BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Brainchip CES 2025 page features LLMs + RAG, Anomoly Detectiob. and Building ML Models with Edge Impulse.
Brainchip CES 2025
https://brainchip.com/ces-2025/
The Edge Impulse partnership is essential for rapid deployment of Akida. NNs require appropriate task-specific models to perform specific tasks. Before Edge Impulse, the models were hand made, a time-consuming and costly task. With EI, models can be assembled automatically using customer data and general models in the same field as the customer's specific tasks.
Then, of course, the models need to be converted to Akida-ese (the spike coding format required for Akida models).
LLMs and RAG, as far as I understand it, means dividing a large universal LLM into subject-specific blocks, and selecting the block or blocks required for the task in hand and using the selected block(s) in configuring Akida. Again, for this to work with Akida the blocks need to be in Akida-usable format, so I would suspect the presence of Edge Impulse at some stage of the process.
https://edgeimpulse.com/all-events/edge-impulse-at-ces
The Edge Impulse team is excited to head back to Vegas in January for another exciting CES!
Our team will be exhibiting with our partners BrainChip, Microchip and CEVA, in addition to walking the floors and meeting with prospects and customers.
Brainchip CES 2025
https://brainchip.com/ces-2025/
LLMs + RAG Demo
See how we’re advancing large language models (LLMs) with Retrieval-Augmented Generation (RAG) for smarter, real-time AI applications.Anomaly Detection Demo
Explore our latest anomaly detection solution running on Raspberry Pi 5. This versatile demo targets multiple verticals, including Industrial IoT, manufacturing, healthcare (wearable devices), cybersecurity, fraud detection, and more.Building ML Models with Edge Impulse
Explore hands-on demos with Edge Impulse, demonstrating how easy it is to build and deploy custom machine learning models directly on the Akida platform.The Edge Impulse partnership is essential for rapid deployment of Akida. NNs require appropriate task-specific models to perform specific tasks. Before Edge Impulse, the models were hand made, a time-consuming and costly task. With EI, models can be assembled automatically using customer data and general models in the same field as the customer's specific tasks.
Then, of course, the models need to be converted to Akida-ese (the spike coding format required for Akida models).
LLMs and RAG, as far as I understand it, means dividing a large universal LLM into subject-specific blocks, and selecting the block or blocks required for the task in hand and using the selected block(s) in configuring Akida. Again, for this to work with Akida the blocks need to be in Akida-usable format, so I would suspect the presence of Edge Impulse at some stage of the process.
https://edgeimpulse.com/all-events/edge-impulse-at-ces
The Edge Impulse team is excited to head back to Vegas in January for another exciting CES!
Our team will be exhibiting with our partners BrainChip, Microchip and CEVA, in addition to walking the floors and meeting with prospects and customers.
KyrieIrving
Emerged
I find this pretty interesting… Could the FG announcement implicate that the Project is already such a success that the contract is already a result of it.
So ESA will use Akida. I think this is the confirmation.
So ESA will use Akida. I think this is the confirmation.
Attachments
Pom down under
Top 20
Boab
I wish I could paint like Vincent
Hey Pom, the podcast is out.
DingoBorat
Slim
Jim Fan
Official Website of Dr. Linxi "Jim" Fan, AI Research Scientist. I obtained my Ph.D. in Computer Science from Stanford, advised by Fei-Fei Li. My primary focus is to develop generally capable autonomous agents. Join me on this grand quest!
(This post is unrelated to the link)
Was watching a humanoid robotics commentary and Jim Fan's name came up, with the above LinkedIn post.
He seems to be one of Nvidia's Top Dogs, in humanoid robotics research.
The implications are just Huge, if he has anything at all, to do with looking at and working with our technology (that's obviously despite the fact, that it's huge anyway, if "anyone" from Nvidia is working with us).
Edit.. And yeah.. Obviously I didn't "read" the short article about "Nvidia GEAR" in your post Doz
..
Last edited:
Came across this in Dec 2023. I cant seem to find the link to it but it came from brn's website. Found it intriguing that the architecture included NVDLA (Nvidia Accelerator) along with NV Encoder (Nvidia encoder) and NV Jpeg (Nvidia encoding, decoding and transcoding of JPEG images).
