar systems that need to operate on battery power over the 2-week-long lunar night.

The third optional challenge is to evolve the neural net individual applications and pipeline through adaptive learning over the course of a simulated mission.
Radiation tolerance and space environment robustness are not addressed directly through this scope. Rather, a provider is expected to use terrestrial grade processors and only after Phase II target radiation tolerant neuromorphic processors potentially developed under Scopes 1 or 2 or from another source. The goal is to achieve space mission capabilities that require system integration of individual neural nets together with minimal overhead conventional software. The continuous mission-long learning complements the capability of Earth operations to adapt software over the course of a mission.

As background, development of individual neural net software is now state of the practice, and a large number of neural net applications can be downloaded in standard formats such as pseudo-assembly level or programming languages such as TensorflowTM (Google Inc), PyTorchTM (Linux Foundation), NengoTM (Applied Brain Research), LavaTM (Intel Cooporation), and others. Published neural nets for aerospace applications can be found, ranging from telescope fine-pointing control to adaptive flight control to medical support for astronaut health. In addition, there are many published neural nets for analogous terrestrial capabilities, such as autonomous driving. Transfer learning and other state-of-practice techniques enable adaptation of neural nets from terrestrial domains, such as image-processing for the image net challenge, to space domains such as Mars terrain classification for predicting rover traction.

Expected TRL or TRL Range at completion of the Project: 2 to 4

Primary Technology Taxonomy:
• Level 1: TX 10 Autonomous Systems
• Level 2: TX 10.2 Reasoning and Acting
Desired Deliverables of Phase I and Phase II:
• Analysis
• Prototype
• Hardware
• Software

Desired Deliverables Description:
The deliverables for Phase I should include at minimum the concept definition of a space capability that could be achieved through a dataflow pipeline/graph of neural nets and identification of at least a portion of the pipeline that can be achieved with existing neural nets that are either already suited for the space domain or provide an analogous capability from an Earth application. The pipeline should at a minimum be mocked up and characterized by parameterized throughput requirements for the individual neural nets, a description of the dataflow and control flow integration of the system of neural nets, and an assignment and mapping from the individual software components to the hardware elements, and an energy/power/throughput estimate for the entire pipeline. Enhanced deliverables for Phase I would include a partial demonstration of the pipeline on some terrestrial hardware platform. A report that illustrates a conceptual pipeline of neural nets for autonomous rovers can be found in the reference authored by Eric Barszcz.

The deliverables for Phase II should include at minimum a demonstration hardware system, using terrestrial grade processors and sensors, that performs a significant portion of the overall pipeline needed for the chosen space capability, together with filling in at least some of the neural net applications that needed to be customized, adapted, or developed from scratch. It is expected that the hardware system would include one or more terrestrial grade neuromorphic processors that do the primary processing, with support from CPUs, GPUs, and FPGAs. An alternative would be an SOC that incorporates a substantial number of neural cores. The demonstration shall include empirical measurement and validation of throughput and power. Enhanced deliverables for Phase II would be a simulation of continuous in situ mission-long adaptation and learning that exhibits significant evolution.

State of the Art and Critical Gaps:
Neuromorphic and deep neural net software for point applications has become widespread and is state of the art. Integrated solutions that achieve space-relevant mission capabilities with high throughput and energy efficiency is a critical gap. For example, terrestrial neuromorphic processors such as Intel Corporation’s LoihiTM, Brainchip's AkidaTM, and Google Inc's Tensor Processing Unit (TPUTM) require full host processors for integration for their software development kit (SDK) that are power hungry or limit throughput. This by itself is inhibiting the use of neuromorphic processors for low SWaP space missions. The system integration principles for integrated combinations of neuromorphic software is a critical gap that requires R&D, as well as the efficient mapping of integrated software to integrated avionics hardware. Challenges include translating the throughput and energy efficiency of neuromorphic processors from the component level to the system level, which means minimizing the utilization and processing done by supportive CPUs and GPUs.

