BRN Discussion Ongoing

[smoothsailing18]

FF.

I have updated my list of Brainchip Engagements as at 1 August, 2025. I use this list for the purpose of ongoing research to keep track of existing projects or to pick up any new projects. Needless to say the list is now quite long so if anyone wants use it for their research activities they are quite welcome:



1. FORD
2. VALEO
3. RENESAS
4. NASA
5. TATA Consulting Services
6.MEGACHIPS
7. MOSCHIP
8.SOCIONEXT
9.PROPHESEE
10. VVDN
11. TEKSUN
12. Ai LABS
13. NVISO now BeEMOTION
14. EMOTION3D
15. ARM
16. EDGE IMPULSE
17. INTEL
18. GLOBALFOUNDRIES
19. BLUERIDGE ENVISIONEERING
20. MERCEDES BENZ
21. ANT 61
22. QUANTUM VENTURA
23.INFORMATION SYSTEM LABORATORIES
24.INTELLISENSE SYSTEMS
25. CVEDIA
26. LORSER INDUSTRIES
27. SiFIVE
28. IPROSILICONE
29.SALESLINK
30. NUMEM
31. VORAGO
32. NANOSE
33. BIOTOME
34. OCULI
35. CIRCLE8 CLEAN TECHNOLOGIES
36. AVID GROUP
37. TATA ELXSI

38. NEUROBUS
39. EDGX
40. EUROPEAN SPACE AGENCY
41 UNIGEN
42. iniVation
43. SAHOMA CONTROLWARE
44. MAGIKEYE
45. MYWAI
46. INFINEON
47. ERICSSON
48. MICROCHIP
49. ONSEMI
50. IPSOLON RESEARCH

51. UBH - HELLAS

52. ACCENTURE

53. FRONTGRADE GAISLER

54. DELL Technologies

55. BOSTON DYNAMICS

56. AIRBUS

57. PARSONS CORPORATION

58. BASCOM HUNTER

59. ExeLANCE IT

60. US AIRFORCE RESEARCH LABORATORY

61. ONSOR

62. ANDES TECHNOLOGY

63. DEGIRUM

64. VEDYA

65. MULTICOREWARE

66. ARQUIMEA

67. LOCKHEED MARTIN
68. RTX - RAYTHEON & COLLINS

69. Nurjana Technologies

70. Chelpis Quantum GROUP

71. MiRLE GROUP

72. BOSCH

73. RENAULT

74. ST MICROELECTRONICS

75. University of Virginia

76. University of Oklahoma

77. Arizona State University
78. Carnegie Mellon University
79. Rochester Institute of Technology
80. Drexel University
81. Cornell Tech - founded by Cornell University & Technion - (Israel Institute of Technology and sponsor of Nanose)
82. University of Western Australia
83. Penn State University

My opinion only DYOR

Fact Finder

 

[Rskiff]

FF.

I have updated my list of Brainchip Engagements as at 1 August, 2025. I use this list for the purpose of ongoing research to keep track of existing projects or to pick up any new projects. Needless to say the list is now quite long so if anyone wants use it for their research activities they are quite welcome:



1. FORD
2. VALEO
3. RENESAS
4. NASA
5. TATA Consulting Services
6.MEGACHIPS
7. MOSCHIP
8.SOCIONEXT
9.PROPHESEE
10. VVDN
11. TEKSUN
12. Ai LABS
13. NVISO now BeEMOTION
14. EMOTION3D
15. ARM
16. EDGE IMPULSE
17. INTEL
18. GLOBALFOUNDRIES
19. BLUERIDGE ENVISIONEERING
20. MERCEDES BENZ
21. ANT 61
22. QUANTUM VENTURA
23.INFORMATION SYSTEM LABORATORIES
24.INTELLISENSE SYSTEMS
25. CVEDIA
26. LORSER INDUSTRIES
27. SiFIVE
28. IPROSILICONE
29.SALESLINK
30. NUMEM
31. VORAGO
32. NANOSE
33. BIOTOME
34. OCULI
35. CIRCLE8 CLEAN TECHNOLOGIES
36. AVID GROUP
37. TATA ELXSI