Not sure what to make out of it. Maybe someone here can help to make sense on why we would need NVDLA, NV Encoder and NV jpeg along with Akida across the Memory controller fabric unless we are aiming for hybrid architecture to offload certain computation between CPU, Akida and Nvidia Accelerator.
Not sure what to make out of it. Maybe someone here can help to make sense on why we would need NVDLA, NV Encoder and NV jpeg along with Akida across the Memory controller fabric unless we are aiming for hybrid architecture to offload certain computation between CPU, Akida and Nvidia Accelerator.
CHIPS
Regular
Latest Holdings, Performance, AUM (from 13F, 13D)
Rothschild Investment Corp /il has disclosed a total of 478 holdings in its most recent SEC filings. The most recent portfolio value is estimated at $1,185,327. Actual assets under management (AUM) is this value plus cash (which is not disclosed). Rothschild Investment Corp /il's largest holdings are Apple Inc. (US:AAPL), JPMorgan Chase & Co. (US:JPM), NVIDIA Corporation (US:NVDA), Alphabet Inc. (US:GOOGL) and Chevron Corporation (US:CVX). Rothschild Investment Corp/il's new positions include Kenvue Inc. (US:KVUE), Patterson-UTI Energy, Inc. (USTEN), iShares Trust – iShares Broad USD High Yield Corporate Bond ETF (US:USHY), iShares Trust – iShares Core S&P Total US Stock Market ETF (US:ITOT) and iShares Trust – iShares MSCI USA Min Vol Factor ETF (US:USM
But is this about BrainChip? I do not find it mentioned anywhere.
Humble Genius
Regular
I like the way these guys think.....
By editor on January 2, 2025 - 04:38
BrainChip (ASX: BRN) gained by 11.54% today to $0.435 per share on 62.26 million shares of volume its second highest volume day of the last twelve months. It is very speculative but so are the quantum computing stocks everybody else is buying. BRN has a chance of one day becoming the world's #1 largest market cap company.
BRN needs to gain another 22.98% to return to its February high of $0.535 per share.
One Stop Systems (OSS) from its current price of $3.35 per share needs to gain another 36.72% to return to its February high of $4.58 per share.
It is highly likely that both OSS and BRN will breakout to well above their February highs this month. OSS has a lot more catching up to do with the largest upside potential and miniscule downside risk.
OSS currently has a healthy 14-day relative strength index (RSI) of only 64.88 vs. BRN's RSI of 77.76.
One day after NIA's initial suggestion of OSS in February when it closed up by 16.28% to $2.50 per share it already had a 14-day relative strength index of 77.20. Three days later when OSS closed on February 20th at $3.05 per share its RSI was extremely overbought at 87.30, but that didn't stop OSS from rising another 50.16% over the following two weeks to hit a high of $4.58 per share.
OSS today having a healthy RSI of 64.88 at $3.35 per share vs. OSS on February 20th having an extremely overbought RSI of 87.30 at a lower share price of $3.05 per share shows that OSS should easily surpass $4.58 per share and hit much higher new 52-week highs this month.
BrainChip (ASX: BRN) doesn't have revenue yet like One Stop Systems (OSS) but we have a feeling it will become big.
It isn't a 100% guaranteed sure thing like OSS but we consider it a "better bet" than any quantum computing stock with no revenue... that's for sure!
BRN and OSS are the only two publicly traded AI companies focused on "edge computing".
Search interest in "Brainchip" peaked in January 2022, when AI wasn't even in play yet:
BrainChip (ASX: BRN) Gains by 11.54% to $0.435 on 62.26M Shares of Volume
By editor on January 2, 2025 - 04:38
BrainChip (ASX: BRN) gained by 11.54% today to $0.435 per share on 62.26 million shares of volume its second highest volume day of the last twelve months. It is very speculative but so are the quantum computing stocks everybody else is buying. BRN has a chance of one day becoming the world's #1 largest market cap company.
BRN needs to gain another 22.98% to return to its February high of $0.535 per share.
One Stop Systems (OSS) from its current price of $3.35 per share needs to gain another 36.72% to return to its February high of $4.58 per share.
It is highly likely that both OSS and BRN will breakout to well above their February highs this month. OSS has a lot more catching up to do with the largest upside potential and miniscule downside risk.
OSS currently has a healthy 14-day relative strength index (RSI) of only 64.88 vs. BRN's RSI of 77.76.
