BRN Discussion Ongoing

[Xray1]

Hi ML,

The Board can only change or influence things which are within their control or sphere of influence.

What would you have the board do differently to avoid the 2nd strike?

Diogenese: you asked the following question ......

" What would you have the board do differently to avoid
the 2nd strike? "

The answer might be quite simple .....

IMO, by thus putting the "Renumeration Reports" currently on hold till such time as the Co actually brings in at the very least a "Break Even" revenue stream.

IMO, if the BOD and Management aren't even prepared to wait for such an event to take effect, then that will only cause me further concerns as to where things are really at with this Co and the take up of Akida IP as well as the actual formalisation of any contracts / Royalties.

In essence, it is also my view that this type of approach can equate to the BOD and Management of putting some of their own personal skin and $$$ into the Co as well as promoting a positive Co sentiment and s/holder satisfaction that things are expected to vastly improve in the not too far distant future.

 

[KKFoo]

Good afternoon, I have some questions for the technical experts in this forum,
1)If you use a Brainchip SNN to analyse the video output from a normal camera it can be used to detect only the changes in the frames and gives you something similar to Prophesee's DVS sensor output?
2) What are the differences between those 2 systems and what are Brainchip+camera advantages over Prophesee DVS sensor?
3) What happen when you use Brainchip SNN with Prophesee DVS sensor? How is it better than Brainchip+normal camera combination?
Appreciate if anybody can help to answer..Thanks

 

[chapman89]

Good afternoon, I have some questions for the technical experts in this forum,
1)If you use a Brainchip SNN to analyse the video output from a normal camera it can be used to detect only the changes in the frames and gives you something similar to Prophesee's DVS sensor output?
2) What are the differences between those 2 systems and what are Brainchip+camera advantages over Prophesee DVS sensor?
3) What happen when you use Brainchip SNN with Prophesee DVS sensor? How is it better than Brainchip+normal camera combination?
Appreciate if anybody can help to answer..Thanks

From Rob Telson-

1. BrainChip has developed an AI Neural Network Accelerator. It can operate independently as a NPU (Neural Processing Unit) or as a companion to a CPU. We license our technology as IP which will be designed into a broader SoC (System on a Chip).
2. The importance of our IP is that eventually over time more and more devices and applications will demand intelligence and more intelligence will be designed into the SoC. BrainChip is very well positioned to capture this market as it continues to gain momentum over the next few years.
3. We feel strongly that we are very well positioned because:
   1. There are not many AI solutions that are IP based
   2. There are very few neuromorphic solutions
   3. We feel neuromorphic will be the architecture of choice because of it’s low power and compute efficiency advantages that will be impactful in the growth of AI
   4. We are the first company to commercialize a neuromorphic IP based solution
4. Prophesee offerings a neuromorphic based sensor. The sensor is a different technology than the NPU. A sensor can leverage the benefits of the neuromorphic architecture (in this case for vision) and deliver very favorable results for the overall system or product. Combined with a neuromorphic NPU it provides an even better solution.
5. Prophesee partnered with Synsense, a neurmorphic “Chip” company a while ago to provide a combined solution of the vision sensor with the neuromorphic chip for very low cost applications, such as toys. Synsense does not license their technology as IP.
6. BrainChip and Prophesee have demonstrated the technical advantages of our combined technologies and how the end product will benefit. Time will tell how many companies will adopt Prophesee’s Vision sensor with BrainChip’s Akida.

 

[jtardif999]

Diogenese: you asked the following question ......

" What would you have the board do differently to avoid
the 2nd strike? "

The answer might be quite simple .....

IMO, by thus putting the "Renumeration Reports" currently on hold till such time as the Co actually brings in at the very least a "Break Even" revenue stream.

IMO, if the BOD and Management aren't even prepared to wait for such an event to take effect, then that will only cause me further concerns as to where things are really at with this Co and the take up of Akida IP as well as the actual formalisation of any contracts / Royalties.