38. NEUROBUS
39. EDGX
40. EUROPEAN SPACE AGENCY
41 UNIGEN
42. iniVation
43. SAHOMA CONTROLWARE
44. MAGIKEYE
45. MYWAI
46. INFINEON
47. ERICSSON
48. MICROCHIP
49. ONSEMI
50. IPSOLON RESEARCH

51. UBH - HELLAS

52. ACCENTURE

53. FRONTGRADE GAISLER

54. DELL Technologies

55. BOSTON DYNAMICS

56. AIRBUS

57. PARSONS CORPORATION

58. BASCOM HUNTER

59. ExeLANCE IT

60. US AIRFORCE RESEARCH LABORATORY

61. ONSOR

62. ANDES TECHNOLOGY

63. DEGIRUM

64. VEDYA

65. MULTICOREWARE

66. ARQUIMEA

67. LOCKHEED MARTIN
68. RTX - RAYTHEON & COLLINS

69. Nurjana Technologies

70. Chelpis Quantum GROUP

71. MiRLE GROUP

72. BOSCH

73. RENAULT

74. ST MICROELECTRONICS

75. University of Virginia

76. University of Oklahoma

77. Arizona State University
78. Carnegie Mellon University
79. Rochester Institute of Technology
80. Drexel University
81. Cornell Tech - founded by Cornell University & Technion - (Israel Institute of Technology and sponsor of Nanose)
82. University of Western Australia
83. Penn State University

My opinion only DYOR

Fact Finder

Impressive. Now SUT$

 

[Fullmoonfever]

We know MB instigated a MOU or whatever with Uni of Waterloo and I was just reading an article about Ontario and how it's a bit of a Tech, AI & neuromorphic hub.

Something I found interesting about The Uni and their Applied Brain Spin off is how a lot of their focus is on SNN models, algos etc to apply to neuromorphic processors, not as much the processors themselves it seems. That could change(d), don't know.

But it got me thinking that maybe, just maybe, MB does def like Akida (still listed as a partner or customer or whatever) but they don't have the requisite experience in the SNN side to really achieve what they want Akida to do and hence the engagement with Waterloo.

Just a thought.

Couple of excerpts below from the article which I presume is part / all AI Gen overseen by this guy. Either way, the info on Waterloo I found interesting.

Joey Mazars​

Contributor & AI Expert

https://uk.linkedin.com/in/joey-mazars

How Neuromorphic Chips Are Reimagining AI by Thinking Like the Brain

Neuromorphic chips—hardware designed to mimic the structure and function of the human brain—are quietly engineering a seismic shift in how AI systems perceive, learn, and react.

autogpt.net

How Neuromorphic Chips Are Reimagining AI by Thinking Like the Brain​

Updated:July 31, 2025

Ontario’s Global Position in Neuromorphic AI​

Globally, neuromorphic hardware has reached notable milestones—Intel’s Loihi chips offer programmable SNNs, IBM’s TrueNorth pioneered the use of 1 million digital neurons, and BrainChip’s Akida platform integrates spiking intelligence into commercial edge devices. What makes Ontario stand out isn’t just research, but readiness. With the Vector Institute, University of Waterloo, silicon innovators like Untether AI and Tenstorrent, and real-world testing grounds like iGaming and robotics, Ontario offers a unique blend of capability and application. It could become North America’s leading launchpad for neuromorphic AI—where theory meets ROI across both factory floors and digital casinos.

Applied Brain Research (Waterloo)​

Applied Brain Research, based in Waterloo, stands as a pillar in the neuromorphic ecosystem. The company is best known for Nengo, a widely adopted software tool for building and simulating spiking neural networks. Nengo supports both research and commercial deployments, enabling models that run efficiently on neuromorphic substrates and low-power accelerators.

Applied Brain Research collaborates extensively with academic institutions and industrial partners to implement adaptive control and edge inference systems that benefit from brain-inspired computing. Their work transforms theory into deployed reality across sectors like robotics, automotive, and industrial automation.

Vector Institute and University of Waterloo​

Two of Ontario’s premier research hubs—Toronto’s Vector Institute and the University of Waterloo—are deeply involved in neuromorphic AI exploration. Vector researchers are advancing algorithms that make spiking networks trainable and efficient, particularly in on-chip learning scenarios. Meanwhile, University of Waterloo labs are publishing on adaptive spiking inference and energy-aware ML models.