One day after NIA's initial suggestion of OSS in February when it closed up by 16.28% to $2.50 per share it already had a 14-day relative strength index of 77.20. Three days later when OSS closed on February 20th at $3.05 per share its RSI was extremely overbought at 87.30, but that didn't stop OSS from rising another 50.16% over the following two weeks to hit a high of $4.58 per share.
OSS today having a healthy RSI of 64.88 at $3.35 per share vs. OSS on February 20th having an extremely overbought RSI of 87.30 at a lower share price of $3.05 per share shows that OSS should easily surpass $4.58 per share and hit much higher new 52-week highs this month.
BrainChip (ASX: BRN) Doesn't Have Revenue Yet but We Have a Feeling About It
By editor on January 1, 2025 - 22:39
BrainChip (ASX: BRN) doesn't have revenue yet like One Stop Systems (OSS) but we have a feeling it will become big.
It isn't a 100% guaranteed sure thing like OSS but we consider it a "better bet" than any quantum computing stock with no revenue... that's for sure!
BRN and OSS are the only two publicly traded AI companies focused on "edge computing".
Search interest in "Brainchip" peaked in January 2022, when AI wasn't even in play yet:
Posted 3 weeks ago, but, given that models are crucial for implementing NNs, well worth a new year reprise:
https://www.linkedin.com/feed/update/urn:li:activity:7272029972876156928/
Rudy Pei reposted this
M Anthony LewisM Anthony Lewis • 3rd+ • 3rd+CTO@BrainChip | AI, Robotics, Disruptive ComputingCTO@BrainChip | AI, Robotics, Disruptive Computing3w • Edited
Yes to Edge AI!
BrainChip has racked up an impressive number of beyond state of the art models for edge inference. These models are based on our own TENNs family of models. hashtag#TENNS are a kind of State-Space Model recently popularized by hashtag#Mamba. We have been working on our variant for a number of year and have already created hardware to run our models.
I am constantly shocked and amazed to see so many incredible results across a broad array of use-cases.
We are at a defining moment in AI and I am proud to be working with an incredible team making it happen at hashtag#BrainChip
Usually, when companies talk about pushing the boundaries of PPA--- Performance, Power and Area (or cost) they can nudge one of these metrics at the cost of another. We are seemingly riding along a Pareto Frontier and "no company shall pass" that frontier... until now.
With TENN+Akida we are seen remarkable performance and power and model size reduction when compared to conventional CNN model. I think we might be able to schedule a funeral for CNN models for temporal processing.
Why does this matter to anyone? AI everywhere is a solution to climbing energy usage of cloud computing, latency, and expensive back-end inference.
I am guessing 80% of LLM tasks can be pushed to the edge right now and with TENNS+Akida, LLM, audio signal processing and more can be pushed into inexpensive chips.
Incredible levels of intelligence at the edge is now possible at very low cost of ownership.
If you are highly skilled in Edge AI, understand building networks from scratch, have a great positive attitude and like working in a highly collaborative atmosphere, consider working for us in R&D here in sunny Southern California. Lets make this happen together!
https://www.linkedin.com/feed/update/urn:li:activity:7272029972876156928/
Rudy Pei reposted this
M Anthony LewisM Anthony Lewis • 3rd+ • 3rd+CTO@BrainChip | AI, Robotics, Disruptive ComputingCTO@BrainChip | AI, Robotics, Disruptive Computing3w • Edited
Yes to Edge AI!
BrainChip has racked up an impressive number of beyond state of the art models for edge inference. These models are based on our own TENNs family of models. hashtag#TENNS are a kind of State-Space Model recently popularized by hashtag#Mamba. We have been working on our variant for a number of year and have already created hardware to run our models.
I am constantly shocked and amazed to see so many incredible results across a broad array of use-cases.
We are at a defining moment in AI and I am proud to be working with an incredible team making it happen at hashtag#BrainChip
Usually, when companies talk about pushing the boundaries of PPA--- Performance, Power and Area (or cost) they can nudge one of these metrics at the cost of another. We are seemingly riding along a Pareto Frontier and "no company shall pass" that frontier... until now.
With TENN+Akida we are seen remarkable performance and power and model size reduction when compared to conventional CNN model. I think we might be able to schedule a funeral for CNN models for temporal processing.
