BRN Discussion Ongoing

Dhm

Regular
Hi @Dhm

Back a long time ago I did a lot of reading about accelerators. Probably at least 2 years now so some may have moved on a little but this is what I took away in a nutshell and in my very lay technophobe language which will likely cause the engineer among us to cringe - @Diogenese :

1. An accelerator is used to speed up data transmission from the sensor/source of the data acquisition to the processor.

2. The processor can be in the Cloud or it can just be in the room where they keep the main computer at an office or factory.

3. Traditionally an accelerator simply compressed all of the data in batches which reduces the bandwidth needed to transmit the data to the processor by sending multiple batches rather than one continuous stream.

4. This works OK in a factory as you can anticipate all the data you are trying to capture and might need to send and have sufficient bandwidth installed to cope with your closed system.

5. Out there on the web where you are sending your data over the 3,4 or 5G network you are in competition with others for the bandwidth and your batches of compressed data can be larger than the available bandwidth, notwithstanding being compressed, and it can still be forced to queue if there is a lot of traffic.

6. I do not understand the full technical reason but if you have lots of compressed data batches (packets) queuing up waiting for a chance to send they can actually become jammed (my word). Since I originally read about this problem I have read a couple of papers where solutions to the log jam were being proposed so not sure about the present situation.

This then leads to using AKIDA as an accelerator and its advantages:

1. AKIDA does not compress the data coming from the sensor.

2. AKIDA actually processes the data coming from the sensor to arrive at an actionable insight or to sort out the needed relevant data to make a decision about what action to take.

3. AKIDA can then either take action at the sensor or it can send that actionable insight to somewhere else as meta data.

4. AKIDA's huge advantage as an accelerator is therefore it is sending only tiny little packets of meta data which can reliably slot in to the bandwidth with ease.

5. AKIDA's huge advantage continues when the meta data reaches the cloud or processor where the action is to be taken because that computer does not have to wade through every single piece of data collected and sent to it and sort out what is relevant before taking action. AKIDA gives it just the relevant data so it can action that data almost instantaneously.

This ability or advantage can be critical in health, automotive, aeronautical, defence and space applications.

My opinion only DYOR
FF

AKIDA BALLISTA
My guess is that an AI accelerator is a circuit which performs basic computationally intensive classification functions in recognizing images or speech or other sensor type data faster than doing the job in software (CNN) on a CPU/GPU/cloud server. The CPU/GPU/cloud server can then perform some programmed function in response to the classification of the data.

AI accelerators in the past have been lumped with fixed model libraries which are difficult to update with additional images/sounds ...

Again in the past, AI accelerators used MAC (multiply accumulate) circuits of 8 bits or more.

Akida does data classification using compact model libraries adapted specifically for Akida using 1-bit to 4-bit "spikes", so Akida is capable of performing the function of an AI accelerator faster and more efficiently that previous AI accelerators ... and much more. [#### PS: The result of the classification by the AI Accelerator uses much less data than the input sensor data. ###]

On-chip one-shot learning makes Akida capable of updating its model library to improve accuracy. Such modifications can be shared via the internet to update other Akidas within a specific group or which use the same model library.

If Anastasi is talking about AI accelerators, I'm sure she understands Akida's capabilities in this field.

Someone needs to update Wiki:
https://en.wikipedia.org/wiki/AI_accelerator#Emergence_of_dedicated_AI_accelerator_ASICs

Emergence of dedicated AI accelerator ASICs[edit]​

While GPUs and FPGAs perform far better than CPUs for AI-related tasks, a factor of up to 10 in efficiency[34][35] may be gained with a more specific design, via an application-specific integrated circuit (ASIC).[citation needed] These accelerators employ strategies such as optimized memory use[citation needed] and the use of lower precision arithmetic to accelerate calculation and increase throughput of computation.[36][37] Some adopted low-precision floating-point formats used AI acceleration are half-precision and the bfloat16 floating-point format.[38][39][40][41][42][43][44] Companies such as Google, Qualcomm, Amazon, Apple, Facebook, AMD and Samsung are all designing their own AI ASICs.[45][46][47][48][49][50] Cerebras Systems has also built a dedicated AI accelerator based on the largest processor in the industry, the second-generation Wafer Scale Engine (WSE-2), to support deep learning workloads.[51][52]
My most humble thank you both @Diogenese and @Fact Finder.
 