Recent pilot programs have shown promise in domains such as traffic analysis and industrial automation, where real-time learning and micro-power consumption are game changers. The cross-pollination between academia and industry in Ontario accelerates the transition from research prototype to deployed solution.

 

[smoothsailing18]

ARM

 

[Diogenese]

FF.

I have updated my list of Brainchip Engagements as at 1 August, 2025. I use this list for the purpose of ongoing research to keep track of existing projects or to pick up any new projects. Needless to say the list is now quite long so if anyone wants use it for their research activities they are quite welcome:



1. FORD
2. VALEO
3. RENESAS
4. NASA
5. TATA Consulting Services
6.MEGACHIPS
7. MOSCHIP
8.SOCIONEXT
9.PROPHESEE
10. VVDN
11. TEKSUN
12. Ai LABS
13. NVISO now BeEMOTION
14. EMOTION3D
15. ARM
16. EDGE IMPULSE
17. INTEL
18. GLOBALFOUNDRIES
19. BLUERIDGE ENVISIONEERING
20. MERCEDES BENZ
21. ANT 61
22. QUANTUM VENTURA
23.INFORMATION SYSTEM LABORATORIES
24.INTELLISENSE SYSTEMS
25. CVEDIA
26. LORSER INDUSTRIES
27. SiFIVE
28. IPROSILICONE
29.SALESLINK
30. NUMEM
31. VORAGO
32. NANOSE
33. BIOTOME
34. OCULI
35. CIRCLE8 CLEAN TECHNOLOGIES
36. AVID GROUP
37. TATA ELXSI

38. NEUROBUS
39. EDGX
40. EUROPEAN SPACE AGENCY
41 UNIGEN
42. iniVation
43. SAHOMA CONTROLWARE
44. MAGIKEYE
45. MYWAI
46. INFINEON
47. ERICSSON
48. MICROCHIP
49. ONSEMI
50. IPSOLON RESEARCH

51. UBH - HELLAS

52. ACCENTURE

53. FRONTGRADE GAISLER

54. DELL Technologies

55. BOSTON DYNAMICS

56. AIRBUS

57. PARSONS CORPORATION

58. BASCOM HUNTER

59. ExeLANCE IT

60. US AIRFORCE RESEARCH LABORATORY

61. ONSOR

62. ANDES TECHNOLOGY

63. DEGIRUM

64. VEDYA

65. MULTICOREWARE

66. ARQUIMEA

67. LOCKHEED MARTIN
68. RTX - RAYTHEON & COLLINS

69. Nurjana Technologies

70. Chelpis Quantum GROUP

71. MiRLE GROUP

72. BOSCH

73. RENAULT

74. ST MICROELECTRONICS

75. University of Virginia

76. University of Oklahoma

77. Arizona State University
78. Carnegie Mellon University
79. Rochester Institute of Technology
80. Drexel University
81. Cornell Tech - founded by Cornell University & Technion - (Israel Institute of Technology and sponsor of Nanose)
82. University of Western Australia
83. Penn State University

My opinion only DYOR

Fact Finder

Parsons acquired Blacksignal last year:

https://www.parsons.com/2024/07/parsons-to-acquire-blacksignal-technologies/

Blaclsignal lists Blue Ridge as one of its companies:

https://www.blacksignal.tech/our-companies

Parsons lists a lot of AI applications:

https://www.parsons.com/ai/

Explore All Of Our AI Capabilities​

• Idea Conception
• Requirements Analysis
• AI Cloud/Edge Architectures
• AI HW/SW Design and Dev
• R&D Prototyping
• Large Scale Data Collection
• Data Preprocessing/Preparation

• Model Verification and Validation
• Model Security
• AI Solution Field Testing
• AI Solution Deployment
• Runtime Environment O&M
• AI Solution Risk Mitigation
• AI Ethics/Privacy/Anti-Bias

• AI Recruiting
• AI Staff Training
• AI Technology Surveillance
• AI Technology T&E (Paladin)
• Data Decommissioning
• AI System Decommissioning
• Synthetic Data Generation

The 2023 Blue Ridge/Brainchip announcement makes reference to Akida 2 technology:

https://www.edgeir.com/brainchip-bl...am-up-to-build-tactical-edge-devices-20230416

BrainChip, a provider of neuromorphic processors for edge AI on-chip processing, has teamed up with Blue Ridge Envisioneering. Blue Ridge Envisioneering will integrate BrainChip Akida processors into Blue Ridge’s designs for tactical edge devices used by defense and intelligence agencies.

