Good Morning Chippers ,
Stolen from the other channel...
From Doz , from a few days ago , Cheers.
Regards,
Esq
I clicked the link and
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
I clicked the link and
Investor Podcast Ep 8: CES & 2025's Future
Investor Podcast Ep 8: CES & 2025's Future Join CEO Sean Hehir and Director of Global Investor Relations Tony Dawe for BrainChip’s latest investor update. They’ll discuss the evolving AI landscape, recent customer agreements, and the company's strategic direction for 2025. Key topics...
brainchip.com
Bravo
If ARM was an arm, BRN would be its biceps💪!
Brain-inspired Computing Is Ready for the Big Time
Neuromorphic pioneer Steve Furber says it's just awaiting a killer app
Edd Gent10 hours ago
6 min read
Edd Gent is a Contributing Editor for IEEE Spectrum.
Steve Temple (left) holding a SpiNNaker chip with Steve Furber (right) in front of a labelled plot of the chip.
Steve Furber
neuromorphic computing spinnaker artificial intelligence neural networks spiking neural networks
Efforts to build brain-inspired computer hardware have been underway for decades, but the field has yet to have its breakout moment. Now, leading researchers say the time is ripe to start building the first large-scale neuromorphic devices that can solve practical problems.
The neural networks that have powered recent progress in artificial intelligence are loosely inspired by the brain, demonstrating the potential of technology that takes its cues biology. But the similarities are only skin deep and the algorithms and hardware behind today’s AI operate in fundamentally different ways to biological neurons.
Neuromorphic engineers hope that by designing technology that more faithfully replicates the way the brain works, we will be able to mimic both its incredible computing power and its energy efficiency. Central to this approach is the use of spiking neural networks, in which computational neurons mimic their biological cousins by communicating using spikes of activity, rather than the numerical values used in conventional neural networks. But despite decades of research and increasing interest from the private sector, most demonstrations remain small scale and the technology has yet to have a commercial breakout.
In a paper published in Nature in January, some of the field’s leading researchers argue this could soon change. Neuromorphic computing has matured from academic prototypes to production-ready devices capable of tackling real-world challenges, they argue, and is now ready to make the leap to large-scale systems. IEEE Spectrum spoke to one of the paper’s authors, Steve Furber, the principal designer of the ARM microprocessor—the technology that now powers most cellphones—and the creator of the SpiNNaker neuromorphic computer architecture.
In the paper you say that neuromorphic computing is at a critical juncture. What do you mean by that?
Steve Furber: We’ve demonstrated that the technology is there to support spiking neural networks at pretty much arbitrary scale and there are useful things that can be done with them. The criticality of the current moment is that we really need some demonstration of a killer app.
The SpiNNaker project started 20 years ago with a focus on contributing to brain science, and neuromorphics is an obvious technology if you want to build models of brain cell function. But over the last 20 years, the focus has moved to engineering applications. And to really take off in the engineering space, we need some demonstrations of neuromorphic advantage.
In parallel over those 20 years, there’s been an explosion in mainstream AI based on a rather different sort of neural network. And that’s been very impressive and obviously had huge impacts, but it’s beginning to hit some serious problems, particularly in the energy requirements of large language models (LLMs). And there’s now an expectation that neuromorphic approaches may have something to contribute, by significantly reducing those unsustainable energy demands.
The SpiNNaker team assembles a million-core neuromorphic system.SpiNNaker
We are close to having neuromorphic systems at a scale sufficient to support LLMs in neuromorphic form. I think there are lots of significant application developments at the smaller end of the spectrum too. Particularly close to sensors, where using something like an event-based image sensor with a neuromorphic processing system could give a very low energy vision system that could be applied in areas such as security and automotive and so on.
When you talk about achieving a large-scale neuromorphic computer, how would that compare to systems that already exist?
Furber: There are lots of examples out there already like the large Intel Loihi 2 system, Hala Point. That’s a very dense, large-scale system. The SpiNNaker 1 machine that we’ve been running a service on [at the University of Manchester, UK] since 2016 had half a million ARM cores in the system, expanding to a million in 2018. That’s reasonably large scale. Our collaborators on SpiNNaker 2 [SpiNNcloud Systems, based in Dresden, Germany] are beginning to market systems at the 5 million core level, and they will be able to run quite substantial LLMs.
Now, how much those will need to evolve for neuromorphic platforms is a question yet to be answered. They can be translated in a fairly simplistic way to get them running, but that simple translation won’t necessarily get the best energy performance.
So is the hardware not really the issue, it’s working out how to efficiently build something on top of it?
