BRN Discussion Ongoing

[Fullmoonfever]

Just a very recent article snapshot of BRN. Liked a couple of focuses mentioned.

BrainChip reflects on its first-quarter developments; will revenue follow in 2023? | Edge Industry Review

BrainChip achievements in the first quarter of 2023 include introducing new products, expanding production capabilities and forming strategic collaborations.

www.edgeir.com

BrainChip reflects on its first-quarter developments; will revenue follow in 2023?​

May 7, 2023 | Abhishek Jadhav
CATEGORIES Edge Computing News | Hardware | Public Companies

BrainChip, a company known for its neuromorphic computing devices, recently reflected on its achievements in the first quarter of 2023. These developments include introducing new products, expanding production capabilities and forming strategic collaborations.

BrainChip’s clients can now integrate the Akida processor family with the Arm Cortex-M85 processor core, enhancing performance and energy efficiency for smart edge devices, the company says.

The company recognizes that many industries utilizing IoT devices must adopt higher performance, security standards and machine learning capabilities. This collaboration between BrainChip and Arm enables their partners to manage complex machine-learning workloads.

“As we move from milestone to milestone, our achievements for the first quarter of 2023 bode well for BrainChip’s growth,” says Nandan Nayampally, the chief marketing officer of BrainChip. “From advancing state of the art with our latest product developments to significantly expanding the ecosystem BrainChip inhabits through industry partnerships, we are pushing the edge of AI at a time of rapid market innovation.”
BrainChip also released the second generation of its Akida platform. It employs vision transformers and spatial-temporal convolution to provide high-performance and power-efficient solutions for the network edge. Renesas uses the Akida neural processors for their neuromorphic computing techniques, which provide highly efficient acceleration for AI models at the edge, company executives say.

BrainChip has also entered the automotive industry by joining forces with emotion3D to offer its Akida processor IP to enhance driver safety and user experience. By partnering with emotion3D and leveraging its computing vision and machine learning software, the combined solution can analyze in-cabin environments through image-based analysis to thoroughly understand objects within a vehicle.

Further, BrainChip has collaborated with AI Labs Inc on application development for predictive maintenance in the industrial AIoT sector. The companies say the combined solution utilizes AI Labs’ Minsky AI engine and BrainChip’s Akida and MetaTF platform to improve system health monitoring in mission-critical applications.

Quick take​

In light of the vast potential of neuromorphic computing, there will likely be more developments in this field throughout the year. It will be fascinating to witness the potential utilization of BrainChip’s product portfolio in space applications, for example. That being said, investors will be keen to see the product’s promise turn into sales as well. Financial results for fiscal 2022 show that the company increased its net losses to $22.1M (compared to $20.9M in 2021), though revenues did increase to $5.1M compared to $1.6M for 2021.

 

[Deadpool]

Good post, this all has the pre 4c feels about it, when people were talking up all the millions that could be there but we’re never gonna be there then all the toys came out of the cot because they weren’t there. I fell for something similar before the ces this year. All the pre build up was great, I’m sure the demos were great, and was hoping for another “Mercedes moment” but nothing material.
ARM can/will be a game changing partner but it’s a tech talk, there’ll be some nice demos, it’ll be interesting and informative with talk of possibilities and potentials but there’ll be nothing material in this event. Very happy to eat humble pie tomorrow if wrong but I expect no change to current flat sp trend and think some expectations need to be tempered.

It's purely the validation that I suspect most are excited about, including myself.
Once and for all the Wanca term will be void, when Arm and partners are shouting Akida from the roof tops and everyone that matters will be educated in this Artificial Intelligence of Things marvel, with its genius game changing attributes.
This may just be the end of the beginning of 4 industrial revolution that is already upon us and gaining traction by the day.
my ramblings only

 

[Bravo]

Trying to stay awake.

 

[TechGirl]

Now I must try and stay wawake.