Relevance / Science Traceability:
• 03-09a (Autonomous self-sensing)
• 04-15 (Collision avoidance maneuver design)
• 04-16 (Consolidated advanced sensors for relative navigation and autonomous robotics)
• 04-23 (Robotic actuators, sensors, and interfaces)
• 04-77 (Low SWaP, “End of arm” proximity range sensors)
• 04-89 (Autonomous Rover GNC for mating)
• 10-04 (Integrated system fault/anomaly detection, diagnosis, prognostics)
• 10-05 (On-Board "thinking" autonomy)
• 10-06 (Creation, scheduling and execution of activities by autonomous systems)
• 10-16 (Fail operational robotic manipulation)

 

[Rise from the ashes]

Pdf for a captain cook for those interested.

 

[Rise from the ashes]

And there we have it - back to pre announcement SP.

How the market doesn’t think this is worth money is beyond the joke. WBT now worth $1.4b and we’re sitting at $900m.

The market is a bit dense mate otherwise we would not have been able to buy sub 5,10,, 20,30,40. Cents. It's a bit slow on realizing what we have.
Oh yeah and $1.12

 

[Rise from the ashes]

Just for interest

 

[Bombersfan]

And there we have it - back to pre announcement SP.

How the market doesn’t think this is worth money is beyond the joke. WBT now worth $1.4b and we’re sitting at $900m.

And there it is, our daily reference to a completely irrelevant WBT sp.

 

[Rise from the ashes]

 

[Proga]

And there we have it - back to pre announcement SP.

How the market doesn’t think this is worth money is beyond the joke. WBT now worth $1.4b and we’re sitting at $900m.

I can't see the SP going anywhere until an application using Akida IP hits the market at commercial scale ie millions in continual royalties. The market has decided we won't believe until we see it.

 

[Fact Finder]

View attachment 31690

Is that the correct data?

Hi @Rise from the ashes

Could well be but my memory had power just listed as 3 to 7 watts in a box with columns covering various numbers.

The warning about heat rings a bell but my recollection is it was a single page and had a statement about being kept confidential.

As I have said I tend not to keep a record once I have found what I want and resolved the issue. I know I posted about it at HC at the time but not sure if I was able to copy and paste the document to HC or if I just gave a link.

Your find confirms that if you put Hailo in a case you could generate sufficient heat to poach an egg but then again it might combust or melt first.

PS: Just read pdf and they provide max power use 8.65 watts depending on application. Getting up around Jetson Nano territory.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Mccabe84]

I can't see the SP going anywhere until an application hits the market at commercial scale ie millions in continual royalties. The market has decided we won't believe until we see it.

Market/shorts have decided to keep the price where it is. 2nd half of the year I expect some more action or it might be time for me to reconsider my investment

 

[GDJR69]

I can't see the SP going anywhere until an application using Akida IP hits the market at commercial scale ie millions in continual royalties. The market has decided we won't believe until we see it.

Once some sales kick in watch the SP snap back the other way like an elastic band that has been stretched as far as possible. In my opinion, this stock will take off like a bushfire with some revenue and/or another good IP agreement. I'm quite sure it will happen and happy to wait for it.

 

[jtardif999]

STMicroelectronics & Lacroix Group were on my radar the other day.

Akida could be in ovens monitoring the fan or in people flow detectors.

Case study 1: AI solution for people counting sensor​

Making buildings smarter is one of the big challenges of today's companies to improve their efficiency. The people flow counting sensor developed by Schneider Electric in partnership with STMicroelectronics enables the counting of the number of people. It also detects whether they are crossing a virtual line in both directions, using a large field of view and a small resolution thermal sensor.

This prototype can count in real-time and with a high level of accuracy the restaurant's attendance, while running on the standard STM32 microcontroller. This is achieved thanks to the artificial intelligence algorithm embedded on the STM32 microcontroller and the use of a thermal infrared technology.

View attachment 31553

​

​

Case study 2: Low-power predictive maintenance + AI at the Edge​

Lacroix Group and its ecosystem are building the future of industrial electronics, in the design and production of industrial embedded systems and connected objects. At the heart of its smart industry strategy, Lacroix Electronics is now experimenting with predictive maintenance on its own assembly lines with the help of STMicroelectronics and its AI ecosystem.

The first trial of the condition monitoring technology is being done on the reflow oven of an automated line that solders component on PCB boards.

