BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Hi Charles2,
As you say, inference is Akida's forte. It is built on highly efficient sparsity in both the activations and in the models which implement N:M coding.
The training of Akida's models has also become a strong point thanks to BRN's partnership with Edge Impulse, and its automatic model generation augmented by Akida's ML.
Akida's GenAI capabilities are also beginning to be built up with the inclusion of long skip in Akida 2. The ability to implement this in silicon relieves the CPU co-processor of the need to implement this in software, leaving the CPU to perform the deductive reasoning or whatever it does.
JDelekto
Regular
The individual in sales, with whom I've exchanged e-mails recently, said they did offer a USB version at one point, but it "did not offer the right level of performance," so they don't offer it anymore.Interesting that the image of Intel Loihi appears in a USB form factor. Wasn’t aware Intel reduced Loihi down to this size. Last I checked it fit in a brief case
Didn’t Brainchip delay producing a USB version of Akida. Didn’t Brainchip also not tape out Akida 2.0.
It makes sense, considering it probably employed the earlier, slower USB specification, and their current I/O interface used by both the PCIe and NVMe devices use dual PCI lanes.
Agree,BrainChip Powers Cyberthreat Tool
January 23, 2025by The Neuron
The integration of BrainChip’s ultra-low power neuromorphic AI into Quantum Ventura’s CyberNeuro-RT technology is poised to revolutionize cybersecurity protection for network access devices. By leveraging the Akida processor, this innovative solution enables on-chip learning and threat detection at the edge, providing a robust defense against cyber threats without relying on cloud-based servers or central processing units.
This collaboration has yielded a cutting-edge technology that can identify and adapt to emerging threats in real-time, utilizing BrainChip’s event-based neuromorphic principles to mimic the human brain and analyze essential sensor inputs with unparalleled efficiency and precision. As the demand for robust cybersecurity solutions continues to grow, this partnership is set to play a pivotal role in safeguarding WiFi access, home routers, small enterprise routers, and other network access devices from malicious attacks, ultimately enhancing the security and integrity of our increasingly interconnected world.
Introduction to Neuromorphic Processing for Cybersecurity
The integration of neuromorphic processing into cybersecurity solutions has the potential to revolutionize the way threats are detected and mitigated. BrainChip Holdings Ltd, a company specializing in ultra-low power, fully digital, event-based, neuromorphic AI, has announced its collaboration with Quantum Ventura to provide cybersecurity protection for WiFi access, home routers, small enterprise routers, and other network access devices. This partnership aims to leverage the Akida processor, BrainChip’s flagship product, to facilitate on-chip learning for deployment of network-specific attack detection.
The increasing need for cybersecurity solutions that can thwart attacks through otherwise unsecure devices connected to the network has driven the development of innovative technologies like Quantum Ventura’s CyberNeuro-RT (CNRT). This technology offering, developed in partnership with Lockheed Martin Co.’s MFC Division and Pennsylvania State University, utilizes BrainChip’s neuromorphic processing capabilities to provide managed cybersecurity support for edge devices that cannot rely on a central server to identify threats and attacks due to cost or power issues. The integration of Akida into CNRT enables the detection of threats across multiple devices that would otherwise be vulnerable to exploitation.
The Akida neural processor and AI IP can identify unknown repeating patterns in vast amounts of noisy data, making it an asset in cyberthreat detection. By learning what normal network traffic patterns look like, Akida can detect malware, attack signatures, and other types of malicious activity. Its unique ability to learn on-device in a secure fashion, without the need for cloud retraining, allows it to quickly adapt to emerging threats.
Neuromorphic Processing and Cyberthreat Detection
Neuromorphic processing, inspired by the human brain’s neural structure and function, offers a promising approach to cyberthreat detection. The Akida processor, with its event-based technology, is inherently lower power than conventional neural network accelerators, providing energy efficiency with high performance for partners to deliver AI solutions previously not possible on even battery-operated or fan-less embedded edge devices. BrainChip’s IP supports incremental learning, on-chip learning, and high-speed inference with unsurpassed performance in micro-watt to milli-watt power budgets, making it ideal for advanced AI/ML devices such as intelligent sensors, medical devices, and autonomous systems.
