BRN Discussion Ongoing

[manny100]

It takes a huge amount of time to get new tech to be accepted.
A change from traditional AI to neuromorphic is a huge step.
We are still at the early adopters stage heading firmly towards the early majority stage. We should be at the early majority stage late this year or perhaps early/mid next year.
It is very difficult to guess timelines for adoption stages but client action 8 months or so indicates that the early adopter stage is progressing quite well.
You may have a different view to me where we are in the stage of adoption.
We are getting there but it takes time. Patience.
Understanding the Technology Adoption Lifecycle

 

[Guzzi62]

It takes a huge amount of time to get new tech to be accepted.
A change from traditional AI to neuromorphic is a huge step.
We are still at the early adopters stage heading firmly towards the early majority stage. We should be at the early majority stage late this year or perhaps early/mid next year.
It is very difficult to guess timelines for adoption stages but client action 8 months or so indicates that the early adopter stage is progressing quite well.
You may have a different view to me where we are in the stage of adoption.
We are getting there but it takes time. Patience.
Understanding the Technology Adoption Lifecycle

Still we have people here screaming for the CEO to be let go! I think that's very premature, especially since the AGM are just around the corner.
Can't you CEO critics at least wait for that until you scream murder!!

Some are strongly against a move to the US! It was inevitable, it was always in the cards, but only when the company is strong enough. You shouldn't do reverse splits, especially without a good, strong income. If you do, it's normally seen as a weak sign.
Again, this will be explained more in detail before the AGM, so why not wait and see when they have to say about it.

Here you can refresh your mind about their thoughts about relocating, well worth a read and think a bit about it.

https://hotcrapper.com.au/threads/ann-brainchip-evaluates-redomiciling-to-us.8468973/

 

[Diogenese]

Listen from 35min. Yann LeCun briefly talks about analog vs digital spiking neural networks. He then refers to some of his colleagues at Meta looking into SNNs for the glasses due to power consumption problems.

There have been a few job adverts with meta referring to SNNs recently.

Hi JB,

Thanks for this.

Obviously Yann is not a follower of Sun Tzu. I think he dismisses SNN too lightly as a competitor to GPU. It does not seem like he has had the time to familiarize himself with Akida (SNN as competitor to NPU? - "not any time soon").

He mentions that CMOS is not sufficiently advanced for SNN, but I think he is talking about analog. He refers to phase change material, which is analog. One of Akida's great practical advantages is that it is implemented in current CMOS.

 

[MadMayHam | 合氣道]

There's 666,666 more shares he is planning to give away (2 more batches) to make a total of 1 million shares being donated this time around.
Expect them all being dumped straight away, as the recipients are generally don't hang on to shares.
Anyway, I suspect he is giving them away before he gets too old (and while they still have little value).

Anyone can point me to info about this donation? This is news to me.

 

[Smoothsailing]

Can some one grab Yann by the ear and March into BRNs office sit him down and don’t let him leave until he listen and it sinks in.
For god sake

 

[yogi]

Our latest hashtag#neuromorphic embedded solution for 24/7 pedestrian detection and tracking with 5W, an EDGE AI FOUNDATION award-winning system, is now ready to be shared with the community from IOP Neuromorphic Computing and Engineering Journal

https://lnkd.in/dVDmvdFy

hashtag#EdgeAI made by SPICES Lab with the KI @ohm from the Technische Hochschule Nürnberg Georg Simon Ohm

 

[IloveLamp]

European Space Agency - ESA has been a pioneer in advancing neuromorphic technologies for space, leveraging a range of pilot activities and studies to explore their potential: | Laurent Hili

European Space Agency - ESA has been a pioneer in advancing neuromorphic technologies for space, leveraging a range of pilot activities and studies to explore their potential: Autonomous landing Autonomous docking Space situational awareness Spectrum monitoring (anti jamming) We are pleased...