The co-developed tactical edge devices will deploy in harsh, resource-constrained environments with limited power access and strict thermal requirements. The BrainChip Akida platform’s proficiency in functioning effectively in extreme situations and its ability to offer advanced artificial intelligence capabilities at the edge makes it suitable for military applications.
...

BrainChip recently announced the development of its latest Akida platform, which is designed for embedded edge AI applications and features 8-bit processing. The new system utilizes vision transformers and Temporal Event Based Neural Nets (TENN) spatial-temporal convolutions to improve performance and power efficiency.

The Brainchip/Blue Ridge partnership should be a neat fit, or rather a significant technical advancement, for Parsons.

 

[JDelekto]

Parsons acquired Blacksignal last year:

https://www.parsons.com/2024/07/parsons-to-acquire-blacksignal-technologies/

Blaclsignal lists Blue Ridge as one of its companies:

https://www.blacksignal.tech/our-companies

Parsons lists a lot of AI applications:

https://www.parsons.com/ai/

Explore All Of Our AI Capabilities​

• Idea Conception
• Requirements Analysis
• AI Cloud/Edge Architectures
• AI HW/SW Design and Dev
• R&D Prototyping
• Large Scale Data Collection
• Data Preprocessing/Preparation
• Model Verification and Validation
• Model Security
• AI Solution Field Testing
• AI Solution Deployment
• Runtime Environment O&M
• AI Solution Risk Mitigation
• AI Ethics/Privacy/Anti-Bias
• AI Recruiting
• AI Staff Training
• AI Technology Surveillance
• AI Technology T&E (Paladin)
• Data Decommissioning
• AI System Decommissioning
• Synthetic Data Generation

The 2023 Blue Ridge/Brainchip announcement makes reference to Akida 2 technology:

https://www.edgeir.com/brainchip-bl...am-up-to-build-tactical-edge-devices-20230416

BrainChip, a provider of neuromorphic processors for edge AI on-chip processing, has teamed up with Blue Ridge Envisioneering. Blue Ridge Envisioneering will integrate BrainChip Akida processors into Blue Ridge’s designs for tactical edge devices used by defense and intelligence agencies.

The co-developed tactical edge devices will deploy in harsh, resource-constrained environments with limited power access and strict thermal requirements. The BrainChip Akida platform’s proficiency in functioning effectively in extreme situations and its ability to offer advanced artificial intelligence capabilities at the edge makes it suitable for military applications.
...

BrainChip recently announced the development of its latest Akida platform, which is designed for embedded edge AI applications and features 8-bit processing. The new system utilizes vision transformers and Temporal Event Based Neural Nets (TENN) spatial-temporal convolutions to improve performance and power efficiency.

The Brainchip/Blue Ridge partnership should be a neat fit, or rather a significant technical advancement, for Parsons.

Wow. What a small world. As a high school student during the late 1980s, I worked as a student programmer for Parsons Engineering Science in Atlanta (they had acquired the small civil engineering firm in 1981).

Those were some of the most enjoyable projects I've worked on in my professional career, including hardware and software solutions for remote monitoring and control of tank pumps, remote calibration and measurement of rainwater levels, or simplifying calculations for treatment plant operations. Most of what they did was very innovative, integrating off-the-shelf components to create very unique automated solutions for their customers.

I can see how Parsons' subsidiaries and companies would benefit from using BrainChip's technology in their solutions, specifically for smart infrastructure projects. While they were a team that focused on Civil Engineering, they would usually contract out to Electrical Engineers for solutions that would make MacGyver proud.