Furber: Yes, I think the last 20 years has seen proof-of-concept hardware systems emerge at the scales required. It’s working out how to use them to their best advantage that is the gap. And some of that is simply replicating the efficient and useful software stacks that have been developed for GPU-based machine learning.
It is possible to build applications on neuromorphic hardware, but it’s still unreasonably difficult. The biggest missing components are the high-level software design tools along the lines of TensorFlow and PyTorch that make it straightforward to build large models without having to go down to the level of describing every neuron in detail.
There’s quite a diversity of different neuromorphic technologies, which can sometimes make it hard to translate findings between different groups. How can you break down those silos?
Furber: Although the hardware implementation is often quite different, the next level up there is quite a lot in common. All neuromorphic platforms use spiking neurons and the neurons themselves are similar. You have a diversity of details at the lower levels, but that can be bridged by implementing a layer of software that matches those lower level hardware differences to higher level commonalities.
We’ve made some progress on that front, because within the EU’s Human Brain Project, we have a group that’s been developing the PyNN language. It is supported by both SpiNNaker, which is a many core neuromorphic system, and the University of Heidelberg’s BrainScaleS system, which is an analog neural model.
But it is the case that a lot of neuromorphic systems are developed in a lab and used only by other people within that lab. And therefore they don’t contribute to the drive towards commonality. Intel has been trying to contribute through building the Lava software infrastructure on their Loihi system and encouraging others to participate. So there are moves in that direction but it’s far from complete.
A member of the SpiNNaker team checks on the company’s million-core machine.Steve Furber
Opinions differ on how biologically plausible neuromorphic technology needs to be. Does the field need to develop some consensus here?
Furber: I think the diversity of the hardware platforms and of the neuron models that are used is a strength in the research domain. Diversity is a mechanism for exploring the space and giving you the best chance of finding the best answers for developing serious, large-scale applications. But once you do, yes, I think you need to reduce the diversity and focus more on commonality. So if neuromorphic is about to make the transition from a largely research-driven territory to a largely application-driven territory, then we’d expect to see that kind of thing changing.
If the field wants to achieve scale will it have to sacrifice a bit of biological plausibility?
Furber: There is a trade-off between biological fidelity and engineering controllability. Replicating the extremely simple neural models that are used in LLMs does not require a lot of biological fidelity. Now, it’s arguable that if you could incorporate a bit more of the biological detail and functionality, you could reduce the number of neurons required for those models by a significant factor. If that’s true, then it may well be worth ultimately incorporating those more complex models. But it is still big research problem to prove that this is the case.
In recent years there’s been a lot of excitement about memristors—memory devices that mimic some of the functionality of neurons. Is that changing the way people are approaching neuromorphic computing?
Furber: I do think that the technologies that are being developed have the potential to be transformative in terms of improving hardware efficiency at the very low levels. But when I look at the UK neuromorphic research landscape, a very significant proportion of it is focused on novel device technologies. And arguably, there’s a bit too much focus on that, because the systems problems are the same across the board.
Unless we can make progress on the systems level issues it doesn’t really matter what the underpinning technology is, and we already have platforms that will support progress on the systems level issues.
The paper suggest that the time is ripe for large-scale neuromorphic computing. What has changed in recent years that makes you positive about this, or is it more a call to arms?
Furber: It’s a bit in-between. There is evidence it’s happening, there are a number of interesting startups in the neuromorphic space who are managing to survive. So that’s evidence that people with significant available funds are beginning to be prepared to spend on neuromorphic technology. There’s a belief in the wider community that neuromorphic’s time is coming. And of course, the huge problems facing mainstream machine learning on the energy front, that is a problem which is desperate for a solution. Once there’s a convincing demonstration that neuromorphics can change the equation, then I think we’ll see things beginning to turn.
Brain-inspired Computing Is Ready for the Big Time
Neuromorphic pioneer Steve Furber says it's just awaiting a killer app
spectrum.ieee.org
Bravo
If ARM was an arm, BRN would be its biceps💪!
Arm secures Meta as first customer for ambitious new chip project, FT reports
By ReutersFebruary 14, 20256:12 AM GMT+11Updated 5 hours ago
Feb 13 (Reuters) - Arm Holdings plans to launch its own chip this year after securing Meta Platforms as one of its first customers, in a major shift to the chip tech provider's model of licensing its blueprints to other companies, the Financial Times reported on Thursday, sending its U.S.-listed shares up about 5%.
This move would put Arm in direct competition with some of its largest customers, including Nvidia, which build their own chips on top of Arm's architecture.
