BRN Discussion Ongoing

[Steve10]

Seriously SynSense are dreaming with their pricing for Speck.

Somebody requested pricing for one unit for home hobby & price is $4k. Try putting that into a smartphone. LOL

Small market only for very expensive low volume industrial applications. Appears to do object classification, object detection, object tracking etc but no learning & it uses CNN (not ideal for Prophesee).

jason​

10 days ago (edited)
Can you purchase one (speck dvs camera) at the hobbyist level (a few hundred dollars?)
Open Neuromorphic
8 days ago
You mean the Speck SoC? Unfortunately, the price for that one is set to 4k$, since it is an ASIC recently introduced.

For those that don't know what Speck is here are details:

Speck™ is a fully event-driven neuromorphic vision SoC. Speck™ is able to support large-scale spiking convolutional neural network (sCNN) with a fully asynchronous chip architecture. Speck™ is fully configurable with the spiking neuron capacity of 320K. Furthermore, it integrates the state-of-art dynamic vision sensor (DVS) that enables fully event-driven based, real-time, highly integrated solution for varies dynamic visual scene. For classical applications, Speck™ can provide intelligence upon the scene at only mWs with a response latency in few ms.

Speck™: Event-Driven Neuromorphic Vision SoC | SynSense

Speck is the world’s first fully event-driven neuromorphic vision SoC using the dynamic vision sensing and the spiking neural network technology.

www.synsense.ai

 

[Steve10]

The PAA can only be bestowed with an estimate of SP for the corresponding periods Dec 2023, 2028, 2030

Who's first?

 

[Fact Finder]

Who's first?

View attachment 32720

I'll go last after we hit $2.75.

 

[Fact Finder]

Seriously SynSense are dreaming with their pricing for Speck.

Somebody requested pricing for one unit for home hobby & price is $4k. Try putting that into a smartphone. LOL

Small market only for very expensive low volume industrial applications. Appears to do object classification, object detection, object tracking etc but no learning & it uses CNN (not ideal for Prophesee).

jason​

10 days ago (edited)
Can you purchase one (speck dvs camera) at the hobbyist level (a few hundred dollars?)
Open Neuromorphic
8 days ago
You mean the Speck SoC? Unfortunately, the price for that one is set to 4k$, since it is an ASIC recently introduced.

For those that don't know what Speck is here are details:

Speck™ is a fully event-driven neuromorphic vision SoC. Speck™ is able to support large-scale spiking convolutional neural network (sCNN) with a fully asynchronous chip architecture. Speck™ is fully configurable with the spiking neuron capacity of 320K. Furthermore, it integrates the state-of-art dynamic vision sensor (DVS) that enables fully event-driven based, real-time, highly integrated solution for varies dynamic visual scene. For classical applications, Speck™ can provide intelligence upon the scene at only mWs with a response latency in few ms.

View attachment 32719

Speck™: Event-Driven Neuromorphic Vision SoC | SynSense

Speck is the world’s first fully event-driven neuromorphic vision SoC using the dynamic vision sensing and the spiking neural network technology.

www.synsense.ai

Actually they might be onto something here. Brainchip's Studio if memory serves me at $3,800 failed to sell in a solid world economy without wars all over the shop and Europe having plenty of energy.

In today's world with banking collapses and wars and trade sanctions rounding up to $4,000 might very well be a winner as it prevents hobbyists and universities trialling it in a side by side comparison with AKIDA and Prophesee who are giving them to Universities along with free course materials and guest lecturers. Not being a listed company anywhere as well allows them to produce material and make claims that might not pass muster as a listed company. Probably a win, win for them.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Diogenese]

Seriously SynSense are dreaming with their pricing for Speck.

Somebody requested pricing for one unit for home hobby & price is $4k. Try putting that into a smartphone. LOL

Small market only for very expensive low volume industrial applications. Appears to do object classification, object detection, object tracking etc but no learning & it uses CNN (not ideal for Prophesee).

jason​

10 days ago (edited)
Can you purchase one (speck dvs camera) at the hobbyist level (a few hundred dollars?)
Open Neuromorphic
8 days ago
You mean the Speck SoC? Unfortunately, the price for that one is set to 4k$, since it is an ASIC recently introduced.

For those that don't know what Speck is here are details:

Speck™ is a fully event-driven neuromorphic vision SoC. Speck™ is able to support large-scale spiking convolutional neural network (sCNN) with a fully asynchronous chip architecture. Speck™ is fully configurable with the spiking neuron capacity of 320K. Furthermore, it integrates the state-of-art dynamic vision sensor (DVS) that enables fully event-driven based, real-time, highly integrated solution for varies dynamic visual scene. For classical applications, Speck™ can provide intelligence upon the scene at only mWs with a response latency in few ms.

