BRN Discussion Ongoing

  • Like
  • Haha
Reactions: 4 users

Damo4

Regular

Ohh thanks for sharing rocket!

What do we think of the below stat?
Customer's designs in concept - IP or Dev kits??

1000005348.png
 
  • Like
  • Fire
  • Love
Reactions: 32 users

Tothemoon24

Top 20
This is interesting ; a YouTube review on the new neuromorphic super computer being built in Sydney.
Also a interesting comment about the review & a man named Elon

IMG_8167.jpeg



Latest thoughts on AI HARDWARE!

Neuromorphic computer chips. (Less power, higher efficiency, designed for neural nets and fast)

A bunch of wires understand the connection / relationship between each other.

(From Quantum perspective just a higher dimensional substrate abstraction layer allowing alignment to natural processes).

Physical representation of LLM’s. Solid option to run LLM’s in an integrated fashion. Software and hardware stack just tighter than current cpu, gpu options.

LLM’s are number / values which the contextual relationship between makes powerful. Just like your body and mind. “Compression is key”

1 gram of human dna contains a potential of 455 EXABYTES.

Neuromorphic is a valid next step option instead of quantum for chips. Been tracking all chips ideas and will have more available data on these.

Probably more likely than quantum near term or silicon variants like those of etch. Etc...



IMG_8168.jpeg

IMG_8169.jpeg
 
  • Like
  • Fire
  • Wow
Reactions: 11 users

IloveLamp

Top 20
This is interesting ; a YouTube review on the new neuromorphic super computer being built in Sydney.
Also a interesting comment about the review & a man named Elon

View attachment 54195


Latest thoughts on AI HARDWARE!

Neuromorphic computer chips. (Less power, higher efficiency, designed for neural nets and fast)

A bunch of wires understand the connection / relationship between each other.

(From Quantum perspective just a higher dimensional substrate abstraction layer allowing alignment to natural processes).

Physical representation of LLM’s. Solid option to run LLM’s in an integrated fashion. Software and hardware stack just tighter than current cpu, gpu options.

LLM’s are number / values which the contextual relationship between makes powerful. Just like your body and mind. “Compression is key”

1 gram of human dna contains a potential of 455 EXABYTES.

Neuromorphic is a valid next step option instead of quantum for chips. Been tracking all chips ideas and will have more available data on these.

Probably more likely than quantum near term or silicon variants like those of etch. Etc...



View attachment 54196
View attachment 54198

Ask him how he knows @Tothemoon24

Great find
 
  • Like
Reactions: 4 users
  • Like
  • Haha
  • Love
Reactions: 8 users

MDhere

Regular
The more I read, reread, look back and listen to old interviews and come back and apply all these learnings to the present including the recent CES2024 and piecing things together over the last 2.5years, my mind is blown away that when Rob Telson said at the end of an interview in July 2021 that this is just the tip of the iceberg, he got that comment DEAD RIGHT.

Sure the financials is not showing fireworks yet but nor was Tesla and other big companies starting out. But what each of these companies have in common with Brainchip is Vision and Expertise to be successful.

I shake my head when I see the crap on crapper and some of the suspect weeds on here and I know for a FACT that they are either shitty try hard negative manipulators or just too plain dumb to not comprehend anything that Brainchip has ever published or not understood anything that they have listened to.

An invaluable interview for example 13 July 2021 discoposse interviewing Rob Telson.

I know its some time back but and its just over an hour but its one hell of a fantastic interview!

Rob said in 2021 its just the tip of the iceberg which we all knows the meaning of that.

I will summarise it soon because listening back to this makes alot of sense to where we are now and I believe the ignition lighter is about to fire up the rocket ! 🗼🔥
 
Last edited:
  • Like
  • Love
  • Fire
Reactions: 75 users

Tothemoon24

Top 20
  • Like
  • Haha
Reactions: 11 users
Some good general observations from CES24 of where AI is at and where it could be heading:


"But this year’s announcements hit differently from previous buzzy developments, such as the Metaverse or adding voice assistant technology to appliances. That’s because nearly every company appeared to be on the same page in 2024.

“It was an almost unanimous tethering to the AI theme … because it has infinite possibilities and a wide range of applications,” said Dipanjan Chatterjee, an analyst at market research firm Forrester."

"It’s also part of a larger shift happening at CES, from technology powering experiences to experiences powered by technology. To help further that change, chipmakers Nvidia and AMD unveiled new processors that will help run the next-generation of AI products.