The integration of neuromorphic processing into cybersecurity solutions can provide high-quality protection against cyber threats. Akida’s on-chip learning can adapt to new threats and redirect unknown threats to the cloud, providing faster and more cost-efficient analysis capabilities than otherwise possible. This approach enables effective edge compute to be universally deployable across real-world applications, reducing latency while improving privacy and data security.
Edge AI and Cybersecurity
Edge AI, which involves processing data at the point of acquisition, is becoming increasingly important in cybersecurity. The ability to analyze data locally, without relying on cloud connectivity, can significantly improve response times and reduce the risk of data breaches. BrainChip’s Akida processor, with its neuromorphic architecture, is well-suited for edge AI applications, enabling Edge learning locally to the chip, independent of the cloud.
The benefits of edge AI in cybersecurity are numerous. By processing data at the edge, latency is reduced, and real-time threat detection becomes possible. Additionally, edge AI can improve privacy and data security by minimizing the amount of sensitive data that needs to be transmitted to the cloud for analysis. As the number of connected devices continues to grow, the importance of edge AI in cybersecurity will only continue to increase.
Quantum Ventura’s CyberNeuro-RT Solution
Quantum Ventura’s CyberNeuro-RT solution is a systems integration and environmental technology innovation company that aims to deliver customer-centric advanced solutions for U.S. Federal & State Governments and Private Sector customers. The CNRT solution, developed in partnership with BrainChip, Lockheed Martin Co.’s MFC Division, and Pennsylvania State University, utilizes neuromorphic processing to provide managed cybersecurity support for edge devices.
The CNRT solution has the potential to revolutionize the way cyber threats are detected and mitigated. By leveraging BrainChip’s Akida processor, Quantum Ventura can provide high-quality protection against cyber threats, reducing the risk of data breaches and improving overall cybersecurity posture. As the demand for effective cybersecurity solutions continues to grow, Quantum Ventura’s CNRT solution is well-positioned to meet the needs of government and private sector customers.
Conclusion
The integration of neuromorphic processing into cybersecurity solutions has the potential to significantly improve the way threats are detected and mitigated. BrainChip’s collaboration with Quantum Ventura to provide cybersecurity protection for WiFi access, home routers, small enterprise routers, and other network access devices is a promising development in this field. As the importance of edge AI in cybersecurity continues to grow, companies like BrainChip and Quantum Ventura are well-positioned to deliver innovative solutions that meet the needs of government and private sector customers.
BrainChip Powers Cyberthreat Tool
The integration of BrainChip's ultra-low power neuromorphic AI into Quantum Ventura's CyberNeuro-RT technology is poised to revolutionize cybersecurity protection for network access devices. By leveraging the Akida processor, this innovative solution enables on-chip learning and threat detection...quantumzeitgeist.com
It looks like there will be no Akida IP silicon in ADAS in the near future, and Akida simulation software appears to be impractical for EVs. As DB says, there's always hope it can make an appearance in infotainment.
However, the automotive NN market is dwarfed by the cybersecurity NN market. This is something I've dreamt of since the DUTH days, although I would like to see Akida in USB format for the retro-fit market.
I do think it is becoming clearer day-by-day that the IP only strategy was a strategic mistake. The barriers to entry were just too great, the time-to-market too long, and the customer base too small, particularly considering their sunk cost bias.
Hindsight, spilt milk, water under the bridge, missed opportunities - we're where we are, and the mis-steps are behind us - now we've learnt the foxtrot, we're ready to tango!