In essence, it is also my view that this type of approach can equate to the BOD and Management of putting some of their own personal skin and $$$ into the Co as well as promoting a positive Co sentiment and s/holder satisfaction that things are expected to vastly improve in the not too far distant future.

I think (and I could have misunderstood this) they stated at the AGM that all remuneration was based on them achieving results in a 1-3 year time frame, so I think they are not collecting on remuneration unless they have achieved the outcomes that have been set in that time frame?

 

[Diogenese]

Good afternoon, I have some questions for the technical experts in this forum,
1)If you use a Brainchip SNN to analyse the video output from a normal camera it can be used to detect only the changes in the frames and gives you something similar to Prophesee's DVS sensor output?
2) What are the differences between those 2 systems and what are Brainchip+camera advantages over Prophesee DVS sensor?
3) What happen when you use Brainchip SNN with Prophesee DVS sensor? How is it better than Brainchip+normal camera combination?
Appreciate if anybody can help to answer..Thanks

With an ordinary camera, a pre-processor is needed to convert the camera signals to spikes for use in Akida.

Because it is always on (no shutter) and only detects pixel illumination changes, the DVS can have a much higher equivalent frame rate, say 10,000 fps.

Prophesee tried Synsense, but it was not accurate enough because it was analog (SPECK) and could not match the Prophesee frame rate. Synsense are pursuing low hanging, non-critical applications (toys). They recently developed a digital NN for sound.

CORRECTION: SPECK is digital.

 

[KKFoo]

From Rob Telson-

1. BrainChip has developed an AI Neural Network Accelerator. It can operate independently as a NPU (Neural Processing Unit) or as a companion to a CPU. We license our technology as IP which will be designed into a broader SoC (System on a Chip).
2. The importance of our IP is that eventually over time more and more devices and applications will demand intelligence and more intelligence will be designed into the SoC. BrainChip is very well positioned to capture this market as it continues to gain momentum over the next few years.
3. We feel strongly that we are very well positioned because:
   1. There are not many AI solutions that are IP based
   2. There are very few neuromorphic solutions
   3. We feel neuromorphic will be the architecture of choice because of it’s low power and compute efficiency advantages that will be impactful in the growth of AI
   4. We are the first company to commercialize a neuromorphic IP based solution
4. Prophesee offerings a neuromorphic based sensor. The sensor is a different technology than the NPU. A sensor can leverage the benefits of the neuromorphic architecture (in this case for vision) and deliver very favorable results for the overall system or product. Combined with a neuromorphic NPU it provides an even better solution.
5. Prophesee partnered with Synsense, a neurmorphic “Chip” company a while ago to provide a combined solution of the vision sensor with the neuromorphic chip for very low cost applications, such as toys. Synsense does not license their technology as IP.
6. BrainChip and Prophesee have demonstrated the technical advantages of our combined technologies and how the end product will benefit. Time will tell how many companies will adopt Prophesee’s Vision sensor with BrainChip’s Akida.

Thanks Chapman for your fast response, I know that Brainchip can handle 5 modalities and can go into many MCUs, but Phrophesee only handle vision, what if we just focus on vision(video), if you want to build a smart door bell, what combination would you use? Brainchip SNN + normal camera+ Renesas RA MCU, PropheseeDVS+Renesas RA MCU, or Brainchip SNN+Prophesee DVS+ Renesas RA MCU or just Renesas RA MCU as demoed in their video? I hope my questions make sense..

 

[KKFoo]

With an ordinary camera, a pre-processor is needed to convert the camera signals to spikes for use in Akida.

Because it is always on (no shutter) and only detects pixel illumination changes, the DVS can have a much higher equivalent frame rate, say 10,000 fps.

Prophesee tried Synsense, but it was not accurate enough because it was analog (SPECK) and could not match the Prophesee frame rate. Synsense are pursuing low hanging, non-critical applications (toys). They recently developed a digital NN for sound.