While things are probably very different from what they were in the 80s, back then, I could envision someone using Akida (if it were available at that time) and cobbling together a custom solution for a client. By the way, most of their clients were large corporations and sometimes involved multi-year contracts regulated by government entities.

Thanks for that trip down memory lane!

 

[Stable Genius]

Vedya talking NPU’s:

Why NPUs shine for edge AI and real-time inference | Vedya Labs posted on the topic | LinkedIn

Do you like to have a control on your data and ensure AI models run instantly at the EDGE? 𝐓𝐡𝐚𝐭’𝐬 𝐰𝐡𝐞𝐫𝐞 𝐍𝐏𝐔𝐬 𝐜𝐨𝐦𝐞 𝐢𝐧. We talk a lot about GPUs when it comes to AI — and rightly so. But as AI shifts to the edge — phones, cameras, sensors — the game changes. NPUs (Neural Processing Units) are...

www.linkedin.com

 

[GazDix]

Hi again smooth.
None of it bothers me, actually sounds pretty good if turns out true. Maybe there was another collaboration or details of partnership announced that I haven't seen, but the red flag for me was the GPT comment "given the strategic nature of the partnership announced in late 2023" when we know the actual IP Licence announcement Dec 2020. The other concern is the formal announcement talks about Renesas required to pay for ongoing support after 2 yrs (paraphrasing), and as you know, engineering revenue (or any revenue) hasn't been great, so I'm a 50/50 on Renesas. Not being a downer by any means, just voicing my thoughts. Conversely, there is obviously lots going on behind the curtains, so I am hopeful, we will all see the tide to turn soonish????
I think GPT is great in many ways, but at the moment, I don't completely trust.

100%.
Experiment with other LLMs and they all cater and adjust to your biases. OpenAi have revealed they keep all your data.
For those who use ChatGPT often, try using the same question about BRN with a family member's account, maybe in a different house or different ISP address with a VPN on. See if you get the same results...

Under 16s good too for now. But don't dare look up history videos in Youtube.

 

[Getupthere]

CEO Tim Cook says Apple ready to open its wallet to catch up in AI — Reuters

Apple (AAPL.O) CEO Tim Cook signaled on Thursday the iPhone maker was ready to spend more to catch up to rivals in artificial intelligence by building more data centers or buying a larger player in the segment, a departure from a long practice of fiscal frugality.

apple.news

 

[Diogenese]

Wow. What a small world. As a high school student during the late 1980s, I worked as a student programmer for Parsons Engineering Science in Atlanta (they had acquired the small civil engineering firm in 1981).

Those were some of the most enjoyable projects I've worked on in my professional career, including hardware and software solutions for remote monitoring and control of tank pumps, remote calibration and measurement of rainwater levels, or simplifying calculations for treatment plant operations. Most of what they did was very innovative, integrating off-the-shelf components to create very unique automated solutions for their customers.

I can see how Parsons' subsidiaries and companies would benefit from using BrainChip's technology in their solutions, specifically for smart infrastructure projects. While they were a team that focused on Civil Engineering, they would usually contract out to Electrical Engineers for solutions that would make MacGyver proud.

While things are probably very different from what they were in the 80s, back then, I could envision someone using Akida (if it were available at that time) and cobbling together a custom solution for a client. By the way, most of their clients were large corporations and sometimes involved multi-year contracts regulated by government entities.

Thanks for that trip down memory lane!

Yes. I've worked on satellite comms, radio navigation, optical fibre projects, ... . This is the most exciting time in the electronics/computing field ever. Having been retired for several years, I'm more than a little envious, so I get my vicarious kicks from following Brainchip.

Maybe one of the new era jobs will be in model development and SNN configuration?