Arm has been somewhat exempt from the AI-linked growth spurts enjoyed by chipmakers because it makes money from AI indirectly by steadily raising licensing fees for its technology and charging royalties for each chip other companies sell.
Rene Haas, Arm CEO, will unveil the first chip that it has made in-house as early as this summer, the FT reported, citing people familiar with the company's plans.
Arm's chip is expected to be a central processing unit (CPU) for servers in large data centres and is built on a base which can then be customised for clients including Meta, the report said.
Production will be outsourced to a manufacturer such as TSMC according to the report.
SoftBank and Meta did not immediately respond to Reuters' requests for comment while Arm declined to comment.
Masayoshi Son, founder of Arm's majority owner SoftBank, has placed Arm at the centre of his push to expand AI infrastructure, with the launch of Arm's own chip considered a step in his larger plans to move into AI chip production, FT reported.
SoftBank is also nearing a $6.5 billion buyout of Oracle-backed chip designer Ampere, a deal considered essential to Arm's own chipmaking project, according to the report.
Bravo
If ARM was an arm, BRN would be its biceps💪!
This article was published online 8 hours ago. Beneath the article I've included a screen-shot from a Raytheon Linkedin post which describes the LTAMDS and it which also includes some comments asking whether BrainChip is involved.Raytheon Demos Advanced Radar System
by
Miles JamisonFebruary 13, 2025, 11:19 amRaytheon, an RTX business, has completed a demonstration of its Lower Tier Air and Missile Defense Sensor, or LTAMDS, at the White Sands Missile Range in New Mexico.
Showcasing LTAMDS Radar Capabilities
RTX said Wednesday the live-fire test showcased the ability of the 360-degree, full-sector radar system to detect and track the Patriot Advanced Capability-2 Guidance Enhanced Missile-T, or PAC-2 GEM-T, missile. The LTAMDS also managed to guide the high-speed cruise missile to intercept a simulated threat.The recent demonstration, part of a comprehensive U.S. Army test program, exhibited the capabilities of the combat-proven effector when guided by the LTAMDS. With the latest live-fire event, the advanced radar system is one step closer to field deployment, particularly its potential integration into the U.S. Army’s Integrated Air and Missile Defense system.
Over a dozen nations have expressed interest and requested information about the LTAMDS. Poland is the first international customer to enter into an agreement to purchase the LTAMDS radar. Raytheon landed the $2.1 billion contract to produce the radar for Poland and the U.S. Army.
“This most recent test represents a significant milestone for both Raytheon and the Army, demonstrating the combat-proven PAC-2 GEM-T interceptor with the transformational LTAMDS radar,” stated Tom Laliberty, president of land and air defense systems at Raytheon. “LTAMDS will further enhance Patriot’s unmatched air defense capabilities, helping the Army and customers around the globe defend against increasingly complex threats.”
Raytheon Demos Advanced Radar System - ExecutiveBiz
Explore the details of the recent live-fire test of Raytheon's advanced radar system.
Thanks for sharing @Rskiff, I like his closing comments, "I believe that significantly more smart devices with very new technology will be around us very soon".
Happy Valentines day everyone, sending love to you all ❤
To my mind, Sean's anecdote about walking the floor of CES25 and talking to exhibitors reinforces the hypothesis that an IP only strategy excludes the small business market sector. They are using COTS NNs, but would prefer to use custom NNs if they could aford it. The thing about Akida is that it is customizabel in both size (number of nodes/NPUs) and application (model development, particularly with our cooperation with Edge Impulse).
Leaving the TENNs algorithm product aside for the moment, a major licence with large customer with near-term production objectives would secure the share price, and clearly, BRN cannot afford to be a chip maker at this point in time. I think this illustrates the need to get a chip maker tied up (as in licensed, not bondage) to produce the various flavours of Akida/TENNs COTS chips to capture the smaller users who cannot afford an IP licence.
A major problem is that most major chip designers have their own in-house AI. Even our early licencee Renesas has its own DRP-AI which it has refined in the interim with N:M coding which may have taken inspiration from Akida. Qualcomm, ARM and Intel each have their own AI implementations.
We seem to have gone cool on the original Socionext/TSMC association, at least for the time being.
So, of our known associates, Global Foundries could be a candidate to produce the Akida/TENNs COTS range, and I wouldn't write off Megachips yet.
Last edited:
Good Afternoon Chippers ,
Once again iv taken the liberty of stealing this from the other forum.
Cheers StockHound , these numbers also correspond with the last update on price compiled by RockerRothsGettyFellerChild LLC which stood at a conservative $7.117 Au on 9th July 2024.