Thanks Diogenese, I thought Akida is the NPU that converts normal video to spikes, no preprocessing is required?

 

[jtardif999]

Good afternoon, I have some questions for the technical experts in this forum,
1)If you use a Brainchip SNN to analyse the video output from a normal camera it can be used to detect only the changes in the frames and gives you something similar to Prophesee's DVS sensor output?
2) What are the differences between those 2 systems and what are Brainchip+camera advantages over Prophesee DVS sensor?
3) What happen when you use Brainchip SNN with Prophesee DVS sensor? How is it better than Brainchip+normal camera combination?
Appreciate if anybody can help to answer..Thanks

Also to add to what @chapman89 has said:
1) Akida is agnostic regarding the sensor (and/or cpu) it is connected to.
2) A normal video stream from a camera sensor is frame based and Akida can process and infer from ordinary pictures at a very high rate at a very low power rate.
3) A DVS sensor such as the technology developed by Prophesy sends the pixel changes in a scene and some ordinary frame based filler data merged together to Akida to process. The pixel changes are akin to events and match to the spike input that Akida expects, so it processes the input much more quickly than conventional video.

Hope this helps

 

[DingoBorat]

Wow, I mean, WooooooooooooooW

Thanks for this offering chappy.

Much, MUCH, food for thought in what your saying, MUCHO, MUCHO GRACIAS Amigo, hope you're on the money ?????

" I like it, I like it, - I like the way, you run, your fingers through my hair " Remember that one ??

Nothing personal, of course ( know what I mean, know what I mean !! Say no more, Say no more )

Thanks for that reality check.

AKIDA BALLISTA >>>>> BINGO maybe - anyone <<<<<

hotty...

This guy probably would have got swatted, long before he finished, in reality..

 

[KKFoo]

Also to add to what @chapman89 has said:
1) Akida is agnostic regarding the sensor (and/or cpu) it is connected to.
2) A normal video stream from a camera sensor is frame based and Akida can process and infer from ordinary pictures at a very high rate at a very low power rate.
3) A DVS sensor such as the technology developed by Prophesy sends the pixel changes in a scene and some ordinary frame based filler data merged together to Akida to process. The pixel changes are akin to events and match to the spike input that Akida expects, so it processes the input much more quickly than conventional video.

Hope this helps

Thanks J, So in other words, normalcamera+Akida=Prophesee DVS sensor, right? On top of that Akida inside the camera can do more such as inferencing what gestures is showing in the video whereas Phrophesee DVS is just a sensor, it cannot differentiate the gestures, right?

 

[Diogenese]

Thanks Diogenese, I thought Akida is the NPU that converts normal video to spikes, no preprocessing is required?

Akida has a preprocessor. This is separate from the NN.

Preprocessing takes time and uses power, and this is necessary for standard video. With DVS, this process is not needed because Akida can handle the native spiking signals from DVS directly.

look at this way, DVS does the pre-processing.

 

[Diogenese]

Thanks J, So in other words, normalcamera+Akida=Prophesee DVS sensor, right? On top of that Akida inside the camera can do more such as inferencing what gestures is showing in the video whereas Phrophesee DVS is just a sensor, it cannot differentiate the gestures, right?

Akida plus normal video camera is much slower than Akida + Prophesee because of the shutter speed of the video camera.

Akida plus Prophesee is ideal for gesture recognition.

 

[KKFoo]

Thanks D, I am getting a clearer picture ,now I understand why we are not in the door bells or toys, it is because "good enough" solutions already exist, Brainchip has to focus on those higher end applications where "good enough" is not going to fly..

 

[JDelekto]

Thanks J, So in other words, normalcamera+Akida=Prophesee DVS sensor, right? On top of that Akida inside the camera can do more such as inferencing what gestures is showing in the video whereas Phrophesee DVS is just a sensor, it cannot differentiate the gestures, right?