Smoothsailing
Regular
I sent an email to BRN today asking to purchase a USB stick and a Wifi router/ modem and if they are available
Waiting for a reply now
Terroni2105
Founding Member
I agree Dio, but I think Sean reiterated again in that recent podcast interview he did that we are (still) an IP company. Someone correct me if my memory is not serving me well (I couldn’t be bothered listening to it again
)
HopalongPetrovski
I'm Spartacus!
Boab
I wish I could paint like Vincent
Hopefully we get there before we're all like this................
Tell me thats not a colostomy bag.
HopalongPetrovski
I'm Spartacus!
No, think it's where he keeps his emergency diaper.Tell me thats not a colostomy bag.
C'mon BrainChip, whilst we can still get it up!
DingoBorat
Slim
The A.I. Model wars are intensifying, with 2 new Chinese ones introduced, in the last several hours.
One by Alibaba no less..
Humble Genius
Regular
Even better would be for Intel to buy us out at US$5 / share. They will then become a behemoth and each and everyone of us will have atleast tripled our money. Some of us will have made 248x our investment and this is better than nothing
Cyw
Regular
Even better is Deepseek, NVidia, Microsoft, Amazon, Google and Meta sign up and buy our technology. I still think it is not if, but when.Even better would be for Intel to buy us out at US$5 / share. They will then become a behemoth and each and everyone of us will have atleast tripled our money. Some of us will have made 248x our investment and this is better than nothing
Pom down under
Top 20
Actually I was walking into my garage and saw my grey Aston Martin ….
Jacket I baught in a second hand shop
Dio, you have at least twice posted about your dissatisfaction when Sean convinced the Board not long after his appointment to abandon the AKD 1000 NSoC approach we all thought was our next phase, I fully understand the reasoning behind this (monetary) decision, and I have been rather outspoken on Seans comments regarding his "too narrow" comments about our brilliant first run chip, this message I have personally mentioned to both Peter and Anil, as I felt it was disrespectful, to which Sean has watered down his comments numerous times since.
Many shareholders will have noticed the traction that AKD 1000 has gained, it's referenced, photographed, researched, published in all tech articles.
You, I and a number of others familar with the IP approach understand the journey is longer, harder, in that we are trying to not only sell the technology in " blocks", we are saying to clients, you go away and secure a fab and take all the financial risk, it's not an easy sell, especially in this current worldwide financial instability.
Personally, with Sean into his 4th year of a 5 year business plan I can't see us changing direction, but I do currently see a few alternate paths being tested.
The key thing for me with our company is, major tier 1 companies continue to engage, partner, collaborate and generally praise our technology, that's the hardest hurdle and we have cleared it with flying colours !
Yes it's painfully slow, but my belief and loyalty are as strong as back in late 2015.
Akida Tech.Make it $10 and you’ve got a dealEven better would be for Intel to buy us out at US$5 / share. They will then become a behemoth and each and everyone of us will have atleast tripled our money. Some of us will have made 248x our investment and this is better than nothing
Pom down under
Top 20
Time is ticking and very fast and me like probably many other share holders are waiting for this one decent contract in the next 2-3 months for Sean to save his job. If not he should beDio, you have at least twice posted about your dissatisfaction when Sean convinced the Board not long after his appointment to abandon the AKD 1000 NSoC approach we all thought was our next phase, I fully understand the reasoning behind this (monetary) decision, and I have been rather outspoken on Seans comments regarding his "too narrow" comments about our brilliant first run chip, this message I have personally mentioned to both Peter and Anil, as I felt it was disrespectful, to which Sean has watered down his comments numerous times since.
Many shareholders will have noticed the traction that AKD 1000 has gained, it's referenced, photographed, researched, published in all tech articles.
You, I and a number of others familar with the IP approach understand the journey is longer, harder, in that we are trying to not only sell the technology in " blocks", we are saying to clients, you go away and secure a fab and take all the financial risk, it's not an easy sell, especially in this current worldwide financial instability.
