BRN Discussion Ongoing

From Weebit nano post on LinkedIn, very interesting.

Bennysmadness said:

From Weebit nano post on LinkedIn, very interesting.
View attachment 91676
View attachment 91677 View attachment 91678

Click to expand...

Thanks Benny,

I did have Weebit in mind for storing models when I mentioned the auxilliary memory.

https://www.weebit-nano.com/news/pr...n-to-advance-intelligent-systems-at-the-edge/

As a Strategic Partner, Weebit brings its low-power, high-performance Resistive RAM (ReRAM or RRAM) technology to this dynamic community. For advanced edge AI chips, Weebit ReRAM provides the dense on-chip non-volatile memory (NVM) needed to store weights for artificial Neural Networks (NNs) with ultra-low power consumption that is critical for edge devices. It also scales to the smaller process geometries used in the fabrication of advanced AI SoCs.

Diogenese said:

Hi Hoppy,

3 months ago in the Roadmap, JT said that GenAI was available as FPGA and they expected silicon (ASIC) within 12 months. Either that was conservative, or GenAI has been prioritized.

To make GenAI semi-autonomous wrt the cloud, it may use RAG (Retrieval Augmented Generation) with a lot of auxiliary memory.

They talk about 1 billion parameters as a yardstick. Given that this could be up to FP32, that's quite a lot of memory for an edge device.

But, apart from that, the nodes will probably still be the TENNs 128 MAC arrays requiring (comparatively) very little adjustment to the tapeout. That said, I suppose that the MACs will need to be adjusted to handle FP32, and maybe that's where the ISA (Instruction Set Architecture) comes in. SO, ok, there will be a lot of tapeout work required*.


* I'm submitting this to Guinnes for the fastest self-contradiction ...

Click to expand...

Has Marcus inadvertently spilled the beans?

Yes, this is off topic I know, BUT I have to comment on my AFL team the Brisbane Lions, congratulations guys and Chris Fagan
our coach, he reminds me of Peter, as in, a really caring individual who puts ego and self aside and considers others first, promotes
positivity and self-belief.

I've been a Brisbane supporter since around 1998, just before or around when Lee Matthews came on board as our coach, we were
around 15th or 16th on the ladder but went on to win the Premiership in 2001,2002 and 2003 and now under Fages we have secured
2024 and 2025, that's a brilliant achievement.

Over my years at the Joondalup Resort Golf Course in Perth I got to meet many Australian sporting heroes in Golf, Tennis, Cricket and
the AFL, including Alastair Lynch and Jason Akermanis of the Brisbane Lions, they love their golf, and I was quietly proud to meet them
on my own patch, so to speak.

So, I guess I have two loves currently, Brainchip and the Brisbane Lions ........ Tech.

ARQUIMEA Research Center’s neuromorphic vision solution for UAVs that deployed Akida with a Prophesee event-based sensor in a low power drone to detect distressed swimmers and surfers in the ocean and prevent drowning now officially has a name: MINERVA - Neuromorphic Vision for Safer Coasts!

“By 𝐫𝐞𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐧𝐠 𝐭𝐡𝐞 𝐧𝐞𝐮𝐫𝐚𝐥 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐢𝐧𝐠 𝐨𝐟 𝐭𝐡𝐞 𝐡𝐮𝐦𝐚𝐧 𝐛𝐫𝐚𝐢𝐧 in silicon-based electronics, neuromorphic technology 𝐩𝐫𝐨𝐜𝐞𝐬𝐬𝐞𝐬 𝐯𝐢𝐬𝐮𝐚𝐥 𝐢𝐧𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐡𝐢𝐠𝐡𝐞𝐫 𝐞𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲, 𝐥𝐨𝐰𝐞𝐫 𝐞𝐧𝐞𝐫𝐠𝐲 𝐜𝐨𝐧𝐬𝐮𝐦𝐩𝐭𝐢𝐨𝐧, 𝐚𝐧𝐝 𝐫𝐞𝐝𝐮𝐜𝐞𝐝 𝐥𝐚𝐭𝐞𝐧𝐜𝐲.
The system integrates multimodal sensing to perform classical computer vision tasks (localization, detection, classification) 𝐰𝐡𝐢𝐥𝐞 𝐨𝐯𝐞𝐫𝐜𝐨𝐦𝐢𝐧𝐠 𝐭𝐡𝐞 𝐜𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬 𝐨𝐟 𝐡𝐨𝐬𝐭𝐢𝐥𝐞 𝐜𝐨𝐚𝐬𝐭𝐚𝐥 𝐜𝐨𝐧𝐝𝐢𝐭𝐢𝐨𝐧𝐬. With this project, we 𝐚𝐝𝐯𝐚𝐧𝐜𝐞 𝐫𝐞𝐦𝐨𝐭𝐞 𝐜𝐨𝐚𝐬𝐭𝐚𝐥 𝐦𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠 𝐚𝐧𝐝 𝐭𝐡𝐞 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧 𝐨𝐟 𝐫𝐢𝐩 𝐜𝐮𝐫𝐫𝐞𝐧𝐭𝐬.”







