BRN Discussion Ongoing

[IloveLamp]

Technology Trend Radar

The BMW Group Trend Radar provides deep insights into our strategy of finding the most innovative products and services for our customers.

www.bmwgroup.com

 

[Fullmoonfever]

Hmmmm ...

What we need to take into consideration is the crucial fact that the product has been in a state of ongoing development for quite a while and this would have played a major part in management's selection of the business model.

For more than 2 years, the company was aware that Akida 2 with TeNNs was in the pipeline, and they would have understood that this would influence the EAPs. The company would have been anxious not to lumber the EAPs with large numbers of "superceded" chips, or at least to give them the choice.

While TeNNs is still not available in silicon, simulation software would have been available for EAPs from the time the patent application was filed, and that would have been undergoing continual improvement. Our partner EAPs would have been in on the secret. EAPs like Valeo and Mercedes have competitive pressures which would have weighed against waiting for Akida 2 SoC, which is one of the reasons why I suspect they have gone ahead with the simulation software as an interim measure. The software was immediately available and readily adaptable as product development progressed.

The development must have been finished, or at least reached a set benchmark, when Anil announced tapeout (about 6 months ago?), but this was then superceded by an overriding announcement that a masked stranger was about to do us a power of good and take over the chip production - since then, not a sausage.

So my Hail Mary speculation is that we are running a parallel software licensing program in parallel with, or in advance of the IP licensing program. Given that this would be in the context of joint developments, the details would be commercial-in-confidence.

All this is derived from my reading of the tea leaves (the crystal ball is being fitted with Gen AI LLMs as we speak).

The Edge Box has sold out, so, with the hindsight of the development of TeNNs, I am reluctant to call it a mistake, or even a tactical error, but it would have been nice to have had more chips for sale.

Gosh - I didn't know I knew that many words!

Well....there is a software licence doc so appears that's an option maybe? Not sure if all encompassing for all software or just certain components.

Haven't snipped the whole thing.

 

[DingoBorat]

I presume the UNIGEN Cupcake AI Edge server chewed up a lot of AKIDA 1000 chips..
In Feb'24 Unigen announced it had produced the 1st lot and that rigorous testing had been completed.

Cupcake Edge AI Server in Full Production - Unigen

Unigen Corporation proudly announces the successful production launch of its highly anticipated Cupcake Edge AI Server. The first units have been produced at our cutting-edge facilities in Hanoi, Vietnam, and Penang, Malaysia, marking a significant milestone in Unigen’s commitment to delivering...

unigen.com

So at this stage, we are not being offered in the Unigen Cupcake and the VVDN Edge Box, is sold out.

Either all available chips have been utilised, of which we have no idea of the starting quantity (but at least a few thousand) or there have been delays in product development.

We won't have a clearer picture (although still muddied) until we get the quarterly results, in just under 3 weeks time.

And does anybody have any idea, of what happened, to the AKD1500 chips that were produced?..

BrainChip Receives First Shipment of AKD1500 Chips in Silicon from Technology Partner GlobalFoundries

BrainChip's latest milestone: AKD1500 chip arrival from GlobalFoundries. Explore the future of semiconductors here.

www.google.com

Could there have only been enough produced, for development boards, as it was the first run?..

 

[Diogenese]

Well....there is a software licence doc so appears that's an option maybe? Not sure if all encompassing for all software or just certain components.

Haven't snipped the whole thing.

View attachment 66397 View attachment 66398 View attachment 66399

That is to cover the MetaTF simulation software which is provided to purchasers of Akida chips. It expressly forbids commercial use.

If there are the supposed licences I imagine, they will be under NDA with the EAPs.

 

[Slade]

You’re embarrassing yourself.

That’s not how I see it.

 

[Diogenese]

Sorry, my bad. I thought this was a collective effort.

View attachment 66311

Well pop goes your PhD my lad!

Plagiarism from the original source is ok, but from another copier ...?

 

[Kachoo]

Hi all,

I would take a look at the following companies for any nuggets on the cupcakes

Apexon they own Ailabs inc as the linkedinnsite directs you to their site.

