BRN Discussion Ongoing

Frangipani

Top 20



F3C1FAC3-9A0C-495D-8012-8FCAE925CCBF.jpeg






489689FC-09E3-4D88-B2F3-76EB5C730D4C.jpeg
7676346A-53D1-41F6-AD9A-446FF563E6ED.jpeg
82266971-03B1-4816-A060-28C2CE3B9A79.jpeg
 
  • Like
  • Love
  • Fire
Reactions: 23 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
Rudy Pei, who moved from BrainChip to NVIDIA earlier this year, has just posted a new paper and a video on event-based data/processing, and the work he did for the PLEIADES network.

Let's hope all of his new colleagues, including Jensen Huang (President and CEO of NVIDIA), take a peek!

Some interesting people have "liked" Rudy's LinkedIn post (see below)!




Rudy Pei am.png





Screenshot 2025-10-29 at 8.43.53 am.png





Screenshot 2025-10-29 at 8.36.14 am.png

Screenshot 2025-10-29 at 8.36.46 am.png





Screenshot 2025-10-29 at 8.35.47 am.png



Screenshot 2025-10-29 at 8.51.43 am.png



Screenshot 2025-10-29 at 8.39.33 am.png

Screenshot 2025-10-29 at 8.36.30 am.png




Screenshot 2025-10-29 at 8.42.31 am.png









 

Attachments

  • Screenshot 2025-10-29 at 8.36.30 am.png
    Screenshot 2025-10-29 at 8.36.30 am.png
    635.1 KB · Views: 29
Last edited:
  • Like
  • Fire
  • Love
Reactions: 22 users

Tothemoon24

Top 20
Rudy Pei, who moved from BrainChip to NVIDIA earlier this year, has just released a paper and a video on event-based data/processing, and the work he did for the PLEIADES network.

Let's hope all of his new colleagues, including Jensen Huang (President and CEO of NVIDIA), take a peek!

Some interesting people have liked Rudy's post (see below)!




View attachment 92494




View attachment 92510




View attachment 92495
View attachment 92501




View attachment 92496

View attachment 92503

View attachment 92504
View attachment 92505



View attachment 92508









Hi Bravo , I was just looking at that post myself will be interesting to see if Rudy responds to the below question which has been recently asked


IMG_1681.jpeg
 
  • Like
  • Fire
  • Thinking
Reactions: 17 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
Hi Bravo , I was just looking at that post myself will be interesting to see if Rudy responds to the below question which has been recently asked


View attachment 92511


Agree @Tothemoon24, it'll be great if Rudy responds, but even if he doesn't I suppose Philip has ensured the connection is evident.

Another interesting "like" from an AI Engineer (Ahabb Sheraz) at Dreambig Semiconductors, which must mean the post touched something he's actively thinking about.



Screenshot 2025-10-29 at 8.58.34 am.png


Screenshot 2025-10-29 at 8.57.47 am.png







I remembered seeing that company name, Dreambig, yesterday when I was doing some research on Athos Silone's Polaris.

Dreambig + Athos are building a central ADAS chip; Akida-like front-ends could be complementary at the edge (cabin, always-on sentry).

Of course, Ahabb's liking of Rudy's post may have nothing to do with DreamBig/Athos’ chiplet vision for Polaris but it's got me pondering over how something like PLEIADES might be woven into it.

Screenshot 2025-10-29 at 8.59.00 am.png
 
  • Like
  • Fire
  • Love
Reactions: 14 users

7für7

Top 20
Sorry if already posted… regarding NVIDIA

 
  • Fire
  • Love
Reactions: 4 users
FF

 
  • Like
  • Fire
  • Love
Reactions: 5 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
Looking a little further into Deniz Cakan, PhD, (Principal Transformational Physicist at Northrop Grumman). Deniz was one of the people who liked Rudy Pei's PLEIADES post.

Obviously Deniz has an interest in neuromorphic computing, given he co-authored the paper below, which discusses neuromorphic computing.