FYI: #PULSARHRI refers to an ARQUIMEA robotics project first presented at the Yokohama IEEE Conference on Robotics and Automation in May 2024 and not to Innatera’s Pulsar neuromorphic ultra-low power microcontroller revealed in May 2025…

---

BrainChip’s previous press releases on this partnership:

cf. https://brainchip.com/brainchip-gives-the-edge-to-search-and-rescue-operations/

Diogenese said:

Thanks Benny,

I did have Weebit in mind for storing models when I mentioned the auxilliary memory.

https://www.weebit-nano.com/news/pr...n-to-advance-intelligent-systems-at-the-edge/

As a Strategic Partner, Weebit brings its low-power, high-performance Resistive RAM (ReRAM or RRAM) technology to this dynamic community. For advanced edge AI chips, Weebit ReRAM provides the dense on-chip non-volatile memory (NVM) needed to store weights for artificial Neural Networks (NNs) with ultra-low power consumption that is critical for edge devices. It also scales to the smaller process geometries used in the fabrication of advanced AI SoCs.


Has Marcus inadvertently spilled the beans?

Click to expand...

Finally, hopefully!

I still believe that embedding Akida and ReRam in the same SoC is a marriage made in heaven. Maybe one day!

Slymeat said:

Finally, hopefully!

I still believe that embedding Akida and ReRam in the same SoC is a marriage made in heaven. Maybe one day!

Click to expand...

Akida + ReRAM + LLM = BigBangBrain SoC for us???

my thougts: technically supported and feasable... large language models, or lightweight versions tailored for edge (that's what BC also do... tweaking and squeezing the code), can operate on an SoC featuring Akida neuromorphic cores and ReRAM memory technology. This "sensor" system can transform sensor intelligence with superior understanding, energy efficiency, and adaptability, while potentially balancing initial integration cost with long-term operational savings and enhanced capabilities compared to traditional sensor systems.

ReRAM as embedded non-volatile memory (>> retain data even when power is off, making it ideal for persistent storage << therefore it could also changed after a one-shot learning process by akida ... no need to store the data in a data center) can enhance on-chip memory storage capacity and speed compared to traditional RAM, but typical embedded memory sizes on SoCs remain limited compared to large GPU RAM. The economies of scale and increasing maturity of neuromorphic and ReRAM technologies may reduce costs in the medium to long term, making such systems financially viable for a range of industries including autonomous vehicles, industrial IoT, and advanced robotics.

This "BigBangBrain"-SoC would realistically be suitable for small to moderate-size models (up to a few billion parameters with optimizations), rather than the largest state-of-the-art models requiring large memory and compute resources found in high-end GPU clusters.

Spoiler?
- - - - -
Erik: So, Jonathan, it sounds like LLM on the edge is not quite here yet, but you're expecting it within one or two years to be capable. Can you explain?

Jonathan: Yeah, I would say less than that. We've demonstrated it, and we expect to have a product by the end of the year.

Source: https://asiagrowthpartners.com/podc...lms-on-the-edge-sean-hehir-ceo-brainchip/p285

If you thought we will have finally a price sensitive announcement today…. We have to disappoint you…

From March 2025 .

Weebit CEO : Coby










With one front indicating WeeBit and Alif Semiconductor . Alif also have a heavy reliance of product from Global Foundaries .

So another project has just recently been created on GitHub.

Not a lot of info and the author is just gv1x2 and no further details though I'll provide additional info I discovered, further in the post.

Whether the GitHub profile is one of the papers authors or someone with relevant datasets I'm not sure as the papers results were via simulations and using SpikingJelly (also under the same GitHub user files). However, the gv1x2 repository also now has the Akida Project and the papers "future work" is to implement their processes on actual neuromorphic hardware.

I'm presuming this is now the new Akida Project repository.

Files appear to involve spectrograms which:

Applications of Spectrograms

• Speech and Language Analysis:
  Analyzing the formant frequencies in vocal sounds to diagnose speech disorders or study language development.
  
  
• Music Production:
  Isolating specific notes or instruments, analyzing musical structure, or even embedding hidden images within songs.
  
  
• Bioacoustics:
  Studying wildlife sounds by visualizing and analyzing calls of birds, insects, and other animals.
  
  
• Audio Forensics and Repair:
  Detecting and isolating problematic noises, helping with the spectral repair of recordings.

On using general searching for the user name I found a reference to the GitHub repository in a recent paper published in Aug this year.


Abstract reveals the authors are working on telemedicine and in particular mental health.

From Convolution to Spikes for Mental Health: A CNN-to-SNN Approach Using the DAIC-WOZ Dataset​

by
Victor Triohin
1,
Monica Leba
2,* and
Andreea Cristina Ionica
3


1
Doctoral School, University of Petroșani, 332006 Petrosani, Romania
2
System Control and Computer Engineering Department, University of Petroșani, 332006 Petrosani, Romania
3
Management and Industrial Engineering Department, University of Petroșani, 332006 Petrosani, Romania
*
Author to whom correspondence should be addressed.
Appl. Sci. 2025, 15(16), 9032; https://doi.org/10.3390/app15169032
Submission received: 28 July 2025 / Revised: 7 August 2025 / Accepted: 14 August 2025 / Published: 15 August 2025
(This article belongs to the Special Issue eHealth Innovative Approaches and Applications: 2nd Edition)

Featured Application​

This work enables energy-efficient, real-time depression screening from speech, with potential applications in mobile health platforms, telemedicine, and low-resource clinical settings.