 

[BrainShit]

When googling for “Brainchip Akida” the other day, I stumbled across an Interesting Engineering article titled “Can an AI chip that mimics the brain beat the data deluge?”, but couldn’t access it at the time, because it was behind a paywall. (IMO it makes far more sense to spend the equivalent of that 1 US$ on half a dozen BRN shares right now…)


View attachment 89044


Today I discovered a link on the BrainChip website under Investor Relations that will take you to a PDF document created by Sarah Marker from Bospar (our new PR & Marketing Agency), which contains screenshots of that same article:

BrainChip’s Neuromorphic Energy-Efficient Computing

Explore BrainChip’s neuromorphic Akida™ processor, delivering ultra-low power, event-based AI for efficient edge computing applications.

investor.brainchip.com

“Brain-inspired chips slash AI power to micro-joules, but 8-bit limits and tooling gaps keep neuromorphic tech niche, for now.”
Learn more

…which links to…


She (or whoever else took those screenshots) unfortunately cut off some lines of text on each page, so in case we have any IE subscribers here on TSE, they may want to double-check whether we missed out on anything important.

View attachment 89045


View attachment 89027
View attachment 89028 View attachment 89029 View attachment 89030

View attachment 89031 View attachment 89032

View attachment 89034
This article - based on an interview with our CMO Steve Brightfield - should serve as a reality check to those shareholders who see virtually everything about BRN through rose-coloured glasses.

One challenge relating to the adoption of neuromorphic technology is that certain applications benefit more than others:

“It is important to note that the power savings are dramatic only in high sparsity scenarios such as static security cameras, sparse sensor data, and anomaly detection. ‘If the scene is a very active video feed with significant motion, resulting in low sparsity, you can expect considerably less savings,’ Brightfield cautioned.”


The last paragraph, where Steve Brightfield is quoted as saying “We’re still in the early stages of understanding how to best leverage neuromorphic architectures”,
might also be rather sobering for quite a few (especially long-term) shareholders. Our CMO then added: “The real test will be whether these efficiency gains can translate into broader applications as the technology matures.”

Now does that sound like our Chief Marketing Officer were talking about our company as being entirely de-risked and about imminent “generational wealth” for shareholders? Or is it rather an honest and realistic assessment of BrainChip’s current market position?

Here's a link to a "readable" version. This page also allows you to "save and display" websites that are behind a paywall / proxy ;-)

Source: https://archive.ph/89ax6

 

[gilti]

Hi Chippers,
This is just a big thank you to all the amazing sleuths out there continuing the ever expanding research into the Brainchip world.
As a technically illiterate man in the street it is such a bonus having these continuous streams of valuable information being presented.
So thanks again for your efforts. It is very much appreciated.
Gilti

 

[Fullmoonfever]

Was googling some search words as I do and this person's CV came up. No longer with BRN but was, back in mid 2023, for 3 mths as a ML Intern working on MB EQXX.

What I found interesting is that it appears MB maybe weren't just testing or playing with Akida for "Hey Mercedes" KWS but also driver distraction via sensors if I read correctly?

AND...over a year after the EQXX demo in early 2022 if this person's dates are correct.

Nice to see pretty damn good anomaly detection performace against a high end, though slightly older, GTX 1080 GPU.

HERE

 

[DingoBorat]

Hi Smoothsailing 18.
Im just not completely trusting of Chat GPT or others, though obviously very good tools in the right context. See the Renesas announcement from december 2020.
First Akida IP License Agreement
cheers

With those kind of "tools" when you ask a question, the way the answer is given, depends a lot on how the question is asked, especially when you are seeking "its" opinion.
It seems to me, they are programmed to be a bit like a dog, that is eager to please and be pat on the head.

If you ask something like "what is the likelihood" it "knows" you are seeking confirmation of something and will try to supply evidence to confirm what you want.

A more reliable way to use it, as I understand it, would be to ask the question in a negative manner, like "what is the argument against" such and such a thing and you will possibly get a more helpful response (if you look at the results in a positive way, if you know what I mean, or even better ask the question in both/multiple ways).

That's just the way I see it currently and don't give much weight to it's "opinion".
Although it can be useful and now it's getting better, I intend to use it more in the Future, I'm getting kind of sick of seeing ChatGPT etc analysis everywhere...