Cupcake Edge AI Server - Unigen

Unigen’s Cupcake Edge AI Server delivers a reliable, high-performance, low-latency, low-power platform for Machine Learning and Inference AI.

unigen.com

On here it state multiple vendors of OS and Ai.

Also for sales Unigen directs you to the Ailabs site for sales in Asia!

 

[Kachoo]

Good Afternoon Diogenese,

Total including CXA trades is 4,905,565 units .... but yes a very gay effort on their behalf.

Regards,
Esq,

Hey Esq gay or straight I'm not the one to judge peoples life style.

But the trading well until there is some activity or interest in new money they will keep playing this game they do.

It's a left hand to right hand when they stop its usually cause they need to wait out and catch some shares from sellers to balance any deficit. I mean we can't trade at 1/4 a cent but they can. Why can't we do that.

If there was more net buyers then sellers then you would see the price move north then again they do these pump and dumps too but usually that is because there is a party taking a position.

Like last time word was a buyer did take a position over January so yeah that activity excites people and they buy driving the price up with the computer programs still.

Really if you broken-hearted trades down if your tracking them Esq I bet the days you see predominantly the 1/4 cent trades they are red days the 1/2 cent day which is normal you would see a net positive day for the most part. Imo

 

[Frangipani]

Well pop goes your PhD my lad!

Plagiarism from the original source is ok, but from another copier ...?

Are you seriously accusing me of plagiarism now?!

 

[Bravo]

Are you seriously accusing me of plagiarism now?!

Could be worse. He accused me of being a lad!

 

[Dallas]

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. | BrainChip

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. Traditional methods have often been computationally intensive or inaccurate, increasing costs and risks for remote healthcare. BrainChip’s #Akida...

www.linkedin.com

 

[Slade]

Are you seriously accusing me of plagiarism now?!

Stop it. You are embarrassing yourself.

 

[Dallas]

💥 PROPHESEE investit 15 millions d'euros pour le développement de l’IA neuromorphique « Made in France ». | Paul-François Fournier

💥 PROPHESEE investit 15 millions d'euros pour le développement de l’IA neuromorphique « Made in France ». Le projet vise au développement d’une nouvelle génération de capteurs de vision neuromorphiques, fabriqués en France, destinés à l’IA embarquée, visant un marché total de 1,3 milliard de...

www.linkedin.com

 

[IloveLamp]

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. | BrainChip

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. Traditional methods have often been computationally intensive or inaccurate, increasing costs and risks for remote healthcare. BrainChip’s #Akida...

www.linkedin.com

 

[DK6161]

Hey Esq gay or straight I'm not the one to judge peoples life style.

But the trading well until there is some activity or interest in new money they will keep playing this game they do.

It's a left hand to right hand when they stop its usually cause they need to wait out and catch some shares from sellers to balance any deficit. I mean we can't trade at 1/4 a cent but they can. Why can't we do that.

If there was more net buyers then sellers then you would see the price move north then again they do these pump and dumps too but usually that is because there is a party taking a position.

Like last time word was a buyer did take a position over January so yeah that activity excites people and they buy driving the price up with the computer programs still.

Really if you broken-hearted trades down if your tracking them Esq I bet the days you see predominantly the 1/4 cent trades they are red days the 1/2 cent day which is normal you would see a net positive day for the most part. Imo

 

[IloveLamp]

https://www.eetimes.com/unleashing-the-potential-of-alternative-deep-learning-hardware/

 

[supersonic001]

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. | BrainChip

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. Traditional methods have often been computationally intensive or inaccurate, increasing costs and risks for remote healthcare. BrainChip’s #Akida...

www.linkedin.com

 

[7für7]

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. | BrainChip

The integration of #AI and #ML is transforming #healthcare by enabling more effective and portable solutions for vital signs analysis and diagnostics. Traditional methods have often been computationally intensive or inaccurate, increasing costs and risks for remote healthcare. BrainChip’s #Akida...

www.linkedin.com

View attachment 66417

This was already posted by @Dallas .. I postet it too and deleted it.

 

[Yoda]

Hmmmm ...