At Northrop, Deniz leads the quantum R&D at the intersection of AIML, materials science, and device physics. His team’s work focuses on designing intelligent systems that accelerate quantum R&D.

In short, his work at the intersection of quantum + AI looks like it lines up very neatly with what BrainChip’s PLEIADES aims to deliver: intelligent, low-latency systems that can sense, decide, and adapt at the edge.

I asked ChatGPT if BrainChip's PLEIADES could help move quantum computing closer to commercialization. See the answer below if interested or scroll past it if you're not.


Screenshot 2025-10-29 at 10.22.13 am.png






Screenshot 2025-10-29 at 10.04.39 am.png




I said:​


Could BrainChip's PLEIADES help bring quantum computing closer to commercialisation?

ChatGPT said:​

That’s a really sharp question — and yes, in a meaningful way, it could help move quantum computing closer to commercialization, though not by “solving” quantum computing outright.

Here’s the nuance:

🔹 The bottleneck today​

The hardest part of scaling quantum systems isn’t just building qubits — it’s stabilising and controlling them in real time.
  • Qubits drift, decohere, or get noisy in unpredictable ways.
  • Each quantum chip, material, or cryostat behaves slightly differently.
  • The tuning and calibration process is slow, data-heavy, and often manual — scientists spend hours just “babysitting” devices to keep them coherent.

🔹 Where PLEIADES-style neuromorphic AI fits​

A system like BrainChip’s PLEIADES (event-driven, polynomial-based temporal processing) could serve as an intelligent control layer that:
  • Detects state changes or drifts as events, in microseconds.
  • Adjusts voltages, fields, or bias currents autonomously, without waiting for batch analysis.
  • Learns the behaviour of each individual device locally, using on-chip adaptation.
  • Runs at low power and deterministic latency, which is vital for cryogenic or shielded setups.
Essentially, it brings real-time intelligence to the hardware level, closing the loop between sensing and actuation — something cloud-scale AI can’t do because of latency and bandwidth limits.

🔹 The commercial implication​

If labs can stabilise and tune qubits faster and keep them stable longer, they can:
  • Improve yield (more usable qubits per wafer),
  • Shorten test and calibration time, and
  • Reduce operational cost and complexity.
That’s the difference between a research-only system and a manufacturable, scalable product line.

✅ In short​

PLEIADES-type neuromorphic controllers won’t build the qubits — but they could make quantum systems smarter, more autonomous, and commercially viable by solving the control and reliability problem that’s holding the industry back.
 
  • Like
  • Fire
  • Love
Reactions: 14 users

7für7

Top 20
Nobody can seriously tell me that with all these partnerships, customers, and connections in every direction, BrainChip can’t generate at least three or four strong commercial contracts. Come on. That would be like filling out a lottery ticket with every possible combination except six numbers… and then somehow the six numbers you *didn’t* tick are the winning ones. Statistically ridiculous.

For me this investment is basically a jackpot that’s already won .. we’re just waiting for payout. That’s my view.

Doesn’t mean I’m not angry. The level of obvious manipulation and shorting we’re seeing is borderline criminal. And yeah, I’ll keep calling it out. That’s my southern European blood .. I don’t sit quiet and pretend it’s normal.

No financial advice DYFOFR
 
  • Like
  • Fire
Reactions: 11 users

7für7

Top 20
IMG_7262.jpeg


It’s not a no! 😑☝️
 
  • Like
  • Fire
  • Thinking
Reactions: 9 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
Interesting to note that Brian Anderson also liked Rudy Pei’s post about BrainChip’s PLEIADES.

Brian leads Project Phasor, a community initiative aimed at fixing the “missing middle” in neuromorphics, building the common tooling needed to train, simulate, and deploy models across different neuromorphic chips.