Jitesh Ubrani, an analyst at market research firm IDC, agreed the chatter around CES felt unique this year as companies had a general understanding of “how ubiquitous and seamless AI will be in the coming years.”

“While many use cases are still unknown, what we do know is that no one wants to be left behind, so they’re starting to invest even before the products and use cases are fully fleshed out,” he said."

"He expects AI to dominate the tech world conversation not only throughout the rest of the year but beyond."

AI future is looking bright 😊
 
  • Like
  • Love
  • Fire
Reactions: 24 users

IloveLamp

Top 20
Hi Lampy , I’ve just put a little bait in the water if he bites I’ll push on with that question
well-ok-then.gif
giphy (13).gif
 
  • Haha
Reactions: 9 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
  • Haha
  • Like
  • Fire
Reactions: 18 users

Diogenese

Top 20
The passage straight after your highlighted section, is also interesting.

"while the use of analog or mixed-signal neuromorphic hardware seems out of scope at the moment due to their intrinsic variability. Hence, we suggest to focus on advanced digital systems such as SpiNNaker2 (Yan et al., 2021) or Loihi2 (Orchard et al., 2021) to further explore neuromorphic hardware for automotive radar processing and automated driving in general"

The references to both mentioned digital solutions, are both from 2021.

We know AKIDA, is not only superior, to both SpiNNaker2 and Loihi2 (because I said so DH 🙄..) but the latter, isn't even commercially available..
Hi DB,


Here's some information on SpiNNaker2.

It uses about 150 ARM M4F cores and does MAC maths.

About 0.75 W per processing element, so about 110 W potential peak consumption, but would run at much lower power due to sparsity and power saving techniques - but certainly not an edge device, although they do produce a robotics version (maybe mains powered?)

SpiNNaker2: Next Level Thinking - Research Articles - Research Collaboration and Enablement - Arm Community

... SpiNNaker2 started with the brain modeling fundamentals of its predecessor. It builds on this with the ability to support conventional neural networks for AI applications, as well as hybrid systems that combine neural networks of different styles, including spiking neural networks (SNN) and deep neural networks (DNN). With 10 million cores, fitting into 16 server racks, it’s a key enabler for real-time AI at massive data rates.


https://niceworkshop.org/wp-content/uploads/2018/05/2-27-SHoppner-SpiNNaker2.pdf
1705190431582.png
 
  • Like
  • Fire
  • Love
Reactions: 12 users

IloveLamp

Top 20
2min 20sec

Interesting choice of words.....

Also, PANASONIC is partnered with more than one of our partners / licensees

 
Last edited:
  • Like
  • Fire
  • Love
Reactions: 9 users

Terroni2105

Founding Member
  • Like
  • Love
  • Fire
Reactions: 43 users

hotty4040

Regular
This is interesting ; a YouTube review on the new neuromorphic super computer being built in Sydney.
Also a interesting comment about the review & a man named Elon

View attachment 54195


Latest thoughts on AI HARDWARE!

Neuromorphic computer chips. (Less power, higher efficiency, designed for neural nets and fast)

A bunch of wires understand the connection / relationship between each other.

(From Quantum perspective just a higher dimensional substrate abstraction layer allowing alignment to natural processes).

Physical representation of LLM’s. Solid option to run LLM’s in an integrated fashion. Software and hardware stack just tighter than current cpu, gpu options.

LLM’s are number / values which the contextual relationship between makes powerful. Just like your body and mind. “Compression is key”

1 gram of human dna contains a potential of 455 EXABYTES.

Neuromorphic is a valid next step option instead of quantum for chips. Been tracking all chips ideas and will have more available data on these.

Probably more likely than quantum near term or silicon variants like those of etch. Etc...



View attachment 54196
View attachment 54198


Well, this is all very interesting, if only, I could make this info compute, in my own tiny Brainchip, that resides in my head, ( which I sometimes loose on occasion's ).....

I'm thinking ( Supercalafragilisticexpialidotious )

I've always been somewhat indecided about what that word meant, but now, I'm not so sure. Maybe this has some answers, to consider Ttm. I need to concentrate my mind on this subject matter, ( edgy subject matter) a lot more IMO.

Truely, ( head over heels ) or ( head in the cloud ), not toooo sure which one, or maybe ( both ) about this new info, re : New Supercomputer being developed.