Why does this matter to anyone? AI everywhere is a solution to climbing energy usage of cloud computing, latency, and expensive back-end inference.
I am guessing 80% of LLM tasks can be pushed to the edge right now and with TENNS+Akida, LLM, audio signal processing and more can be pushed into inexpensive chips.
Incredible levels of intelligence at the edge is now possible at very low cost of ownership.
If you are highly skilled in Edge AI, understand building networks from scratch, have a great positive attitude and like working in a highly collaborative atmosphere, consider working for us in R&D here in sunny Southern California. Lets make this happen together!
Fullmoonfever
Top 20
This guys name popped up in a basic google keyword search so I thought I'd see if been noticed before and @cosors picked him up originally back in 2023 during his internship. Whilst not with us currently, think he's a researcher at the University of Nebraska-Lincoln, I see our Tech Director (I think) has left a recommendation on Mujahirs profile. Looks like he made a couple of nice impacts / inputs with AKD1000.
thestockexchange.com.au
www.linkedin.com
Click here to view Richard Bohl’s profile
Richard Bohl
“I had the pleasure of mentoring Mujahir Abassi in Brainchip's highly competitive Summer Internship 2023 and then extended his work with Brainchip through the fall of 2023. During this time Mujahir completed several projects relating to Yolo object detection model analysis and performance modeling for the next version of Brainchip's AI accelerator IP. Mujahir produced slide sets with operator-level analysis of Yolo v2, v4, v5 and v7 to assist in Brainchip's AI accelerator architectural exploration. Mujahir also created “proxy” model architectures which were models compatible with Brainchip's present hardware that emulated the precise number of MACs required by models not currently supported by Brainchip's hardware. These "proxy" models were first created manually via spreadsheets and then Mujahir automated the process through a series of Python programs. By measuring the performance of the "proxy" models on existing Brainchip hardware, an accurate performance estimate of the emulated model running on Brainchip's future hardware can be calculated. Mujahir's work on these "proxy" models proved invaluable in multiple customer engagements for Brainchip. Training of a neural network model and creation of an inference pipeline for an anomaly detection demo to be presented at CES 2024 was another of Mujahir’s contributions while at Brainchip. The model had to be trained and quantized for execution on Brainchip’s AKD1000 accelerator IC. Furthermore, the inference pipeline needed to manage capturing floating-point sensor data, binning for integer input to the AKD1000, and processing results on a RPi4 host embedded in a custom demo fixture. Many times, Mujahir's close attention to detail found latent issues with existing AI models and Python benchmarking code. Mujahir also demonstrated a near ideal balance of investigating an issue on his own versus asking for assistance. He consistently backed up questions with descriptions of steps taken on his own to solve a problem. This combination of detailed analysis with the intuition of when to ask for assistance is invaluable in a fast-paced, startup environment such as that of Brainchip. The combination of engineering talent and conscientious nature exhibited by Mujahir throughout his work at Brainchip highlights his potential for great contributions to future employers.”
BRN Discussion Ongoing
Once again mate, great darts!! ;)👌 one hundred and eighty
thestockexchange.com.au
Mujahir Abbasi - KATSH DIGITAL ID | LinkedIn
I have over seven years of experience in data science, machine learning, and data… · Experience: KATSH DIGITAL ID · Education: California State University, Los Angeles · Location: Los Angeles Metropolitan Area · 500+ connections on LinkedIn. View Mujahir Abbasi’s profile on LinkedIn, a...