View attachment 32719

Speck™: Event-Driven Neuromorphic Vision SoC | SynSense

Speck is the world’s first fully event-driven neuromorphic vision SoC using the dynamic vision sensing and the spiking neural network technology.

www.synsense.ai

A really exciting parlour game is to combine darts with pin-the-tail-on-the-donkey.

 

[wasMADX]

I was reading Science Illustrated Issue#96 p.38 https://pubhtml5.com/kvtj/jffv/Science_Illustrated_Australia_2023/.

It says, animals given a mirror to look in who try to remove a mark put on the forehead (under prior sedation) are judged to be self-aware.
Macaques don't try to remove the mark unless they are rewarded if they touch it. Before that, they don't care. Other tests show they can be clever, e.g. can remember 80% of numbers in a test vs 40% by humans.

I was wondering if the same principle can be applied, or is already, to Akidas i.e. can learn by rewarding a prior ignored piece of important data.

 

[Diogenese]

Popular Science:

Why the Air Force wants 1,000 new combat drones​

Story by Kelsey D. Atherton • Yesterday 10:00 pm

https://www.msn.com/en-au/news/tech...1&cvid=004d1971adbe4ab88b61e54026f6a087&ei=76

Scroll down to the video:
1:25 - "a high performance aircraft capable of autonomous action"

Will this use the brains of the Boeing "Loyal Wingman" developed in Australia?

https://en.wikipedia.org/wiki/Boeing_MQ-28_Ghost_Bat

The Boeing MQ-28 Ghost Bat, previously known as the Boeing Airpower Teaming System (ATS) and the Loyal Wingman project, is a stealth, multirole, unmanned aerial vehicle in development by Boeing Australia for the Royal Australian Air Force (RAAF). It is designed as a force multiplier aircraft capable of flying alongside manned aircraft for support and performing autonomous missions independently using artificial intelligence.[4]

 

[Mugen74]

Hey I like cats, but any company that develops "cat like" intelligence, is going to be a failure.

How often is something like a C.A.I. doorbell, going to "feel" like working?

I can already imagine the frustration


"Hey! I know you know it's me, let me in damn it!"


View attachment 32712

Uh Oh! Lookout world.

 

[Euks]

Not a bad little read about all the players in the infotainment systems market.

No mention of Brainchip but interesting none the less.

Automotive Infotainment Systems Market - Growth, Trends, COVID-19 Impact, And Forecasts (2023 - 2028)

Currently, The Automotive Infotainment Systems market is valued at USD 24.54 billion, and it is expected to reach USD 36.60 billion, registering a CAGR of 6.76% during the forecast period. The impact of the COVID-19 pandemic on the automotive infotainment systems market was inevitable as the...

uk.sports.yahoo.com

 

[TopCat]

I came across this link on the ARM LinkedIn page about some new software kit developed for the Cortex M range. What I found interesting though is a comment at the bottom from an ecosystem partner Qeexo praising it. When I searched for info on Qeexo I discovered that they have been recently acquired by TDK to strengthen their ML and Edge position

Introducing Keil MDK Version 6

Keil MDK Version 6 delivers new features that are optimized for the entire Cortex-M and Ethos-U processor portfolio.

community.arm.com

The IoT runs on Arm, look at the Cortex-M processor family, which is undoubtedly the de-facto industry standard for embedded compute with over 9,500 different microcontrollers from 38 different silicon vendors available today. The demand for more machine learning (ML) capabilities in IoT applications is growing exponentially, and our roadmap has evolved to meet these needs with the introduction of the Cortex-M55and Cortex-M85 microcontrollers (MCUs), designed to address high-performance ML and DSP workloads, and Neural Processing Units such as the Ethos-U microNPU family which accelerates ML inference in resource-constrained embedded and IoT devices.

As IoT devices get smarter, developers are facing increasing software complexity that requires new development flows that create optimized ML models combined with efficient device drivers. It is therefore crucial that the software development platforms and tools we are providing to the ecosystem are evolving alongside our processor roadmap. The Keil Microcontroller Development Kit (Keil MDK) is the most comprehensive software development solution for Arm Cortex-M-based embedded, IoT, and ML edge device applications, and today I am excited to share that we are releasing Keil MDK Version 6, optimized for the entire Arm Cortex-M and Ethos-U processor portfolio. The compute capabilities of today’s microcontrollers require MLOps for deploying machine learning models and DevOps for validating complex software workloads.