A normal camera's video will need to be processed for each frame in order to determine the differences in the images and generate the spikes needed to feed to Akida. The Prophesee sensor converts that video to the spikes directly that can be processed by Akida. As @Diogenese points out, the Prophesee sensor has a much higher framerate than a normal camera and a step to convert a change in a normal camera's image to a spike train is not required.

Both the Prophesee sensor and Akida are complementary technologies, one does not substitute the other. Their combination may provide the best power savings and performance while eliminating the need for any additional hardware/software translation between the two.

Regarding your earlier combination about the addition of the MCU into the mix, such as with an intelligent doorbell, a very low-powered MPU would most likely be used in order to add some programmability and interface with other hardware, such as Bluetooth or WiFi antenna and perhaps updating models on the device. When it comes to inferencing, the MCU does not need to be used to do this, as that can be handled by Akida directly.

 

[JDelekto]

Thanks D, I am getting a clearer picture ,now I understand why we are not in the door bells or toys, it is because "good enough" solutions already exist, Brainchip has to focus on those higher end applications where "good enough" is not going to fly..

I should point out that while there are "good enough" solutions that exist for toys and doorbells, batteries and electricity do not have a non-zero cost involved.

A small anecdote: I live in a very old house that was built in the late 40's and wasn't wired for a doorbell. I couldn't believe that barbarians used to knock on doors back in the day, go figure.

Joking aside, I purchased a Ring doorbell and because I couldn't easily hardwire it, I opted to go the "recharging" route. After a couple of months, I would have to recharge the thing at what seemed like every few days. I would have welcomed a "solar-powered" solution, but a panel to keep it charged would probably be larger than the device itself.

It became such a bother that I ditched it and opted to go with a simple wireless chime I could plug inside the house and a button, powered by a calculator battery and affixed with an adhesive strip. It only needs to detect the button press and alert the internal chime. It's been over a year and I have not yet had to change the battery.

There are probably several similar solutions that exist that can be improved by magnitudes in performance and power savings. Solutions that use neuromorphic sensors and processors will give the company that employs them an extreme competitive advantage.

 

[buena suerte :-)]

Thanks buena suerte >>>>> BOB <<<<<

for that concise reply. Now, I'm in the know on this score. I can see the significance for the next meeting ( AGM ) will be very interesting indeed.

I must brush up I reckon from here on in, of my understanding about s/h controls IMVHO.

Thanks for your time in replying, much appreciated. And good luck with your investments.

Regards, hotty...

Akida Ballista

Hey hotty,

My pleasure... A nice easy explanation of what it all means!

It is something for all shareholders to consider when voting!!!

Some might not be happy with the whole BRN situation at the moment but if the votes go the way of a 'second strike' then that IMO would be an absolute nightmare for BRN moving forward (no confidence!!)

When a ‘second strike’ occurs, the shareholders will vote at the same AGM to determine whether all the directors will need to stand for re-election.....

Have a great weekend and Good luck to you et al

 

[buena suerte :-)]

If solid progress is being made and evident that AKIDA is being adopted despite little revenue it would be madness to 2nd strike.

Absolutely!!

 

[buena suerte :-)]

Thanks for posting this, I was about to Google what it would entail.

I genuinely couldn't think of anything worse than losing any of the talented board members we have.
If people think voting it down, then having them re-elected any way as a middle-finger message is going to help you're on another planet.
When they are already under criticism for performance I couldn't imagine they SP should any of them be cast out.
Given the majority of the histeria stems the SP more than the company engaging with current/potential clients, I Believe with 99.99% certainty we wouldn't be having this conversation if the SP was closer to $1.

It's getting pathetic now, if it's not out of your system, maybe drop your final say and let's move on.

No probs... Cheers Damo

 

[Schnitzel lover]

I may be wrong but revenue has actually increased year over year?

Remains to be seen whether revenue is increasing yoy unless they can pull a $5-6mill qtr or half in the H2 2023 half..

Sean said lumpy revenue, it’s happening.
The company has been partnering with 1-2 companies every month, clearly stating they will adopt akida. They will not buy a board and simply evaluate it, they said they will adopt and drive akida into products.