Click here to view Richard Bohl’s profile
Richard Bohl
“I had the pleasure of mentoring Mujahir Abassi in Brainchip's highly competitive Summer Internship 2023 and then extended his work with Brainchip through the fall of 2023. During this time Mujahir completed several projects relating to Yolo object detection model analysis and performance modeling for the next version of Brainchip's AI accelerator IP. Mujahir produced slide sets with operator-level analysis of Yolo v2, v4, v5 and v7 to assist in Brainchip's AI accelerator architectural exploration. Mujahir also created “proxy” model architectures which were models compatible with Brainchip's present hardware that emulated the precise number of MACs required by models not currently supported by Brainchip's hardware. These "proxy" models were first created manually via spreadsheets and then Mujahir automated the process through a series of Python programs. By measuring the performance of the "proxy" models on existing Brainchip hardware, an accurate performance estimate of the emulated model running on Brainchip's future hardware can be calculated. Mujahir's work on these "proxy" models proved invaluable in multiple customer engagements for Brainchip. Training of a neural network model and creation of an inference pipeline for an anomaly detection demo to be presented at CES 2024 was another of Mujahir’s contributions while at Brainchip. The model had to be trained and quantized for execution on Brainchip’s AKD1000 accelerator IC. Furthermore, the inference pipeline needed to manage capturing floating-point sensor data, binning for integer input to the AKD1000, and processing results on a RPi4 host embedded in a custom demo fixture. Many times, Mujahir's close attention to detail found latent issues with existing AI models and Python benchmarking code. Mujahir also demonstrated a near ideal balance of investigating an issue on his own versus asking for assistance. He consistently backed up questions with descriptions of steps taken on his own to solve a problem. This combination of detailed analysis with the intuition of when to ask for assistance is invaluable in a fast-paced, startup environment such as that of Brainchip. The combination of engineering talent and conscientious nature exhibited by Mujahir throughout his work at Brainchip highlights his potential for great contributions to future employers.”
Arm Tech Predictions for 2025 and Beyond
Arm's 2025 tech predictions cover the growth of AI, future of silicon designs and key trends across different technology markets.
newsroom.arm.com
Frangipani
Regular
The first model to feature the all-new Mercedes-Benz Operating System (MB.OS) will actually be the 2025 CLA EV. Its world premiere is scheduled for March, with deliveries slated to start in (Northern hemisphere) summer.
Markus Schäfer on LinkedIn: As 2024 draws to a close, we pause to take stock and to look forward…
As 2024 draws to a close, we pause to take stock and to look forward. Reflecting on the past year it feels like conquering Mount Everest – challenging yet…
A closer look at the upcoming CLA | Mercedes-Benz Group
In a first drive, Mercedes Benz CEO Ola Källenius gives detailed insights into the next-generation CLA at the Mercedes-Benz test facility in Immendingen.
Frangipani
Regular
Speaking of Mauro Diamant - I just revisited his company website for the first time in months, mostly to check whether it still had that eclectic mix of nature photos (it does!) that one wouldn’t exactly expect on a website describing itself as the “virtual Stop-for-Top shop for best-in-class digital IP in Israel”.
According to his LinkedIn profile, Mauro Diamant remains the sales representative of BrainChip in Israel - strangely, though, the IPro Silicon IP website no longer mentions our company at all…
Instead, the logos of two other companies are now prominently displayed next to that of SiFive: Baya Systems (the company that came out of stealth in June and where Nandan Nayampally is now CCO) and Xiphera. This trio is referred to as being IPro’s “leading vendors” - which suggests IPro actually represents more than three IP vendors - but weirdly the reference to “groundbreaking neuromorphic intelligence” seems to point in the direction of Baya Systems, not in ours?
“IPro provides a comprehensive suite of IP solutions, including cutting-edge IP technologies from SiFive, Xiphera and Baya.” This statement also suggests there are more than those three vendors to choose from.
Scrolling further down, however, it sounds as if those three companies are the only vendors that IPro currently represents in Israel?
In Mauro Diamant’s LinkedIn profile, which suggests he still represents BrainChip, Baya Systems is not (yet) listed - as for Xiphera, his profile states he became their sales representative a few weeks ago… Another update incoming?
View attachment 70136
View attachment 70137
View attachment 70138
View attachment 70139
View attachment 70141
View attachment 70140
Three months ago, I had noticed the following about our (former) sales representative in Israel, Mauro Diamant, and his company’s website:
“According to his LinkedIn profile, Mauro Diamant remains the sales representative of BrainChip in Israel - strangely, though, the IPro Silicon IP website no longer mentions our company at all…
Instead, the logos of two other companies are now prominently displayed next to that of SiFive: Baya Systems (the company that came out of stealth in June and where Nandan Nayampally is now CCO) and Xiphera. This trio is referred to as being IPro’s “leading vendors” - which suggests IPro actually represents more than three IP vendors - but weirdly the reference to “groundbreaking neuromorphic intelligence” seems to point in the direction of Baya Systems, not in ours?