Qeexo AutoML is designed to make machine learning accessible to a wider range of users, including businesses, researchers, and developers who may not have deep knowledge of ML. Based on sensor data, Qeexo AutoML automatically builds machine learning solutions that are optimized for Arm processors and is used in building a wide variety of applications including: industrial, mobile, home IoT, medical, and automotive. Combined with Keil MDK and Arm Virtual Hardware, the complete end-to-end ML workflow for embedded developers is now easier than ever.” Michael Gamble, Director Product Management, TDK Qeexo.

TDK to acquire Qeexo to enable complete smart edge platforms

/PRNewswire/ -- TDK Corporation (TSE: 6762) (CEO & President: Noboru Saito, hereinafter "TDK") announced that today TDK has agreed to acquire Qeexo, Co. (CEO:...

www.prnewswire.com

• TDK to acquire Qeexo, Co, a leading developer of automated machine-learning (ML) platform that accelerates the development of tinyML models for low power, always-on intelligent platforms
• TDK aims to further strengthen its ML expertise and simplify ML application development to become a leader in delivering smart edge solutions
• Acquisition enables TDK to accelerate the transition to Industry 4.0 with smart edge solutions

 

[Lex555]

All the MCU's will not be for AI applications.

I would allow approx. 1% for AI enabled MCU's x 30B in 2023 = 300M chips TAM.

In three years I would allow 5-10% for AI enabled MCU's x 42B = 2.1-4.2B chips TAM.

2.1-4.2B chips x 10% BRN market share = 210-420M chips x 30c BRN royalty = $63-126M revenue.
(I have allowed 30c for royalty, however, AI enabled MCU's appear to be priced at around $20 x 2-3% royalty = 40-60c)

With 30% market share = $189-378M revenue x 60% EBITDA = $113.4-226.8M x 0.7 if taxed in Australia = $79.4-158.8M NPAT x PE60 =$4.76-9.52B MC.

It appears that most of the MCU market future growth will be in AI enabled MCU's.

By 2030 there will be approx. 70B MCU's per year - 30B in 2023 = 40B per year growth within 7 years.

Most likely 80% of the 40B MCU future growth will be for AI applications = 32B / 70B total = 45.7% of MCU's will be AI enabled by 2030 when there will be mass adoption.

So in 2030 if 32B MCU's are AI enabled x 10% BRN market share = 3.2B x 30c royalty = $960M revenue x 60% EBITDA = $576M x 0.7 ATO = $403.2M NPAT x PE60 = $24.2B MC. With 30% market share MC will be $72.6B.
(PE could be as high as 100. Nvidia is currently trading at PE 148.65)

Revenue growth will be similar to J-curve as mass adoption takes place. We are at the early adoption phase at the moment with early mass adoption expected within 3-6 years.

True, AI for MCU’s may only account for 1% of current market but perhaps let’s add some more use cases for existing MCU market including low power consumption, scalability and compact size.

Do you think royalty would be 30c per chip, my research is indicating an average sale price of $1 per chip and from Pitt St report I think off the top of my head iP royalty can charge between 5-15%, higher the more unique it is.

 

[GazDix]

Don’t get sucked into the ChatGPT vortex Steve.

She/he is a dirty little lying bastard most of the time

Yes, this is very true and sometimes obviously so.
There is a cryptocurrency that is looking at decentralising machine learning called Bittensor (Tao). They are also starting a Bittensor Chat which will have the same function as Chat GPT 3/4 pretty much but it will be decentralised. Anyway, here is what ChatGPT worked out:

 

[Steve10]

True, AI for MCU’s may only account for 1% of current market but perhaps let’s add some more use cases for existing MCU market including low power consumption, scalability and compact size.

Do you think royalty would be 30c per chip, my research is indicating an average sale price of $1 per chip and from Pitt St report I think off the top of my head iP royalty can charge between 5-15%, higher the more unique it is.

Renesas' RA6 MCU's sell for about $20 for volume. The new ARM Cortex-M85 based chips for top range RA MCU's to be released should be at least $20. Qualcomm pays about US 20-30c per chip to ARM & they are on a low royalty they agressively negotiated.

Trying to estimate royalty is impossible. Simply using ARM's pricing model of 2-3% for high volume. The numbers should be used a guide to see what is possible in different market sectors & market overall. They highlight that the TAM is huge & small percentages will result in significant revenue for BRN.

Can more accurately estimate/forecast once revenue commences & details of chip numbers are made available.