Abstract​

Depression remains a leading cause of global disability, yet scalable and objective diagnostic tools are still lacking. Speech has emerged as a promising non-invasive modality for automated depression detection, due to its strong correlation with emotional state and ease of acquisition. While convolutional neural networks (CNNs) have achieved state-of-the-art performance in this domain, their high computational demands limit deployment in low-resource or real-time settings. Spiking neural networks (SNNs), by contrast, offer energy-efficient, event-driven computation inspired by biological neurons, but they are difficult to train directly and often exhibit degraded performance on complex tasks. This study investigates whether CNNs trained on audio data from the clinically annotated DAIC-WOZ dataset can be effectively converted into SNNs while preserving diagnostic accuracy. We evaluate multiple conversion thresholds using the SpikingJelly framework and find that the 99.9% mode yields an SNN that matches the original CNN in both accuracy (82.5%) and macro F1 score (0.8254). Lower threshold settings offer increased sensitivity to depressive speech at the cost of overall accuracy, while naïve conversion strategies result in significant performance loss. These findings support the feasibility of CNN-to-SNN conversion for real-world mental health applications and underscore the importance of precise calibration in achieving clinically meaningful results.

Excerpt:

Future research will aim to address the limitations identified in this study, including the need for evaluation on neuromorphic hardware and the exploration of more complex network architectures. Nevertheless, the present work contributes meaningfully to the expanding body of literature focused on bridging the gap between high-performance deep learning models and biologically inspired, energy-efficient architectures. The findings presented herein offer empirical support for the viability of properly configured SNNs as effective alternatives to CNNs in sensitive applications such as automated depression detection from speech data.

Data Availability Statement​

Data available upon request at https://dcapswoz.ict.usc.edu/ (accessed on 3 May 2024). All codes involved in obtaining the results presented in this paper are available at https://github.com/gv1x2/ANN2SNN-SpikingJelly- (accessed on 20 July 2025).

Doz said:

View attachment 91697

Click to expand...

Hope he doesn’t get a call…(cough cough)

Esq.111 said:

Evening Mccabe84 & Fellow Chippers ,

Nice find .

Just saw this ........ Mentions Neuromorphic..... if only there was a company which dabbled in Neuromorphic Compute...

Pushing the Boundaries of Memory:What’s New with Weebit and AI | Weebit | A Quantum Leap In Data Storage

Blog article: ReRAM scales to advanced nodes, cuts power and cost, and lays the foundation for the next generation of AI, from in-memory-compute to neuromorphic systems.

www.weebit-nano.com

Regards,
Esq.

Click to expand...

26 min mark

Love the A.I. Cowboy's work ❤
Using Jetson and AKIDA to drop it's wattage from 120 to 9 is massive!
Also looking like a definite tie up with Weebit? ❤
Who's this Markus May guy again though? Rings a bell but..

I don't know about Sean meeting his 9 million dollar target, but it popped into my head this morning, maybe the perfect word to describe how he's been acting lately in podcasts etc..

Smug.
He's looking pretty smug.

I think at the very least, the quarterly is going to show a decent improvement over the last, which was around 1.4 million?
Still nowhere, where we need, but I think we may be finally about to witness some traction, with this Company

DingoBorat said:

I don't know about Sean meeting his 9 million dollar target, but it popped into my head this morning, maybe the perfect word to describe how he's been acting lately in podcasts etc..

Smug.
He's looking pretty smug.

I think at the very least, the quarterly is going to show a decent improvement over the last, which was around 1.4 million?
Still nowhere, where we need, but I think we may be finally about to witness some traction, with this Company

Click to expand...

My money is on megachips extending the license agreement with BRN for $9 million

Doz said:

View attachment 91697

Click to expand...

Thank you for posting that message Doz, I have only ever engaged once with Tony, and if you read how he starts out his
above reply to Gary, you can see that he is being guarded, which is what we all would expect......BUT goes on to share some
really positive news, including high praise for our engineering teams, like I say, he knows how good our technology is, so
taking the time to acknowledge his tech team is "extremely important".

That's the little bit of news bite that we would like to see more of from the top, nothing was given away that could be construed
as breaching company protocol, so I say, take a step back Sean and witness how a little bit of communication can go a long way
to satisfying your biggest supporter base.

More than baby steps, we appear to be taking rather wide steps forward now! AKD Tech

TECH said:

More than baby steps, we appear to be taking rather wide steps forward now! AKD Tech

Click to expand...

Maybe true, but our share price keeps on going

Doz said:

View attachment 91697

Click to expand...

Thank you but where is the link to it?

CHIPS said:

Thank you but where is the link to it?

Click to expand...