Akida Ballista >>>>> Will it have AKIDA ( 'S ) INSIDE at all, by chance, anyone KNOW ? <<<<<

"HOPE SO"

Would be worth a real Anne, if it did !!






Hotty...
 
  • Like
  • Love
  • Haha
Reactions: 7 users

Tothemoon24

Top 20

Unfortunately paper is behind a paywall , combination of neuromorphic & drones works for me​

IMG_8175.jpeg





Excited to share that our work "EV-Planner: Energy-Efficient Robot Navigation via Event-Based Physics-Guided Neuromorphic Planner" is now available as an Early Access article in IEEE Robotics and Automation Society's Robotics and Automation Letters. The work done in collaboration with Rohan Kumar Manna and our advisor Prof. Kaushik Roy from Purdue University explores the frontier of vision-based autonomous navigation using Neuromorphic Vision (through novel DVS Sensors and Spiking Neural Networks) along with Physics-based AI. This enables energy-efficient trajectory planning with obstacle avoidance capabilities. The entire setup is implemented using Robot Operating System (ROS) and tested in the Gazebo simulation environment on a Parrot Bebop2 drone.

Code available 🔐 : https://lnkd.in/ghjPpES4
Video Link: https://lnkd.in/gxKhqhGm
ArxiV Link: https://lnkd.in/gycyrnCE

EV-Planner: Energy-Efficient Robot Navigation via Event-Based Physics-Guided Neuromorphic Planner​

Publisher: IEEE
Cite This


Sourav Sanyal; Rohan Kumar Manna; Kaushik Roy
Sign In or Purchase


Abstract:Vision-based object tracking is an essential precursor to performing autonomous aerial navigation in order to avoid obstacles. Biologically inspired neuromorphic event cameras are emerging as a powerful alternative to frame-based cameras, due to their ability to asynchronously detect varying intensities (even in poor lighting conditions), high dynamic range, and robustness to motion blur. Spiking neural networks (SNNs) have gained traction for processing events asynchronously in an energy-efficient manner. On the other hand, physics-based artificial intelligence (AI) has gained prominence recently, as they enable embedding system knowledge via physical modeling inside traditional analog neural networks (ANNs). In this letter, we present an event-based physics-guided neuromorphic planner (EV-Planner) to perform obstacle avoidance using neuromorphic event cameras and physics-b
 
Last edited:
  • Like
  • Fire
Reactions: 7 users
This is interesting ; a YouTube review on the new neuromorphic super computer being built in Sydney.
Also a interesting comment about the review & a man named Elon

View attachment 54195


Latest thoughts on AI HARDWARE!

Neuromorphic computer chips. (Less power, higher efficiency, designed for neural nets and fast)

A bunch of wires understand the connection / relationship between each other.

(From Quantum perspective just a higher dimensional substrate abstraction layer allowing alignment to natural processes).

Physical representation of LLM’s. Solid option to run LLM’s in an integrated fashion. Software and hardware stack just tighter than current cpu, gpu options.

LLM’s are number / values which the contextual relationship between makes powerful. Just like your body and mind. “Compression is key”

1 gram of human dna contains a potential of 455 EXABYTES.

Neuromorphic is a valid next step option instead of quantum for chips. Been tracking all chips ideas and will have more available data on these.

Probably more likely than quantum near term or silicon variants like those of etch. Etc...



View attachment 54196
View attachment 54198

The way she is describing neuromorphic computing (although she does mention TrueNorth) it sounds like she is completely oblivious, to the existence of AKIDA? 🤔
 
  • Like
Reactions: 2 users
Ohh thanks for sharing rocket!

What do we think of the below stat?
Customer's designs in concept - IP or Dev kits??

View attachment 54197
Isn't it the same thing? (the Dev kits allow that, with our IP?).

It looks like 8 customer chip designs, with our IP included, in concept (late stage) of development.
Edit - "concept" by definition, means early stage.
Even the Company's blurb (on AKD1000) says this.
"Each of our tiered programs brings you from concept to working prototype with varying levels of model complexity and sensor integration. In addition, our AI experts provide training and support to make the process efficient and smart"


These would be major OEM type customers.

It may be worth, seeing if we can get further clarification, of the meaning of that point (the 8) from the Company..
 
Last edited:
  • Like
Reactions: 7 users

Damo4

Regular
Isn't it the same thing? (the Dev kits allow that, with our IP?).

It looks like 8 customer chip designs, with our IP included, in concept (late stage) of development.

