Hi Bravo,Report suggests BAE's February 2024 acquisition of Ball Aerospace will help position it for a role in the US Iron Dome defense system.
Intriguingly, Ball Aerospace have an established partnership with the AFRL collaborating on various research and development projects. In 2021, Ball Aerospace, in collaboration with Booz Allen Hamilton, secured a contract under AFRL's Trusted and Elastic Military Platforms and Electronic Warfare System Technologies (TEMPEST) program. This initiative focuses on developing cybersecurity tools to bolster the resilience of Air Force weapon systems against cyberattacks.
As per @manny100's post yesterday when he said regarding the recent podcast with Sean, "when talking about the about the US AFRL contract and the unnamed defence contractor BRN is working closely with, he made it clear that this creates other opportunities for BRN".
I wonder if other opportunities or application use cases that the AFRL are looking into might be cybersecurity related?
View attachment 77816
View attachment 77819
Ball Aerospace Acquisition Positions BAE For U.S. Iron Dome Work | Aviation Week Network
BAE Systems’ acquisition of Ball Aerospace will help position the company for a potential role on the proposed Iron Dome for America missile defense project.aviationweek.com
LinkedIn post 1 month ago...
View attachment 77817
And then, there's also this...
View attachment 77821
Booz Allen, Ball Aerospace Secure Potential $200M Contract for AFRL Cybersecurity Effort
The Air Force Research Laboratory in Ohio has awarded Booz Allen Hamilton and Ball Aerospace & Technologies contracts to develop cybersecurity tools designed to enhance the resilience of the service’s weapon systems against cyberattacks. In support of the Trusted and Elastic Military Platforms...potomacofficersclub.com
Sean in the podcast said the US AFRL was about TENNs. No mention of AKIDA.Hi Bravo,
Rob Whyte's reference to hybrid architecture could encompass Qualcomm's hybrid architecture where AI loads are shared between CPU, GPU, and NPU depnding on the type of load.
https://www.qualcomm.com/content/da...I-with-an-NPU-and-heterogeneous-computing.pdf
As previously mentioned, most generative AI use cases can be categorized into on-demand, sustained, or pervasive. For on-demand applications, latency is the KPI since users do not want to wait. When these applications use small models, the CPU is usually the right choice. When models get bigger (e.g., billions of parameters), the GPU and NPU tend to be more appropriate. For sustained and pervasive use cases, in which battery life is vital and power efficiency is the critical factor, the NPU is the best option.
What this suggests to me is that Qualcomm do not have the confidence that Hexagon can handle all use cases competently.
A rider on that is that TL mentioned that Akida is moving to 16 bit, which indicates higher precision is required for some use cases - so it's coming down to a real arm wrestle.
I wonder what applications could require such precision.
I doubt it's cybersecurity because of the limited vocabulary and syntax.
Could it be for AFRL micro-doppler radar? In MDR, you are looking for very small variations in the frequency of the reflected radar signal.
https://www.aumanufacturing.com.au/brainchips-human-like-ai-wins-it-us-air-force-radar-contract 20241210
The AFRL MDR article mentions TENNs, which can be implemented as an algorithm, but we have been told that Akida and TENNs are better together.
Late edition:
On the other hand, $1.8M is pretty mingy for developing a 16-bit version of Akida.
I reckon you would really enjoy this interview, ( if you haven't already seen it )
Palmer Luckey ( Anduril founder ) 32 years old.
Just took over $22 Billion IVAS contract from Microsoft 4 days ago.
Started Occulus when he was 19.
After watching this, I'm just praying that he's aware of BrainChip
Would be nice to have this bloke on side.
New regional sale manger Korea-
Check out this job at BrainChip :
https://www.linkedin.com/jobs/view/4155477814
Footware already?Sounds promising!
- Build and maintain strong relationships with key SOC customers, understanding their needs and providing tailored IP solution
That guy again… He had a crack at BrainChip a few months on LinkedIn. Same thing - querying whether we were neuromorphic and implying we were shysters. Tony Lewis replied from memory. Tony needs to contact him and put him straight. This guy could be doing a lot of damage if he his circulating doubt and mistruths.@Diogenese @uiux you guys might be able to help us understand on below which was posted on LinkedIn as one of the guy coment on BrB and the answer below got me confused.
Peter Sillan you bring up good points, and highlight the key gap to realizing this groundswell : a lack of available, usable, interoperable hashtag#neuromorphic Silicon
But are you sure BrainChip is neuromorphic? To date only their marketing material says so. In disclosures at NICE, ICONS last year their CTO opened the kimono a little, and conceded that they’re really just an Event-based General Matrix Engine (ie more CNN than SNN) pipeline - although not sure if he still has CTO role, he was a new hire trying to be more transparent.
To be clear : event-based is a forward step, but maybe one of the smallest pieces of the key Neuromorphic advances, but the published brainchip patents only show changes that are more incremental, than revolutionary. And even for that, they’ve had to “roll their own” TENN. So it’s clearly an uphill battle to do just this, as a one-off chip maker, solo solution.
But if intellectually honest, just event-based but still systolic array fmac engine, is not a big differentiation from today’s GPU/TPU/AI chip arch.
An independent view of the landscape I would refer you to Catherine Schuman’s excellent paper
Perhaps a weakness as I can be overly impressed by academic accolades but graduating in the top 5% at Williams College is no humdrum accomplishment as Williams is the highest ranked small college in the US....although those from Amherst might disagree.![]()
Brian Malone on LinkedIn: BrainChip’s IP for Targeting AI Applications at the Edge - EE Times Podcast
I thought that this was an excellent episode with the team at BrainChip. They discuss recent advances like TENNs (Temporal Event-based Neural Networks). I…www.linkedin.com
View attachment 77841
View attachment 77842
View attachment 77843