.
NOTE , The attached has been generated by ChatGPT.
Dam it , still can't cut and past the whole picture.
The crapper,
Poster , StockHound81
Today @ 11:00
If a savvy Chipper would be so kind as to retrieve this from the crapper site , and post for all to have a ponder.
Thankyou in advance.
#Well i certainly made mess of that , sorry all
Regards,
Esq.
Once again iv taken the liberty of stealing this from the other forum.
Cheers StockHound , these numbers also correspond with the last update on price compiled by RockerRothsGettyFellerChild LLC which stood at a conservative $7.117 Au on 9th July 2024.
.NOTE , The attached has been generated by ChatGPT.
The crapper,
Poster , StockHound81
Today @ 11:00
If a savvy Chipper would be so kind as to retrieve this from the crapper site , and post for all to have a ponder.
Thankyou in advance.
#Well i certainly made mess of that , sorry all
Regards,
Esq.
Last edited:
Mea culpa
prəmɪskjuəs
Nah, my middle name’s Messy, Esq, and I don’t even put vanilla beans in me beer.Good Afternoon Chippers ,
Once again iv taken the liberty of stealing this from the other forum.
Cheers StockHound , these numbers also correspond with the last update on price compiled by RockerRothsGettyFellerChild LLC which stood at a conservative $7.117 Au on 9th July 2024.
NOTE , The attached has been generated by ChatGPT.
View attachment 77588
View attachment 77590
View attachment 77589
View attachment 77591
View attachment 77592 View attachment 77593 View attachment 77594 View attachment 77595 Dam it , still can't cut and past the whole picture.
The crapper,
Poster , StockHound81
Today @ 11:00
If a savvy Chipper would be so kind as to retrieve this from the crapper site , and post for all to have a ponder.
Thankyou in advance.
#Well i certainly made mess of that , sorry all
Regards,
Esq.
MMc.
Taproot
Regular
Spiking Neural Network Modeler (SpiNNMo) Phase II | SBIR.gov
legacy.www.sbir.gov
In order to meet the Navy’s need for a spiking neural network testing platform, ChromoLogic proposes to develop a Spiking Neural Network Modeler (SpiNNMo) capable of simulating a variety of neuromorphic hardware platforms. SpiNNMo is able to extract relevant performance parameters from a neuromorphic chip and then predict the chip’s performance on new networks and data. In this way SpiNNMo can predict accuracy, latency and energy usage for a wide variety of hardware platforms on a given neural network and dataset. This will allow the Navy to test the performance of new spiking neural network architectures and chipsets before the chips are widely available and therefore speed neuromorphic adoption.
rgupta
Regular
So there is a clarity on DOD deal today about the contract. The contract is about changing existing code into neurophonic ( akida) which means the name of second partner will never be available because it is a top secret.
It may also means that Isleki is bluff us that TD had responded to him differently.
Dyor
7für7
Top 20
Hummm what a surprise
FJ-215
Regular
Appendix 4C & Quarterly Activities Report for the Period Ended 31 December 2024
"BrainChip is currently in negotiations with a major defence industry contractor, to enter into a sub-
contractor agreement for the completion of the contract award."
If negotiations have reached a conclusion, is it unreasonable to expect that BRN would inform the market via the ASX platform? Don't care if the subcontractor is named or not. Don't care how many spelling mistakes are in the ann.
Just keep the market updated. Can't be that hard.
As we discovered earlier the Navy is transitioning to Neuromorphic AI. It appears that is the way all defence and security will go.
Humble Genius
Regular
Good Afternoon Chippers ,
Once again iv taken the liberty of stealing this from the other forum.
Cheers StockHound , these numbers also correspond with the last update on price compiled by RockerRothsGettyFellerChild LLC which stood at a conservative $7.117 Au on 9th July 2024.
NOTE , The attached has been generated by ChatGPT.
View attachment 77588
View attachment 77590
View attachment 77589
View attachment 77591
View attachment 77592 View attachment 77593 View attachment 77594 View attachment 77595 Dam it , still can't cut and past the whole picture.
The crapper,
Poster , StockHound81
Today @ 11:00
If a savvy Chipper would be so kind as to retrieve this from the crapper site , and post for all to have a ponder.
Thankyou in advance.
#Well i certainly made mess of that , sorry all
Regards,
Esq.
Question: Esq: what price are you refering to when you mention, "...last update on price compiled by RockerRothsGettyFellerChild LLC which stood at a conservative $7.117 Au on 9th July 2024."?
BTW a buyout between 0-12 months at $7p/s would be great
Similar threads