What we need to take into consideration is the crucial fact that the product has been in a state of ongoing development for quite a while and this would have played a major part in management's selection of the business model.

For more than 2 years, the company was aware that Akida 2 with TeNNs was in the pipeline, and they would have understood that this would influence the EAPs. The company would have been anxious not to lumber the EAPs with large numbers of "superceded" chips, or at least to give them the choice.

While TeNNs is still not available in silicon, simulation software would have been available for EAPs from the time the patent application was filed, and that would have been undergoing continual improvement. Our partner EAPs would have been in on the secret. EAPs like Valeo and Mercedes have competitive pressures which would have weighed against waiting for Akida 2 SoC, which is one of the reasons why I suspect they have gone ahead with the simulation software as an interim measure. The software was immediately available and readily adaptable as product development progressed.

The development must have been finished, or at least reached a set benchmark, when Anil announced tapeout (about 6 months ago?), but this was then superceded by an overriding announcement that a masked stranger was about to do us a power of good and take over the chip production - since then, not a sausage.

So my Hail Mary speculation is that we are running a parallel software licensing program in parallel with, or in advance of the IP licensing program. Given that this would be in the context of joint developments, the details would be commercial-in-confidence.

All this is derived from my reading of the tea leaves (the crystal ball is being fitted with Gen AI LLMs as we speak).

The Edge Box has sold out, so, with the hindsight of the development of TeNNs, I am reluctant to call it a mistake, or even a tactical error, but it would have been nice to have had more chips for sale.

Gosh - I didn't know I knew that many words!

Thanks Diogenese, that's a great post with a lot of insight.

I think another way of viewing the situation is that perhaps the company had no practical choice but to pursue an IP model (because of all the factors you have identified and one additional factor I would add - the capital required to produce the chips) .

The idea of a 'parallel software licensing program' has also occurred to me too. It seems unlikely that we are providing the software to partners without protections around the IP. Some may recall I asked Tony some time ago if there is a partnership agreement in place for these partnerships and he said yes. I'd be surprised if there weren't terms in the agreements effectively licensing, and at least regulating, the use of the IP. We can only hope that these arrangements turn into profitable ones in due course. I've always viewed these partnerships as JV's to develop the product.

Yours Yoda

 

[Bravo]

July 22 Deadline for ASCR Call for Papers: Workshop on Neuromorphic Computing for Science​

July 10, 2024 by staff Leave a Comment

July 10, 2024 — The U.S. Department of Energy’s Advanced Scientific Computing Research (ASCR) program has announced a Monday July 22 deadline (11:59 pm ET) for position papers for a workship on neuromorphic computing for science.
The workshop website can be found here. Notification of position acceptance will be issues on Friday, Aug. 2. For meeting technical questions, please contact: Todd Munson, Todd.Munson@science.doe.gov
The 2024 Workshop on Basic Research Needs for Neuromorphic Computing, to be held Thursday and Friday, Sept. 12-13 in the greater Washington, DC area, will inform and draft a set of grand challenges for advancing the field of neuromorphic computing and developing proof of principle neuromorphic circuits applicable for High Performance Computer (HPC) acceleration for scientific discovery, and brainstorm ideas needed for a successful, robust, and world leading basic research program.
​

Engineering novel neuromorphic computing systems with functionalities, capabilities, and energy efficiency similar to biological brains is one of the most exciting and challenging scientific endeavors of our time. This workshop aims to identify key research needs, challenges, and next steps necessary to develop biologically-realistic neuromorphic circuits primitives that capture the functionality of neural systems found in nature. Moreover, simulating neuromorphic computing primitives integrated into networks will be key to understanding their behavior at scale, particularly for those computing architectures where full-scale commercial fabrication is not yet readily accessible. Appropriate neuroscience datasets and metrics will have to be established to vet proposed neuromorphic circuits.