www.linkedin.com

Mission accomplished, ESA workshop Morpheus dedicated to “Edge AI and Neuromorphic Hardware Accelerators” took place the last 6th and 7th of May at European Space Research and Technology Centre… | Laurent Hili

Mission accomplished, ESA workshop Morpheus dedicated to “Edge AI and Neuromorphic Hardware Accelerators” took place the last 6th and 7th of May at European Space Research and Technology Centre (ESTEC) in Noordwijk (The Netherlands). Many thanks to ALL the participants and colleagues who...

www.linkedin.com

GR801 | Gaisler

GR801 is a radiation-hardened SoC designed to enable reliable artificial intelligence (AI) applications in space. The GR801 device is part of our GRAIN (Gaisler Research Artificial Intelligence NOEL-V) product line.

www.gaisler.com

Frontgrade Gaisler Launches New GRAIN Line and Wins SNSA Contract to Commercialize First Energy-Efficient Neuromorphic AI for Space Applications | Gaisler

The Swedish National Space Agency (SNSA) has awarded us a contract to commercialize the first neuromorphic System on Chip (SoC) device for space applications.

www.gaisler.com

Episode 37: Neuromorphic Computing in Space

Episode 37: Neuromorphic Computing in Space BrainChip CEO Sean Hehir had the chance to sit down with Frontgrade Gaisler General Manager, Sandi Habinc this week for a conversation on space technology and neuromorphic computing. This episode of our This is Our Mission podcast covers the joint...

brainchip.com

 

[7für7]

So… we go to da moon?

European Space Agency - ESA has been a pioneer in advancing neuromorphic technologies for space, leveraging a range of pilot activities and studies to explore their potential: | Laurent Hili

European Space Agency - ESA has been a pioneer in advancing neuromorphic technologies for space, leveraging a range of pilot activities and studies to explore their potential: Autonomous landing Autonomous docking Space situational awareness Spectrum monitoring (anti jamming) We are pleased...

www.linkedin.com

 

[Bravo]

I reckon that guy's full of shit Frangipani..
I don't care how many letters he has after his name.

I don't think there's any way, that AKIDA properly configured, would be 91 times less efficient than LoiHi.

Either he's doing something very wrong, or his research is being "paid" for by Intel.

Hi Dingo,

If we're talking about real-time cybersecurity applications, the results from the 2025 white paper titled "BrainChip Akida: A Game Changer in AI Computing for Cybersecurity" indicate that Akida offers superior performance in terms of accuracy and energy efficiency.

Benchmark tests compared BrainChip's Akida 1000 and Intel's Loihi 2 using the UNSW TON-IoT dataset, which includes nine attack classes.

The findings were as follows:

• Accuracy: Akida achieved 98.4% accuracy in classifying nine traffic types, surpassing Loihi 2's 90.2%.
• Power Consumption: Akida consumed just 1 watt during inference, compared to Loihi 2's 2.5 watts.
• Model Size Reduction: Akida reduced its model size by 75%, slightly more than Loihi 2's 72.4% reduction.
• Attack Detection: Akida demonstrated higher accuracy in detecting specific attack types, such as Denial of Service (DOS) attacks (95.4% for Akida vs. 52% for Loihi 2).

https://brainchip.com/wp-content/uploads/2025/01/BrainChip_White-Paper-A-Game-Changer-in-AI-Computing-for-Cybersecurity_v3.pdf?utm_source=chatgpt.com

 

[DingoBorat]

Hi Dingo,

If we're talking about real-time cybersecurity applications, the results from the 2025 white paper titled "BrainChip Akida: A Game Changer in AI Computing for Cybersecurity" indicate that Akida offers superior performance in terms of accuracy and energy efficiency.

Benchmark tests compared BrainChip's Akida 1000 and Intel's Loihi 2 using the UNSW TON-IoT dataset, which includes nine attack classes.