View attachment 36105

 

[cassip]

these are the news from IBM for May, 9th:

IBM unveils new watsonx, AI and data platform​

...The new AI platform launch comes over a decade after IBM's software called Watson got attention for winning the game show Jeopardy. IBM at the time said Watson could “learn” and process human language. But Watson's high cost at the time made it a challenge for companies to use, according to Reuters reporting.
..."When something becomes 100 times cheaper, it really sets up an attraction that's very, very different," said Krishna. "The first barrier to create the model is high, but once you've done that, to adapt that model for a hundred or a thousand different tasks is very easy and can be done by a non-expert."...
He added that IBM was also embracing a more open ecosystem and partnering with open-source AI software development hub Hugging Face and others.

IBM said companies can use the watsonx platform to train and deploy AI models, automatically generate code using natural language and use various large language models built for different purposes such as chemical creation or climate change modeling.

IBM unveils new watsonx, AI and data platform

International Business Machines Corp on Tuesday launched watsonx, a new artificial intelligence and data platform to help companies integrate AI in their business.

www.reuters.com

 

[TechGirl]

30 minutes away, ARMageddon tired

 

[MDhere]

im in

 

[MDhere]

That's not it.

huh says olease waitthe webinar will begin soon

 

[SERA2g]

Webinar just ended.

If anyone cares to look into the people that asked the questions, I've listed their names and questions below for reference.

There were additional questions but they were from the below people. I was just wanting to capture all names (rather than questions) to search for dots.

Nandan presented well. He mentioned we're working with an agricultural company as an example for object detection.

Nandan said in conversation 'working on state of the art de noising' (not sure if brainchip or a customer using akida), and Brainchip are 'putting a paper out on TENN shortly'.

Todd answered a few questions. Rob didn't say anything until he said goodbye.

The tech talk will be made available so you can watch a replay if you'd like to. I assume it can be found on the Arm website.

Ludovico Rella 11:20 pm

How much is your hardware used for Edge Inference, and how much does it enable forms of Edge Training and Distributed Learning (Federated, Swarm, etc.)? Is the whole model loaded on devices, or loaded layer by layer? What about pruning/quantization/sparcity? Thank you very much this is a very exciting talk!


Shyama Sastha Krishnamoorthy Srinivasan 11:31 pm

With an AI model that is compute-intensive, it seems pretty straightforward as to how Akida helps meet the requirements. But, even if the computation is done on raw signals, the load will still be considerable. So, how does this affect the power load? Also, even if it somehow works with lower power, it still needs to manage the heat over time. How does it wok here?


Krishna Vemireddy 11:39 pm

In terms of silicon area, how does AKida compare with

* Cortex M-85 only

* and also M-85+ Ethos


amir arjmand 11:41 pm

which Board do you suggest to deploy our model on it?


Nikola Trajic 11:42 pm

Could you please tell what are Akida SNN capabilities of on-device learning? An impression is that breakthroughs are in inference speed up mostly, with keeping up the accurace with pruning and other optimizations.


Frederik Schack 11:43 pm [Hi @FrederikSchack ]

How many neurons with 4 bit weights can Akida-P emulate, with on chip weights?


Ali Çolak 11:45 pm

There was a slide about MFCC + CNN that is efficiencly working to use in keyword spoting. Are the MFCC steps taking as input to CNN or MFCC and also other type of data, which is fedding to CNN, taking as input


Alex Epstine 11:47 pm

Please, tell me, is it possible to use a TENN to reduce the noise of the audio stream?


Remy Pottier 11:48 pm

SNN have been known do be difficult to train. how have you solved the problem?


Gautham Sai Vadicherla 11:50 pm

could you explain more about TENN, what existing networks can it replace, how could it circumvent FFTs and what are it's limitations respect to the applications


Hugo Martin 11:55 pm

On the medical front, an example I often ponder is Akida being able to identify disease through touching the skin. According to my understanding of Akida this is theoretically possible. Am I correct in my reasoning?

 

[cosors]

Did he passively mentioned or confirm Deere with the big agricultural company?