Digging further into what Brain's been up to, I noticed he commented 5 days ago on Meta’s acquisition of Rivos, suggesting it could "work out really well" for the neuromorphic community. According to press reports, Meta’s move is partly about reducing its dependence on NVIDIA GPUs for AI workloads.

Rivos itself isn’t a neuromorphic company, it's a Silicon Valley startup building data-center AI accelerators around RISC-V CPUs + Rivos-designed GPGPU cores, but as Brian pointed out, it could become the training platform for neuromorphics - a bridge between large-scale AI accelerators and edge neuromorphic chips like Akida. That fits perfectly with the virtualization and compiler-pipeline work Brian has been developing through Project Phasor.



Screenshot 2025-10-29 at 11.29.32 am.png









Screenshot 2025-10-29 at 11.15.59 am.png








Screenshot 2025-10-29 at 11.10.55 am.png







Screenshot 2025-10-29 at 11.16.15 am.png
 
Last edited:
  • Love
  • Fire
  • Like
Reactions: 6 users

Frangipani

Top 20
Here comes the BrainChip LinkedIn post about the multi-year strategic agreement with Parsons - better late than never… 😉


BBD4D7D0-327B-438C-829F-38C136452256.jpeg
 
  • Like
  • Fire
  • Love
Reactions: 18 users

TheDon

Regular
Here comes the BrainChip LinkedIn post about the multi-year strategic agreement with Parsons - better late than never… 😉


View attachment 92528
I think they forgot to put the multi billion dollar agreement. Fix it BrainChip! ;)

TheDon
 
  • Haha
  • Fire
  • Like
Reactions: 8 users

Frangipani

Top 20

View attachment 92479 View attachment 92480



View attachment 92481 View attachment 92482


8C064F95-8A49-4718-9B0F-31D7940F5C8F.jpeg
 
  • Like
  • Fire
Reactions: 7 users

Tothemoon24

Top 20
Agree @Tothemoon24, it'll be great if Rudy responds, but even if he doesn't I suppose Philip has ensured the connection is evident.

Another interesting "like" from an AI Engineer (Ahabb Sheraz) at Dreambig Semiconductors, which must mean the post touched something he's actively thinking about.



View attachment 92516

View attachment 92517






I remembered seeing that company name, Dreambig, yesterday when I was doing some research on Athos Silone's Polaris.

Dreambig + Athos are building a central ADAS chip; Akida-like front-ends could be complementary at the edge (cabin, always-on sentry).

Of course, Ahabb's liking of Rudy's post may have nothing to do with DreamBig/Athos’ chiplet vision for Polaris but it's got me pondering over how something like PLEIADES might be woven into it.

View attachment 92518
Well he didn’t say no 🙌

IMG_1682.jpeg
 
  • Like
  • Thinking
  • Fire
Reactions: 14 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
Well he didn’t say no 🙌

View attachment 92531


It's a bit like... 🤭



a630f1f6035cca9be9be84687c16079d.gif


But on a serious note, it sounds like the kind of carefully worded response you’d expect from someone who has contractual or confidentiality obligations to meet.

Even mentioning whether something is or isn't BrainChip IP might breach strict non-disclosure terms.

At least he didn't say something like “You’ll have to reach out to Dunkin’ Donuts directly.”

So, the way I look at it is, it's like acknowledging that the material does relate to BrainChip’s proprietary work without actually saying so.
 
  • Like
  • Fire
  • Thinking
Reactions: 9 users

7für7

Top 20
NVIDIA share price would explode and our would drop as usual on massive news if they would sign any contract … LoL 😂 (Aside from that “Rudy” topic) fun fact at its best
 

7für7

Top 20
New partner in… according to our partner website




IMG_7263.jpeg
 
Last edited:
  • Fire
  • Like
  • Wow
Reactions: 10 users

7für7

Top 20
Come on cockroaches… you have still one hour to dump it to 18-18.5… come on

Edit 2 hours… 🤷🏻‍♂️
 
Last edited:
  • Haha
Reactions: 1 users

Diogenese

Top 20
It's a bit like... 🤭



View attachment 92532

But on a serious note, it sounds like the kind of carefully worded response you’d expect from someone who has contractual or confidentiality obligations to meet.