In the development of new circuits and methodologies for neuromorphic computing, it is critical that there is close collaboration among circuit designers, computer engineers, computational neuroscientists, and algorithms and simulation researchers. This workshop aims to bring together a diverse range of experts across three complementary technical areas.
Submit position paper to the technical areas below:

1. Neuroscience algorithms and translation to neuromorphic analog circuits
   This technical area is driven by the fundamental question “What are the key neuromorphic circuit primitives that are needed to capture the full functionality of critical biological computing mechanisms?”. The goal of the activities in this space is to understand what principles and circuit structures of brain organization and dynamics underpin its functionality and robustness capabilities and how these principles can be translated into functionally-equivalent neuromorphic circuits and systems that could be practically implemented (with available technology?). Ideas related to neuromorphic computing principles inspired from brain regions/functions (cortical, hippocampus, thalamus, sensing, motor control, etc.) are sought after. Topics related to neuromorphic approaches and emulations of small invertebrate brains are also of interest.
2. Technologies and prototyping of neuromorphic analog primitives
   This technical effort is driven by the fundamental question “What are the technologies needed to demonstrate and prototype key neuromorphic circuit primitives?” Ideas related to novel neuromorphic circuits based on new devices and designs, and new principles guided by neuroscience-inspired functionality are of interest. Ideas related to emerging analog technologies that provide orders of magnitude in performance, parallelism, energy efficiency, tunability range, temporal delays, etc., and that mimic the biological behavior and robustness of key primitives are welcomed. Also of interest are topics related to high neuromorphic connectivity capabilities, e.g. optoelectronic technologies and photonic interconnects.
3. Scalable integration for neuromorphic computing modeling
   The fundamental question driving this technical area is “What are the critical characteristics for effective large-scale simulation of neuromorphic circuits and systems?” New approaches are needed to create simulations of large-scale biologically-realistic neural networks, diverse synapse connectivity, and sophisticated network activity. Of interest are ideas related to novel methods to integrate and to scale up the simulation of the neuromorphic circuit primitives using high-performance computing in order to understand their interactions in the context of hundreds of millions of neurons and synapses. Also welcomed are novel methodologies for the efficient exploration of the large co-design space between neuromorphic algorithms and circuit technologies.

When discussing the technical idea and how it fits in the technical area(s) and the overall vision of the workshop, include a discussion on the benchmarks, metrics, and/or datasets requirements for neuromorphic computing for your proposed implementation.
Submission Guidelines
The structure for the ideal position paper may include several of the below themes:

1. Neuroscience-inspired computing principles
2. Translation to analog microelectronic circuits
3. Modeling and simulation approaches
4. Performance metrics, data requirements, and energy efficiency

The position paper should be an individual submission, one paper per investigator. The format is one page (plus one extra page for figures, captions and references only) with an 11-point font, submitted in a Word or PDF document. The primary theme should be mentioned during the submission, a secondary theme is optional.

Accepted position papers will be made public.
The following information is being collected during the submission process:

1. Author email
2. Author first name
3. Author last name
4. Author organization
5. Position paper title
6. Position paper theme(s)
7. Position paper abstract
8. Position paper file

Submissions will be reviewed by the workshop’s organizing committee using criteria of overall quality, relevance, likelihood of stimulating constructive discussion, and ability to contribute to an informative workshop report. Unique positions that are well presented and emphasize potentially transformative research directions will be given preference.
Workshop Organizers
Co-Chairs

• Gina Adam, George Washington University
• Garrett Kenyon, Los Alamos National Laboratory
• Thomas Potok, Oak Ridge National Laboratory

Organizing Committee

• Giorgio Ascoli, George Mason University
• Frances Chance, Sandia National Laboratory
• Yiran Chen, Duke University
• Joseph Friedman, University of Texas at Dallas
• Cory Merkel, Rochester Institute of Technology
• Maryam Parsa, George Mason University
• Midya Parto, University of Central Florida
• Catherine Schuman, University of Tennessee Knoxville
• Shinjae Yoo, Brookhaven National Laboratory
• Yuping Zeng, University of Delaware

July 22 Deadline for ASCR Call for Papers: Workshop on Neuromorphic Computing for Science

July 10, 2024 -- The U.S. Department of Energy's Advanced Scientific Computing Research (ASCR) program has announced a Monday July 22 deadline (11:59 pm - Read more from Inside HPC & AI News.

insidehpc.com