The findings were as follows:

• Accuracy: Akida achieved 98.4% accuracy in classifying nine traffic types, surpassing Loihi 2's 90.2%.
• Power Consumption: Akida consumed just 1 watt during inference, compared to Loihi 2's 2.5 watts.
• Model Size Reduction: Akida reduced its model size by 75%, slightly more than Loihi 2's 72.4% reduction.
• Attack Detection: Akida demonstrated higher accuracy in detecting specific attack types, such as Denial of Service (DOS) attacks (95.4% for Akida vs. 52% for Loihi 2).

View attachment 82322

https://brainchip.com/wp-content/uploads/2025/01/BrainChip_White-Paper-A-Game-Changer-in-AI-Computing-for-Cybersecurity_v3.pdf?utm_source=chatgpt.com

"Power Consumption: Akida consumed just 1 watt during inference, compared to Loihi 2's 2.5 watts"

Yeah, it makes no sense Bravo..

If we take it as LoiHi using 2 watts, then this wally is saying AKIDA is 91 times less efficient, or using 182 watts??..

Is my simple mass wrong ...

Sounds like that other Australian university wally last year (from memory an electrical engineer) who said AKIDA was using 180 watts, presumably because that was the power supply rating on the PC Shuttle?..

 

[Bravo]

"Power Consumption: Akida consumed just 1 watt during inference, compared to Loihi 2's 2.5 watts"

Yeah, it makes no sense Bravo..

If we take it as LoiHi using 2 watts, then this wally is saying AKIDA is 91 times less efficient, or using 182 watts??..

Is my simple mass wrong ...

Sounds like that other Australian university wally last year, who said AKIDA was using 180 watts, presumably because that was the power supply rating on the PC Shuttle?..

Yeah, it's a massive discrepancy. All I can think is that the white paper demonstrates performance in certain cybersecurity applications, whereas Dr. Tarek M. Taha was looking at Controller Area Network (CAN) anomaly detection.

Neuromorphic processors like Akida and Loihi are optimized for specific types of tasks and maybe anomaly detection is one that suits Loihi better???

Without knowing what measurement methodologies have been employed or whether, as you say, any peripheral components had been included By Dr. Taha in the power measurements, then it's impossible to know how it could lead to such differing results.

 

[Diogenese]

Yeah, it's a massive discrepancy. All I can think is that the white paper demonstrates performance in certain cybersecurity applications, whereas Dr. Tarek M. Taha was looking at Controller Area Network (CAN) anomaly detection.

Neuromorphic processors like Akida and Loihi are optimized for specific types of tasks and maybe anomaly detection is one that suits Loihi better???

Without knowing what measurement methodologies have been employed or whether, as you say, any peripheral components had been included By Dr. Taha in the power measurements, then it's impossible to know how it could lead to such differing results.

Yes, it's important to compare apples with apples.

Loihi 2 has a range of 1 bit to 32 bits, so comparing a 1 bit Loihi 2 implementation with a 4 bit Akida could skew the results ..,.

... and then there's Akida 2/TENNs.

But as you indicated, without the full paper we're tilting at windmills.

 

[manny100]

Loihi is not commercially available so using it in tests against AKIDA is really just to show that the commercially available AKIDA stands up.
There are no other decent neuromorphic chips (that can perform required cybersecurity functions) to test AKIDA against.
So its us by default in any case.
There has to be 'hard/can't to fix' problems with Loihi otherwise after all this time it would be commercially available. Is Loihi a Dud??
Most chips are specialised eg vision. AKIDA is a general purpose winner.
An example is Bascom Hunter using AKIDA for 5 different puposes in the one device as shown below (and its only using the AKIDA1000 - 5 of them.
General purpose - multi function - that is our EDGE.
APPLICATIONS
A single SNAP Card, deployed on an airborne ISR platform, occupying only one 3U VPX slot, could deploy up to five complex machine learning algorithms across each of its onboard processors to perform the following missions simultaneously:
• Automated Target Recognition (ATR) on full motion video feeds and imagery (to include 4k and higher)
• Real-time detection and identification of threat radars and their acquisition/ operating modes
• Detection of FISINT (Foreign Instrumentation Signature Intelligence) and/or hacking across the airframe’s 1553 communications bus
• Perform communications analysis to include speech-to-text and foreign language translation of intercepted communications
• Fuse multiple infrared cameras (e.g., SWIR, MWIR, LWIR) to provide a combined infrared operating picture on the ground Multi-user, simultaneous modulation/demodulation