View attachment 31544

Artificial Intelligence @ ST​

STM32Cube function pack for high performance STM32 with artificial intelligence (AI) application for Computer Vision.
View attachment 31547



Artificial Intelligence (AI) condition monitoring function pack for STM32Cube.
View attachment 31548

STM32Cube function pack for ultra-low power IoT node with artificial intelligence (AI) application based on audio and motion sensing.
View attachment 31549

https://www.st.com/content/st_com/en/campaigns/artificial-intelligence-at-the-edge.html

Give your product an Edge​

Simple, fast, optimized. Our extensive solutions​

for embedded AI.​

A set of tools to enable Edge AI​

on STM32 MCU, MPU and smart sensors​

Embedded AI can improve many solutions in a simple, fast, and cost-effective way.
Predictive maintenance, IoT products, smart buildings, asset tracking, people counting and more.
Learn how these applications can become smarter by making data meaningful with machine learning!

Be inspired by the possibilities of Edge AI - STMicroelectronics - STM32 AI

stm32ai.st.com

Be inspired by the possibilities of Edge AI - STMicroelectronics - STM32 AI

stm32ai.st.com

Be inspired by the possibilities of Edge AI - STMicroelectronics - STM32 AI

stm32ai.st.com

Customers:
View attachment 31554 View attachment 31555 View attachment 31556 View attachment 31557

Give your product an Edge

Free SW tools to embed Machine Learning on STM32 MCU and MPU. Edge AI can improve your product in a simple, fast, and cost-effective way.

stm32ai.st.com

All those years ago SNAP (Akida’s grand father) was counting cars from a bridge over a California freeway - sparked my initial interest in BrainChip and my continued love for this technology. Now counting people entering a building to conserve energy, beneficial ai hey.

 

[Fact Finder]

It is interesting how a group of investors called shorts are not referred to as FCB’s (Future Confirmed Buyers).

Shorts do not own shares in a company after they take their position. A Short or FCB after taking a position has a future obligation to buy in this case Brainchip shares to return them to the lender. They cannot avoid this obligation and in most cases have given security to the lender to guarantee they have the capacity to buy back.

So when a company has FCB’s shareholders know there are guaranteed buyers for their shares if they need to sell.

Yet shareholders worry about FCB’s.

The FACT that FCB’s exist gives rise to panic buying when an event of unexpected origin incites new retail shareholders interest.

My opinion only DYOR
FF

AKIDA BALLISTA

PS: Obviously ‘the more to come’ over the next couple of weeks has been missed by most.

 

[Rise from the ashes]

Hi @Rise from the ashes

Could well be but my memory had power just listed as 3 to 7 watts in a box with columns covering various numbers.

The warning about heat rings a bell but my recollection is it was a single page and had a statement about being kept confidential.

As I have said I tend not to keep a record once I have found what I want and resolved the issue. I know I posted about it at HC at the time but not sure if I was able to copy and paste the document to HC or if I just gave a link.

Your find confirms that if you put Hailo in a case you could generate sufficient heat to poach an egg but then again it might combust or melt first.

PS: Just read pdf and they provide max power use 8.65 watts depending on application. Getting up around Jetson Nano territory.

My opinion only DYOR
FF

AKIDA BALLISTA

Yes she's definitely fit for central heating purposes.
If I kept all the information gathered regarding this stock and similar companies in paper form I'd have to rent out a large warehouse. I even gave up with storing it electronically as I'm not really great with labelling folders.
One humongous jumble sale of information that I honestly couldn't be stuffed sorting through. .

 

[robsmark]

And there it is, our daily reference to a completely irrelevant WBT sp.

It has every relevance to us:
- a tech stock in the same industry;
- a similar target audience;
- a pre-revenue tech stock on the ASX;
- facing the same “economic headwinds”;
- a similar market cap;
- a less developed company…

Need I go on?

 

[Pappagallo]

And there we have it - back to pre announcement SP.

How the market doesn’t think this is worth money is beyond the joke. WBT now worth $1.4b and we’re sitting at $900m.

Swings and roundabouts. WBT will overshoot and come back, just like BRN did with the Mercedes news. Then the momentum will shift again at some point, probably after an IP license announcement or two. I have a feeling there will be a few “lead changes” between these two over the coming years.