These would be major OEM type customers.
Oh for sure, one is simply a precursor to the other.
I've been thinking about this a lot today, especially about the choice of wording.
My biggest questions are:

1. How do they define a "customer"? (does it exclude collaborations/partners)
2. How many customers make up the 8 products?
3. If a customer is deemed as paying for IP (rather than for engineering services), does that mean Megachips is servicing many products with Akida with each node?
4. Are any known connections actually customers of our customers?
 
  • Like
  • Thinking
  • Love
Reactions: 7 users

Tothemoon24

Top 20
IMG_8177.jpeg



🥁 It's official: "Edge AI Movement" stole the show at this year's CES!

What gives? Because it greatly embodies the popular idealism of "AI Everywhere. For Everyone."🌐

Edge AI gives wings (of #Inference) to thin clients, puts 32-bit MCUs on steroids, and more importantly, translates directly into a new market for tens of billions of affordable (<$10 each) AI/ML-capable chips (e.g., NPUs); in contrast to millions of exorbitant ($10k-$100k each) DPUs/GPUs/DLAs currently required for implementing Cloud AI.

📋 Enclosed is a list of 24 leading semiconductor startups in the field of Edge AI (in alphabetical order; Q1 ’24 Edition):

🍦Alif Semiconductor
🍦AONDevices, Inc.
🍦Axelera AI
🍦Blaize
🍦BrainChip
🍦DEEPX
🍦Edge Impulse
🍦Efinix, Inc.
🍦Esperanto Technologies, Inc
🍦Eta Compute
🍦Expedera Inc.
🍦Flex Logix Technologies, Inc.
🍦GreenWaves Technologies
🍦HALIO, Inc.
🍦INUITIVE
🍦Kinara, Inc.
🍦Mythic
🍦Perceive
🍦Quadric
🍦Roviero Inc.
🍦SiMa.ai
🍦Syntiant Corp.
🍦Tenstorrent Inc.
🍦Untether AI

⚾ Knowing when to pivot is an essential survival skill for every startup. More than one of the honorees above have pivoted from an IC model to an IP one (and vice versa), given this everchanging and increasingly competitive market.

🎲 6 more notables in this esteemed cohort of Edge AI chip startups:

🍭Ceremorphic, Inc.
🍭Condor Computing Corporation
🍭Edgecortix
🍭Fabric Cryptography
🍭MIPS (Not a startup per se but you get the idea🖖)
🍭Rain AI

In closing, AI implemented near/on/at the Edge—in a distributed or, in some cases, decentralized fashion—offers a perfect complement to a largely centralized, Cloud-based AI, while injecting new energy and life (with neural networks) to an arguably lukewarm 5G/IoT market. Not only #semiconductor #startups but also megacorporations have taken notice; from #AMD to #AnalogDevices, from #Intel to #Infineon, from #Nvidia to #NXP, #Google to #Qualcomm, and the list goes on... Exciting times ahead!🏇

📝Notes:
#AI: Artificial Intelligence
#IoT: Internet of Things
#MCU: Microcontroller Unit
#ML: Machine Learning
#NPU: Neural Processing Unit
#DPU: Data Processing Unit
#GPU: Graphics Processing Unit
#DLA: Deep-Learning Accelerator
#IC: Integrated Circuit
#IP: Intellectual Property

Additional reading:
🏷️Xnor.ai (acquired by Apple): https://lnkd.in/dsT2ZuG9
🏷️NeuRRAM by Stanford University: https://lnkd.in/gDGD2TRK
🏷️Edge AI and Vision Alliance (Blog): https://lnkd.in/gBZHWMj7
🏷️Plumerai (Blog): https://blog.plumerai.com
🏷️Deloitte (TMT Predictions 2020): https://lnkd.in/g_a3XJDh
🏷️Notes on #CES2023: https://lnkd.in/gDg2x32n
🏷️Notes on #CES2022: https://lnkd.in/g7vsjNCr

➟To be continued.

#SemiconductorIndustry #Semiconductors #EdgeAI #EdgeComputing #EdgeSecurity #EdgeNetwork #EdgetoCloud #CES2024 #CIM #5G #SiP #HeterogeneousIntegration #AdvancedPackaging #AIIoT #DistributedComputing #Blockchain #CloudAI #CloudComputing

IMG_8176.jpeg
 
  • Like
  • Love
  • Fire
Reactions: 33 users
Top Bottom