 

[Diogenese]

Popular Science:

Why the Air Force wants 1,000 new combat drones​

Story by Kelsey D. Atherton • Yesterday 10:00 pm

https://www.msn.com/en-au/news/tech...1&cvid=004d1971adbe4ab88b61e54026f6a087&ei=76

Scroll down to the video:
1:25 - "a high performance aircraft capable of autonomous action"

Will this use the brains of the Boeing "Loyal Wingman" developed in Australia?

https://en.wikipedia.org/wiki/Boeing_MQ-28_Ghost_Bat

The Boeing MQ-28 Ghost Bat, previously known as the Boeing Airpower Teaming System (ATS) and the Loyal Wingman project, is a stealth, multirole, unmanned aerial vehicle in development by Boeing Australia for the Royal Australian Air Force (RAAF). It is designed as a force multiplier aircraft capable of flying alongside manned aircraft for support and performing autonomous missions independently using artificial intelligence.[4]

We're not listed as a key company:
https://www.boeing.com.au/products-...echnology/boeing-airpower-teaming-system.page

... but do any of the Key Companies have AI expertise?

The MQ-28 Ghost Bat (formerly known as the Airpower Teaming System) is an uncrewed teaming aircraft designed to be a force multiplier in support of advanced multi-mission air combat operations.

Boeing Australia, supported by the Royal Australian Air Force and a local industry team, has produced the first military aircraft to be designed, manufactured and flown in Australia in over 50 years.

MQ-28 Ghost Bat development vehicles continue to be manufactured at Boeing’s Fishermans Bend facility in Victoria. Advanced robotics, composite materials and digital engineering have enabled key manufacturing innovations for the MQ-28 in the areas of robotic drill and fill, shimless assembly and full-size determinant assembly to significantly reduce assembly costs, compared with traditional methods.

Mission system and payload development continues, with Boeing’s digital engineering capabilities used to validate air vehicle teaming behaviours and examine the interaction of the MQ-28 Ghost Bat aircraft with other crewed and uncrewed assets.

The first flight of MQ-28 Ghost Bat was achieved in February 2021, with the test program expanding the flight envelope as the program evolves.



https://www.boeing.com.au/featured-...-capabilities-in-all-digital-environment.page

Boeing, RAAF validate Airpower Teaming System operational capabilities in all-digital environment​

• Latest digital tests progress mission system readiness

December 08, 2021

Boeing Australia and the Royal Australian Air Force (RAAF) have advanced the operational readiness of the Airpower Teaming System aircraft, successfully demonstrating multiple payloads, semi-autonomous behaviours and crewed-uncrewed teaming in the digital environment.

Given that NASA's SiFive RISC-V HPSC development may well include Akida2, and NASA's statement that it's not just for space (ie, defence etc) there is a possibility that GhostBat will include the HPSC, and there is a possibility that HPSC will include Akida. So, while we may not have directly been part of the GhostBat project, we may be part of the HSPC project.

 

[Steve10]

FYI

 

[Deleted member 2799]

Off topic

I'm not a baseball fan, but two Korean colleagues watched the Mexico-Japan semi-final in the office, it wasn't bad I must say...I never understood the rules, but today I somewhat understood! Tomorrow is the final between Japan and the USA...I'm taking bets...happy to accept Akida shares too! Let's go, go go!

 

[Steve10]

FYI - smart money buying the dip. It's the big funds.

When dumb money buying peaks the market peaks & vice versa.

 

[DingoBorat]

FYI

View attachment 32726

Love your work Steve10, but graphs and charts like that are a lie.

A weighted market, will always "recover".

 

[Kachoo]

Well I must say a word about my feelings.
Firstly i never imagined we would pull back this far but many events ans forces not honest have pushed us to the point of fear.

Fear makes humans react in ways they normally would not. It's likely our biggest decision maker.

1. FOMO we buy buy buy it's going to the moon
2. Paper losses scare tge crap out of people we sell sell sell.
3. Fear of death accident or misfortune.
4. Fear to fail at something we don't do.
We all have many things were fear will lead to decisions that are not the best in the long run. It's easy to push one and manipulate them with fear.

This is what is happening hear with people. They are fear full some are more savy and are lurking to get even bigger discounts if they can.

Then one thing I will say is that do I feel discomfort with SP action sure we all do. I feel fear when the wife asked how Brainchip is doing.

But I look for logic in this time.
The product is real the product is generating excitement and has generated sales. The product is going to me marketed and sold by one of the Top 5 chip makers. There is more Megachips is working and helping to make a market ans design.

Oh u forgot if Akida was a bust would NASA still be evaluating us and digging deeper in research and case use.

Ant61 is putting us in space to run a robot. Really why would they waste there robot with our chip is shit.

So science and the manufacturers are starting to use and study our capabilities and accept Nueromorphic Computing.

Schools are starting to teach it.

Some trading or investing WANCA will say this product is not marketable.

Who knows better?

 

[Shadow59]

Love your work Steve10, but graphs and charts like that are a lie.

A weighted market, will always "recover".

Hi DB are we talking "Lies, Damn Lies and Statistics"?