Would prefer to see them licensing Akida. Something I’d like to hear if they intend to do so or if that’s not required because of third party partners with license already.

Development cycles takes time with IP, 2-3-4 years, it is not Brainchips role to update at every achievement of their partners that they make and announce where they are up to!

Still IP licenses would be indicative of their intent.

An example, Renesas/Meagchips license the IP, Brainchip would not be allowed to tell shareholders exactly where they are up to in the cycle, that is up to Renesas.
Brainchips hands are tied running with the IP model, it is simply not up to Brainchip when all the boxes are ticked with prototype, design, test, implement, production with those companies to announce everything.

Sean stated at the AGM that one of their licensees would have akida in commercial products “hopefully” this year.
Now let me ask you all, if there’s been delays, is that Brainchips fault? Absolutely not.

Unknown if delayed.

As Antonio said, you don’t hear what’s being designed into the iPhone 16 or the Galaxy S26, it’s just not how it works. Yes some less sophisticated companies may announce things, but most won’t. We all need to get used to it otherwise our money is best invested elsewhere.

Let me say this, let’s say Brainchip was a private company not listed on a stock exchange, and we looked at Brainchips technology, the board/management, the massive growing ecosystem which is 1-2 a month, the industries in which these companies are involved in and service, the recognition we and neuromorphic are getting, would we be complaining? No we wouldn’t.
Show me another neuromorphic company that has the ecosystem we have? Show me another neuromorphic company that has the experience of the board we have?
No other company has what we have.

But it is a listed company and the cost of capital gets exceedingly more expensive as a company’s market valuation declines..

People are using Chris stevens departure as some type of negative? How is it a negative?? Look at his employment history, it wasn’t exactly stable to begin with and at the end of the day he was in it for money, he clearly got a better offer at Untether AI and good luck to him, maybe he couldn’t handle selling IP? He never sold Neuromorphic IP before, nor has anybody in the world for that matter.

Well he was pouted as the head of sales in Sean’s 2022 AGM address, a narrative that was positively positioned for BRNs sales efforts..

When Nandan left Amazon did that mean Amazon was a bad sign for Amazon? NO.

When Duy-Loan left Texas Instruments did that mean it’s a bad sign for Texas Instruments? NO.

When Rob Todd Antonio Nandan and our Japanese Sales guy all left ARM at different stages did that mean it was a bad sign for ARM? NO!!

When our recent product development manager just left Facebook (Meta) and joined Brainchip does that mean that it’s looking bad for Meta? NO.

When Sean left Hewitt Packer (HP) as Global Ecosystem boss and joined Brainchip did that mean that HP was going under? NO.

People leave their jobs for all sorts of reasons.
People need to stop looking at Chris stevens departure with the eyes of a shareholder and start looking at it like he got a better opportunity which he did!!

That’s your opinion.. Many feel differently.

This argument and the same people on this forum who keep projecting their emotional insecurity of their investment are just that, emotional.

If you see your investment go down it’s a human emotion to get emotional about loss of capital value..

I just read a post above that the share price may go to 6 cents? Please let it go to 6 cents.

If BRN goes to 6c, it is indeed in big trouble and things are worse than we hope.. As the listed public market where shares can be traded every day, progress is often if not always priced in 3-6 months in advance.

With Renesas coming to market in the next few months, large tier ones working on driving akida into applications, with akida being rocketed into space in January, with neuromorphic being written about more and more and with the technology and board and ecosystem that Brainchip have and growing, I’ll keep buying shares and reap the rewards, whilst others suffer in fear.

Throwing good money after bad if that’s your investment approach based on future hopes.. Go ask Mark Minervini or Gil Morales or any good investor with a measurable record if they’d buy Brainchip shares.. I can tell you the answer would be not now.. Are they wrong? Their records would suggest not..