 

[manny100]

Sorry but i had to give this concept its own post:
There has to be 'hard/can't to fix' problems with Loihi otherwise after all this time it would be commercially available. Is Loihi a Dud??
That that leave AKIDA as a general purpose Neuromorphic AI event based on chip learner out there on its own?

 

[manny100]

Sorry but i had to give this concept its own post:
There has to be 'hard/can't to fix' problems with Loihi otherwise after all this time it would be commercially available. Is Loihi a Dud??
That that leave AKIDA as a general purpose Neuromorphic AI event based on chip learner out there on its own?

To answer my own question I think at this stage Loihi is a dud.
When, if ever it is made commercially available it will take some time to gain industry confidence.
Brainchip is in the Early adoption stage of the cycle heading towards the Early Majority stage whereas Intel Loihi has not even made it to the Innovation stage because it has not got a commercially available product.
We are 'light years' ahead of Loihi right now.
My opinion of course.

 

[Diogenese]

Loihi is not commercially available so using it in tests against AKIDA is really just to show that the commercially available AKIDA stands up.
There are no other decent neuromorphic chips (that can perform required cybersecurity functions) to test AKIDA against.
So its us by default in any case.
There has to be 'hard/can't to fix' problems with Loihi otherwise after all this time it would be commercially available. Is Loihi a Dud??
Most chips are specialised eg vision. AKIDA is a general purpose winner.
An example is Bascom Hunter using AKIDA for 5 different puposes in the one device as shown below (and its only using the AKIDA1000 - 5 of them.
General purpose - multi function - that is our EDGE.
APPLICATIONS
A single SNAP Card, deployed on an airborne ISR platform, occupying only one 3U VPX slot, could deploy up to five complex machine learning algorithms across each of its onboard processors to perform the following missions simultaneously:
• Automated Target Recognition (ATR) on full motion video feeds and imagery (to include 4k and higher)
• Real-time detection and identification of threat radars and their acquisition/ operating modes
• Detection of FISINT (Foreign Instrumentation Signature Intelligence) and/or hacking across the airframe’s 1553 communications bus
• Perform communications analysis to include speech-to-text and foreign language translation of intercepted communications
• Fuse multiple infrared cameras (e.g., SWIR, MWIR, LWIR) to provide a combined infrared operating picture on the ground Multi-user, simultaneous modulation/demodulation

Hi manny,

I wonder if there's any overlap between this:

"Detection of FISINT (Foreign Instrumentation Signature Intelligence) and/or hacking across the airframe’s 1553 communications bus"

and the DoE/Quantum Ventura/Akida cybersecurity implementation.

They both need models capable of being used to detect aberrant code.

 

[equanimous]

Sorry but i had to give this concept its own post:
There has to be 'hard/can't to fix' problems with Loihi otherwise after all this time it would be commercially available. Is Loihi a Dud??
That that leave AKIDA as a general purpose Neuromorphic AI event based on chip learner out there on its own?

Can Intel commercialise Loihi when BRN has the significant SNN IP rights..