Even mentioning whether something is or isn't BrainChip IP might breach strict non-disclosure terms.

At least he didn't say something like “You’ll have to reach out to Dunkin’ Donuts directly.”

So, the way I look at it is, it's like acknowledging that the material does relate to BrainChip’s proprietary work without actually saying so.

In the associated youjube video, Rudy refers to Centaurus:

https://arxiv.org/abs/2501.13230#:~:text=We introduce Centaurus, a class of networks composed,can be treated as tensor contractions during training.

[Submitted on 22 Jan 2025 (v1), last revised 9 Apr 2025 (this version, v2)]

Let SSMs be ConvNets: State-space Modeling with Optimal Tensor Contractions​


We introduce Centaurus, a class of networks composed of generalized state-space model (SSM) blocks, where the SSM operations can be treated as tensor contractions during training. The optimal order of tensor contractions can then be systematically determined for every SSM block to maximize training efficiency. This allows more flexibility in designing SSM blocks beyond the depthwise-separable configuration commonly implemented. The new design choices will take inspiration from classical convolutional blocks including group convolutions, full convolutions, and bottleneck blocks. We architect the Centaurus network with a mixture of these blocks, to balance between network size and performance, as well as memory and computational efficiency during both training and inference. We show that this heterogeneous network design outperforms its homogeneous counterparts in raw audio processing tasks including keyword spotting, speech denoising, and automatic speech recognition (ASR). For ASR, Centaurus is the first network with competitive performance that can be made fully state-space based, without using any nonlinear recurrence (LSTMs), explicit convolutions (CNNs), or (surrogate) attention mechanism. The source code is available as supplementary material on this https URL

Rudy's linkedin shows he started with Nvidia in February 2025.

From Rudy's arvix bibliography:

https://arxiv.org/search/cs?searchtype=author&query=Pei,+Y+R
arXiv:2405.12179 [pdf, ps, other]

cs.LG cs.AI
PLEIADES: Building Temporal Kernels with Orthogonal Polynomials
Authors: Yan Ru Pei, Olivier Coenen
Abstract: We introduce a class of neural networks named PLEIADES (PoLynomial Expansion In Adaptive Distributed Event-based Systems), which contains temporal convolution kernels generated from orthogonal polynomial basis functions. We focus on interfacing these networks with event-based data to perform online spatiotemporal classification and detection with low latency. By virtue of using structured temporal… ▽ More
Submitted 23 October, 2025; v1 submitted 20 May, 2024; originally announced May 2024.
Comments: 14 pages, 3 figures
Journal ref: 39th Conference on Neural Information Processing Systems
(NeurIPS 2025)

The original paper was submitted in May 2024. Version 2 was submitted in October 2025, so what happened up to January 2025 is clearly BRNs property. What happened after that depends on Rudy's employment contract with BRN and US law. Rudy's lab notebooks (and those of his co-workers) may shed some light.

https://natlawreview.com/article/ownership-inventions-former-employers-lessons-after-bio-rad-v-itc

Bio-Rad’s cited cases were deemed distinguishable. According to the court, none held that a “significant contribution” to a post-employment invention could fall within an assignment provision limited to the term of employment. (The court indicated a different outcome could result if a later-claimed invention was conceived during the period of employment.) As an example of assignment obligations not limited to the employment period, the court pointed to a case (Stanford) where the relevant provisions expressly extended to inventions conceived “as a consequence” of employment.
 
  • Like
  • Love
  • Wow
Reactions: 6 users

walderamaa

Emerged
Kommt schon, Kakerlaken … ihr habt noch eine Stunde, um es auf 18-18,5 zu bringen … kommt schon

2 Stunden bearbeiten…🤷🏻‍♂️
Du nervst
 
Top Bottom