People actually think that it’s like selling a new energy drink to the servo or the convenience store, we are talking about a technology that is ahead of its time, never been worked on before until recently.

Ill take my chances with Brainchip over any other neuromorphic company out there.

Good luck with that risk management profile.. 1 in 100 with BRNs hype make it as a company, so as much as I want to agree with you, the stats say it’s unlikely.

Let me say it again, if Brainchip was not a publicly listed company, and we knew the rapidly growing ecosystem, the partners we have, the multiple different industries, the experience of the board & management we have, the technology, the growing recognition, the soon to hit the market in mass volume mcu’s containing akida and everything else, we would be jumping for joy, only question we all want to know is how long is it going to take, but it is a wait I am happy to endure.

I’d like to think it’s that simple as it’s only a matter of time, but there are many assumptions there. Until all those show up in revenues then it’s all just speculation..

I’d consider success a few new IP licenses signed by the next AGM, and some royalties in the H2 2023 report, and atleast something in the single digit $millions In royalties > $2-3mill by the H1 2024 half yearly report..

If MCUs are indeed mass produced and used then happy days, no more worries with a low market valuation or need to worry about shorter term capital costs…

Tough market to crack, but hopefully as you say it’s just a matter of time..

 

[Diogenese]

GPT SNN

This paper was referred to by Dylan Muir (Synsense) in the recent interview with Sally Ward-Foxton discusses the use of SNNs in GPT. Who's in the right place, right time?

2302.13939.pdf (arxiv.org) https://arxiv.org/pdf/2302.13939.pdf

As the size of large language models continue to scale, so does the computational resources required to run it. Spiking Neural Networks (SNNs) have emerged as an energy-efficient approach to deep learning that leverage sparse and event-driven activations to reduce the computational overhead associated with model infer ence. While they have become competitive with non-spiking models on many computer vision tasks, SNNs have also proven to be more challenging to train. As a result, their performance lags behind modern deep learning, and we are yet to see the effectiveness of SNNs in language generation. In this paper, inspired by the Receptance Weighted Key Value (RWKV) language model, we successfully implement ‘SpikeGPT’, a generative language model with binary, event-driven spiking activation units. We train the proposed model on two model variants: 45M and 216M parameters. To the best of our knowledge, SpikeGPT is the largest backpropagation-trained SNN model to date, rendering it suitable for both the generation and comprehension of natural language. We achieve this by modifying the transformer block to replace multi-head self attention to reduce quadratic com putational complexity O(N2) to linear complexity O(N) with increasing sequence length. Input tokens are instead streamed in sequentially to our attention mecha nism (as with typical SNNs). Our preliminary experiments show that SpikeGPT remains competitive with non-spiking models on tested benchmarks, while main taining 20× fewer operations when processed on neuromorphic hardware that can leverage sparse, event-driven activations.
...
Recall Self-Attention. The self-attention operation lies at the heart of Transformers. In Transformers, self-attention takes an input sequence X, and applies a scaled dot product attention.


Artificial Neural Networks (ANNs) have recently achieved widespread, public-facing impact in Natural Language Processing (NLP), but has come with a significant computational and energy consumption burden across training and deployment. As examples, training GPT-3 was projected to use 190,000 kWh of energy [1; 2; 3]. Deploying ChatGPT into every modern word processor will witness millions of users in need of on-demand inference of large language models [4]. SNNs, inspired by neuroscientific models of neuronal firing, offer a more energy-efficient alternative by using discrete spikes to compute and transmit information [5]. Spike-based computing combined with neuromorphic hardware holds great potential for low-energy AI [6 (Mike Davies loihi); 7; 8], and its effectiveness in integration with deep learning has been demonstrated through numerous studies [9; 10; 11; 12; 13; 14]. At this stage, the performance of Spiking Neural Networks (SNNs) in NLP and generation tasks remains relatively under-investigated. While SNNs have shown competitiveness in computer vision tasks such as classification and object detection [15; 16; 17], they have yet to attain similar success in generative models. Th