 

[Tothemoon24]

Neuromorphic AI Hardware: Brain-Inspired Chips Powering the Future

Neurons, not transistors, may decide the next decade of computing. Neuromorphic hardware—silicon that imitates the spikes, sparsity, and co-located memory of the brain—now delivers triple-digit gains in energy efficiency while unlocking real-time learning at the edge.

www.linkedin.com

 

[IloveLamp]

https://www.eetimes.com/hybrid-ai-models-blend-deep-learning-with-neuromorphic-ideas/

In this article from EE Times, BrainChip Scientific Advisory Board Member, Dr. | BrainChip

In this article from EE Times, BrainChip Scientific Advisory Board Member, Dr. Jason Eshraghian, explores how researchers are combining the strengths of deep learning with neuromorphic computing to create more efficient, brain-inspired AI systems. https://lnkd.in/gdDyYbBU #Neuromorphic...

www.linkedin.com

 

[stockduck]

Innatera claims world’s first mass-market neuromorphic microcontroller for the sensor edge​

Interviews | May 21, 2025

Innatera claims world’s first mass-market neuromorphic microcontroller for the sensor edge

Innatera in the Netherlands has launched the first commercial microcontroller using a neuromorphic architecture for sensor applications.

www.eenewseurope.com

"The Pulsar chip has a heterogenous architecture that combines analog and digital neuromorphic blocks with a traditional convolutional neural network accelerator and a RISC-V core. "
....

I really would like to know what kind of IP this "digital neuromorphic block" looks like?

The last 4c announcement gives me some hope.....

 

[FuzM]

https://www.eetimes.com/hybrid-ai-models-blend-deep-learning-with-neuromorphic-ideas/

In this article from EE Times, BrainChip Scientific Advisory Board Member, Dr. | BrainChip

In this article from EE Times, BrainChip Scientific Advisory Board Member, Dr. Jason Eshraghian, explores how researchers are combining the strengths of deep learning with neuromorphic computing to create more efficient, brain-inspired AI systems. https://lnkd.in/gdDyYbBU #Neuromorphic...

www.linkedin.com

View attachment 89251

Didn't realise that state space models was developed by Prof Chris Eliasmith from University of Waterloo.

The very same person that Mercedes has signed an MoU for research collaboration with.

Mercedes MoU

 

[TheDrooben]

https://www.eetimes.com/what-it-will-take-to-build-a-resilient-automotive-compute-ecosystem/

Interesting article by Dipti Vachani, SVP GM Automotive, Arm

Happy as Larry

 

[Diogenese]

Didn't realise that state space models was developed by Prof Chris Eliasmith from University of Waterloo.

The very same person that Mercedes has signed an MoU for research collaboration with.

Mercedes MoU

Hi Fuz,

Now there's a bit of morphic resonance.

It was only 3 days ago that I mentioned Eliasmith:

• #102,948

in relation to this:

TENNs-PLEIADES: Building Temporal Kernels with Orthogonal Polynomials​

Yan Ru Pei, Olivier Coenen

https://arxiv.org/html/2405.12179v3

...

The seminal work proposing a memory encoding using orthogonal Legendre polynomials in a recurrent state-space model is the Legendre Memory Unit (LMU) [33], where Legendre polynomials (a special case of Jacobi polynomials) are used. The HiPPO formalism [11] then generalized this to other orthogonal functions including Chebyshev polynomials, Laguerre polynomials, and Fourier modes. Later, this sparked a cornucopia of works interfacing with deep state space models including S4 [12], H3 [2], and Mamba [10], achieving impressive results on a wide range of tasks from audio generation to language modeling. There are several common themes among these networks that PLEIADES differ from. First, these models typically only interface with 1D temporal data, and usually try to flatten high dimensional data into 1D data before processing [12, 37], with some exceptions [21]. Second, instead of explicitly performing finite-window temporal convolutions, a running approximation of the effects of such convolutions are performed, essentially yielding a system with infinite impulse responses where the effective polynomial structures are distorted [31, 11]. And in the more recent works, the polynomial structures are tenuously used only for initialization, but then made fully trainable. Finally, these networks mostly use an underlying depthwise structure [14] for long convolutions, which may limit the network capacity, albeit reducing the compute requirement of the network.
[33]↑Aaron Voelker, Ivana Kajić, and Chris Eliasmith.Legendre Memory Units: Continuous-time representation in recurrent neural networks.Advances in neural information processing systems, 32, 2019. [Uni of Waterloo]


Our Pleiades paper differentiates our SSM from the Legendre polynomial which Eliasmith proposed.