 

[IloveLamp]

Better late than never

(85k followers on LinkedIn)

🚀 𝐁𝐫𝐚𝐢𝐧𝐂𝐡𝐢𝐩 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐞𝐬 𝐀𝐤𝐢𝐝𝐚 𝟏𝟎𝟎𝟎 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫: 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐄𝐝𝐠𝐞 𝐀I 💻✨ | YALLO Group

🚀 𝐁𝐫𝐚𝐢𝐧𝐂𝐡𝐢𝐩 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐞𝐬 𝐀𝐤𝐢𝐝𝐚 𝟏𝟎𝟎𝟎 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫: 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐄𝐝𝐠𝐞 𝐀I 💻✨ BrainChip has unveiled the Akida 1000 processor, a state-of-the-art AI chip designed for edge-based applications. Featuring 1.2 million artificial neurons and 10 billion synapses, the Akida 1000 is optimized for low-latency...

www.linkedin.com

 

[Guzzi62]

Better late than never

(85k followers on LinkedIn)

🚀 𝐁𝐫𝐚𝐢𝐧𝐂𝐡𝐢𝐩 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐞𝐬 𝐀𝐤𝐢𝐝𝐚 𝟏𝟎𝟎𝟎 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫: 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐄𝐝𝐠𝐞 𝐀I 💻✨ | YALLO Group

🚀 𝐁𝐫𝐚𝐢𝐧𝐂𝐡𝐢𝐩 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐞𝐬 𝐀𝐤𝐢𝐝𝐚 𝟏𝟎𝟎𝟎 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫: 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐄𝐝𝐠𝐞 𝐀I 💻✨ BrainChip has unveiled the Akida 1000 processor, a state-of-the-art AI chip designed for edge-based applications. Featuring 1.2 million artificial neurons and 10 billion synapses, the Akida 1000 is optimized for low-latency...

www.linkedin.com

View attachment 82339

Akida1000 uhm, when we should be on to the 2nd generation already, where is it, is no one tapping it out or what?

The 2nd generation, by the way, is called what? Akida2000 or 2? I vote for 2 myself, LOL.

Looking at the company website, they cal it 2nd generation.

Technology

Technology Performance Where It Matters Most Smarter AI at the Edge Starts with Sparsity AkidaTM is built on the principle of sparsity, the idea that accurate AI results come from processing only the most meaningful data. By focusing only on what’s necessary, Akida enables longer battery

brainchip.com

 

[Diogenese]

Better late than never

(85k followers on LinkedIn)

🚀 𝐁𝐫𝐚𝐢𝐧𝐂𝐡𝐢𝐩 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐞𝐬 𝐀𝐤𝐢𝐝𝐚 𝟏𝟎𝟎𝟎 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫: 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐄𝐝𝐠𝐞 𝐀I 💻✨ | YALLO Group

🚀 𝐁𝐫𝐚𝐢𝐧𝐂𝐡𝐢𝐩 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐞𝐬 𝐀𝐤𝐢𝐝𝐚 𝟏𝟎𝟎𝟎 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫: 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐄𝐝𝐠𝐞 𝐀I 💻✨ BrainChip has unveiled the Akida 1000 processor, a state-of-the-art AI chip designed for edge-based applications. Featuring 1.2 million artificial neurons and 10 billion synapses, the Akida 1000 is optimized for low-latency...

www.linkedin.com

View attachment 82339

Well we had Akida 1.0, the engineering sample chip, then Akida 1.1, the demonstration chip, so now we have the COTS Akida 1000 (but we don't know who's making it?) - (nothing to do with "fake it til you make it") ... and we've still got a cupboard full of Akida 1500s, and a display cabinet of Akida 2 FPGAs, as well as an array of designer Picos and M2s. Haven't seen a TENNs chip in the wild yet.

Is this part of the new multi-product line Brainchip - IP, software, chips, engineering support (with Edge Impulse?)?

If we do have a supply of the "limited" Akida 1000 chips, we may start to see a quicker uptake. IP is a very hard sell with a miniscule number of market participants with the necessary cash and capability.

It looks like we have had to build a market for the chips and then get a foundry to make the chips. We will still be involved in designing special purpose SoCs incorporating 3rd party IP/processors. Akida 1000 and 1500 have been taped out and can be made on request. We've had conflicting information on the tapeout of Akida 2. The Akida 2 FPGA is only a shadow of the ASIC. TENNs is the key for high speed temporal applications like micro-Doppler radar with Raytheon/RTX.