Personally, with Sean into his 4th year of a 5 year business plan I can't see us changing direction, but I do currently see a few alternate paths being tested.
The key thing for me with our company is, major tier 1 companies continue to engage, partner, collaborate and generally praise our technology, that's the hardest hurdle and we have cleared it with flying colours !
Yes it's painfully slow, but my belief and loyalty are as strong as back in late 2015.
Smoothsailing
Regular
The recent upgrade to the BRN website to include a detailed list of were Pico will be going and into what products to me means they are about to sign a deal Very soon
Go brainchip
Tothemoon24
Top 20
The good Doctor is an absolute 
️
DeepSeek R1's breakthrough results are a boon for BrainChip and the extreme-edge AI industry. Here's how it will impact inference at the extreme edge:
DeepSeek has shattered the "No one ever got fired for going with IBM" mentality, liberating customers from the behemoths' models. This paves the way for a Cambrian explosion of specialized LLMs delivered by smaller companies. It's akin to breaking the 4-minute mile – now that it's been done, we're poised to witness an explosion of innovation. This is excellent news for smaller companies like BrainChip.
Now imagine AI models running on whispers of energy, delivering powerful performance without massive data centers or cloud connectivity. This is the promise of extreme-edge AI, which is BrainChip's sweet spot.
While DeepSeek is an impressive transformer based model, TENNs technology, based on State Space Models (SSMs), currently offers a superior solution for extreme-edge applications:
1. No explosive KV cache: Unlike traditional transformer models that rely on rapidly expanding key-value (KV) caches, TENNs sidesteps this issue with fixed memory requirements, regardless of input length. This fundamental property enables LLMs at the extreme edge.
2. Competitive Performancereliminary experiments pitting the DeepSeek 1.5B model against TENNs 1.2B model in various tasks, such as trip planning and simple programming, showed comparable or slightly better results for TENNs.
3. Extremely Low Training Costs: BrainChip's focus on small specialized models means training costs are less than a premium economy flight from Los Angeles to Sydney.
DeepSeek's success highlights areas where TENNs can be further enhanced. We can leverage many of the tricks learned from DeepSeek to improve their TENNs LLM even more.
The future of extreme-edge AI is bright, with DeepSeek demonstrating that small companies can compete effectively in this space.
️DeepSeek R1's breakthrough results are a boon for BrainChip and the extreme-edge AI industry. Here's how it will impact inference at the extreme edge:
DeepSeek has shattered the "No one ever got fired for going with IBM" mentality, liberating customers from the behemoths' models. This paves the way for a Cambrian explosion of specialized LLMs delivered by smaller companies. It's akin to breaking the 4-minute mile – now that it's been done, we're poised to witness an explosion of innovation. This is excellent news for smaller companies like BrainChip.
Now imagine AI models running on whispers of energy, delivering powerful performance without massive data centers or cloud connectivity. This is the promise of extreme-edge AI, which is BrainChip's sweet spot.
While DeepSeek is an impressive transformer based model, TENNs technology, based on State Space Models (SSMs), currently offers a superior solution for extreme-edge applications:
1. No explosive KV cache: Unlike traditional transformer models that rely on rapidly expanding key-value (KV) caches, TENNs sidesteps this issue with fixed memory requirements, regardless of input length. This fundamental property enables LLMs at the extreme edge.
2. Competitive Performance
reliminary experiments pitting the DeepSeek 1.5B model against TENNs 1.2B model in various tasks, such as trip planning and simple programming, showed comparable or slightly better results for TENNs.3. Extremely Low Training Costs: BrainChip's focus on small specialized models means training costs are less than a premium economy flight from Los Angeles to Sydney.
DeepSeek's success highlights areas where TENNs can be further enhanced. We can leverage many of the tricks learned from DeepSeek to improve their TENNs LLM even more.
The future of extreme-edge AI is bright, with DeepSeek demonstrating that small companies can compete effectively in this space.
Last edited:
Similar threads