“IPro provides a comprehensive suite of IP solutions, including cutting-edge IP technologies from SiFive, Xiphera and Baya.” This statement also suggests there are more than those three vendors to choose from.
Scrolling further down, however, it sounds as if those three companies are the only vendors that IPro currently represents in Israel?
In Mauro Diamant’s LinkedIn profile, which suggests he still represents BrainChip, Baya Systems is not (yet) listed - as for Xiphera, his profile states he became their sales representative a few weeks ago… Another update incoming?”
Not unexpectedly, Mauro Diamant’s LinkedIn profile has since been updated.
Oddly, though, there is no more trace of BrainChip whatsoever in his work experience section…
BrainShit
Regular
He just assumed that Rothschild is invested in BrainChip.
Attachments
I like the way these guys think.....
BrainChip (ASX: BRN) Gains by 11.54% to $0.435 on 62.26M Shares of Volume
By editor on January 2, 2025 - 04:38
BrainChip (ASX: BRN) gained by 11.54% today to $0.435 per share on 62.26 million shares of volume its second highest volume day of the last twelve months. It is very speculative but so are the quantum computing stocks everybody else is buying. BRN has a chance of one day becoming the world's #1 largest market cap company.
BRN needs to gain another 22.98% to return to its February high of $0.535 per share.
One Stop Systems (OSS) from its current price of $3.35 per share needs to gain another 36.72% to return to its February high of $4.58 per share.
It is highly likely that both OSS and BRN will breakout to well above their February highs this month. OSS has a lot more catching up to do with the largest upside potential and miniscule downside risk.
OSS currently has a healthy 14-day relative strength index (RSI) of only 64.88 vs. BRN's RSI of 77.76.
One day after NIA's initial suggestion of OSS in February when it closed up by 16.28% to $2.50 per share it already had a 14-day relative strength index of 77.20. Three days later when OSS closed on February 20th at $3.05 per share its RSI was extremely overbought at 87.30, but that didn't stop OSS from rising another 50.16% over the following two weeks to hit a high of $4.58 per share.
OSS today having a healthy RSI of 64.88 at $3.35 per share vs. OSS on February 20th having an extremely overbought RSI of 87.30 at a lower share price of $3.05 per share shows that OSS should easily surpass $4.58 per share and hit much higher new 52-week highs this month.
BrainChip (ASX: BRN) Doesn't Have Revenue Yet but We Have a Feeling About It
By editor on January 1, 2025 - 22:39
BrainChip (ASX: BRN) doesn't have revenue yet like One Stop Systems (OSS) but we have a feeling it will become big.
It isn't a 100% guaranteed sure thing like OSS but we consider it a "better bet" than any quantum computing stock with no revenue... that's for sure!
BRN and OSS are the only two publicly traded AI companies focused on "edge computing".
Search interest in "Brainchip" peaked in January 2022, when AI wasn't even in play yet:
MDhere
Top 20
Agree with all that agree that Mercedes is involved with us. Why would a trusted brand yell from the rooftops that they are using Akida and cause the share price to go into the $2s and then no longer use us. That in my opinion makes zero sense in a major brand doing this without conviction for Akida.
And for Brainchip to have Mercedes logo on their website says it all ❤. Love Mercedes love Brainchip. ❤ 'Best in class' ❤
Happy Sunday fellow Brners
7für7
Top 20
Agree with all that agree that Mercedes is involved with us. Why would a trusted brand yell from the rooftops that they are using Akida and cause the share price to go into the $2s and then no longer use us. That in my opinion makes zero sense in a major brand doing this without conviction for Akida.
And for Brainchip to have Mercedes logo on their website says it all ❤. Love Mercedes love Brainchip. ❤ 'Best in class' ❤
Happy Sunday fellow Brners
Mercedes no longer has the quality it once did. “Best in class” is a thing of the past when you look at what the Hyundai Group is putting on the road these days, especially with Genesis, or what the Chinese manufacturers like BYD, NIO, and now Lotus are achieving in the electric vehicle sector. The German automotive industry must be feeling the pressure. There are countless complaints about new Mercedes models due to poor interior quality, with everything creaking and rattling, not to mention the engines now coming from China. Whether this can still be called “best in class” is for each person to decide. But it’s still better than American cars, that’s for sure.
Not only my opinion
Similar threads