 

[Falling Knife]

Great presentation by the team and good response to Q&A, it was recorded so will be available later to review.

I don't know why I was thinking we would get to see a live demo of a product incorporating ARM Cortex and Akida.
A static presentation was once again showing us what we already know, AKIDA is agnostic, it touches on all sensory faculties and is more efficient on many fronts than other related processes.

My two favourite slides

I still don't understand how the $ come when one company goes through this process and decides to volume produce a product incorporating an ARM cortex M85 WITH akida IP
Will ARM pay the IP License on the customers behalf or will the company pay ARM and BRN directly for each part of the product?

Not sure this will support the SP tomorrow. Although I'm very surprised after 40k in the last report we held up so well

 

[SERA2g]

Did he passively mentioned or confirm Deere with the big agricultural company?

No, but we know it is.

 

[Sirod69]

BrainChip
@BrainChip_inc

Did you miss the @Arm Tech Talk with the BrainChip team? You can watch the video presentation here: AI at the Edge: Going Cloudless with Arm and BrainChip

Arm Tech Talk from Brainchip: AI at the Edge: Going Cloudless with Arm and BrainChip​

 

[Deleted member 118]

BrainChip
@BrainChip_inc

Did you miss the @Arm Tech Talk with the BrainChip team? You can watch the video presentation here: AI at the Edge: Going Cloudless with Arm and BrainChip

Arm Tech Talk from Brainchip: AI at the Edge: Going Cloudless with Arm and BrainChip​

 

[Sirod69]

Edge Impulse

As we look ahead to the Embedded Vision Summit later this month, Edge AI and Vision Alliance's Jeff Bier sheds some light on the biggest trends in embedded AI.

Multimodal perception
“AI everywhere"
Faster, cheaper and more capable processors
Low-code/no-code development (like Edge Impulse!)
Generative AI

Key Trends Shaping the Embedded AI Landscape​

............

As the program chair of the Embedded Vision Summit, I’ve spent the last six months reviewing hundreds of session proposals focused on practical computer vision and embedded AI, putting me in a unique position to see some of the evolving trends in this field. But in noticing these trends, I always try to ask: Why? What is driving this trend? What problem is being solved?
.................

—Jeff Bier is the president of consulting firm BDTI, founder of the Edge AI and Vision Alliance, and the general chair of the Embedded Vision Summit.

https://www.eetimes.com/key-trends-shaping-the-embedded-ai-landscape/

 

[Beebo]

ARM TechTalk: the JUICIEST bit for me was the discussion around the very last question on whether or not TENN can reduce noise in an audio stream!

The brief reply: YES, we are working on de-noising applications. Combine that with our ability to learn, now you can do contextual de-noising, etc.

Are we working on that with a “communications” company (?)

Let’s go BRN!

 

[DK6161]

Description:
This Tech Talk will highlight the ease of deploying best-in-class power efficient AI combining BrainChip's Akida Neuromorphic AI Accelerator with the Arm Cortex-M family., including the Arm Cortex-M85. The presentation is designed to showcase the disruptive potential of the technology combination, including a look at the 2nd generation of Akida to create compelling, cloudless Edge AI solutions for advanced vision, video object detection and vital signs prediction, that weren’t possible before.

Have to say that I am a tad excited to see the demo.

Well that was disappointing. No bloody demo

 

[Deleted member 118]

ARM TechTalk: the JUICIEST bit for me was the discussion around the very last question on whether or not TENN can reduce noise in an audio stream!

The brief reply: YES, we are working on de-noising applications. Combine that with our ability to learn, now you can do contextual de-noising, etc.

Are we working on that with a “communications” company (?)

Let’s go BRN!

There was more juice in this bowl then there was in the discussion

 

[Moonshot]

Well that was disappointing. No bloody demo

Agree, was a lunch and learn type presentation. Can’t wait to tune in for the electric toothbrush voice on arm talk next week.

Why is Nandan doing these and not Rob?

 

[Schnitzel lover]

Bring on 2025