  • Like
  • Love
  • Fire
Reactions: 13 users

wilzy123

Founding Member

View attachment 21657
Hey - just a tip - you can actually get the URL for said tweet by hitting share/send and then "copy URL"... then you can just paste that URL into your post here and it will natively place the tweet into your post (i.e. the actual tweet instead of just a screenshot of it). The upshot is that it is then more likely to be engaged with. (y)
 
  • Like
  • Love
  • Fire
Reactions: 11 users

Proga

Regular
Cant stop buying, managed to get a good parcel at $0.595, i hope thats the bottom as I cant afford to buy anymore 😳
Same, 60c and same (I'm not prepared to sell others to buy more). I've bought 4 days in a row now.

@FJ-215 most have been in here or the other joint for well over 12 months so will be in the green. I've been averaging up.
 
  • Like
  • Fire
Reactions: 9 users
Hey - just a tip - you can actually get the URL for said tweet by hitting share/send and then "copy URL"... then you can just paste that URL into your post here and it will natively place the tweet into your post (i.e. the actual tweet instead of just a screenshot of it). The upshot is that it is then more likely to be engaged with. (y)

yeah thanks @wilzy123 i normally do just that from my phone, but today off lappie and got thrown by different interface...... Cheers

i'll stick with phone tech !!
 
  • Love
  • Like
Reactions: 2 users

Quercuskid

Regular
Oh Well!
Now that we know you're all set, I'll ring Sean and tell him to take his foot off the brakes. 🤣


That would really be appreciated, thanks 😁 😜
 
  • Haha
  • Like
  • Fire
Reactions: 5 users

wilzy123

Founding Member
yeah thanks @wilzy123 i normally do just that from my phone, but today off lappie and got thrown by different interface...... Cheers

i'll stick with phone tech !!

Just hit this thing if using lappy :)

1668055877228.png
 
  • Like
  • Love
Reactions: 7 users

jk6199

Regular
  • Haha
  • Like
Reactions: 3 users

Vladsblood

Regular
Chimera yet another reference to Greek mythology
First Brainchip Akida
Second ARM ethos
Now Quadric Chimera
What is it with AI chips and Greek
It’s in the secret 🤫 sauce of Peter Van Der Madge’s making. That’s the connection
 
  • Love
  • Like
Reactions: 2 users

HopalongPetrovski

I'm Spartacus!
  • Like
Reactions: 1 users

TopCat

Regular
I noticed PainChek was listed as a partner on the Nviso page. It was on a page seemingly dedicated to Brainchip technology. What a great use of Akida , if indeed it is Akida, especially for the elderly who can’t communicate
31706DC0-24C9-4852-926F-F48A0CE62B14.jpeg
C3172AB6-7F49-4D11-9691-C2C6CD4BB9B6.jpeg


The first PainChek® clinical pain assessment in aged care was at Barossa Village in South Australia. Three years on, PainChek® has now established itself as the new standard in pain assessment and has expanded across the whole of Australia and around the globe.

PainChek® Universal is a medical device and has regulatory clearance in Australia, UK, European Union, Canada, Singapore and New Zealand.

PainChek® is now licensed in over 1,500 aged care facilities covering four regions of the world and more than 850,000 cumulative pain assessments have been conducted to date1. This has enabled better pain management for people who cannot reliably verbalise their pain – particularly those living with dementia and cognitive impairment.

Introducing the Universal Pain Assessment Solution​

As part of the next step in the evolution of pain assessment, PainChek® continues to innovate with the Universal Pain Assessment Solution – enabling best practice pain management for all people, everywhere.

The PainChek® Universal solution enables better pain management by:

      • Giving a voice to those who cannot reliably verbalise their pain, through a unique AI driven assessment tool
      • Document the pain for those who can self-report, and
      • Providing the right pain assessment solution for those whose ability to communicate fluctuates
All directly available at the point of care. Click here for more information

PainChek® is a fully digital healthcare solution and can be operated from up to 3m away from the person being assessed. This aligns with social distancing requirements, minimising infection risks during both the current COVID-19 crisis, and any future viral outbreaks which may impact the vulnerable.

Finally, PainChek® Universal helps to keep all your processes aligned as it integrates directly into Care Management Systems. Pain data is safely stored in one place from which powerful real-time reporting via PainChek® Analytics supports your critical clinical decision-making.

 
  • Like
  • Love
  • Fire
Reactions: 41 users

wilzy123

Founding Member
Last edited:
  • Haha
  • Love
  • Fire
Reactions: 5 users

Tony Coles

Regular
  • Like
  • Haha
Reactions: 4 users

chapman89

Founding Member
This is interesting although I’m not surprised.
A person by the name of Manish Pandey who works for Synopsys as Fellow, Vice President of R&D is also a adjunct professor at Carnegie Mellon University with Machine Learning, Systems and Computer Architecture.

Now we know that Brainchip has the “BrainChip University AI Accelerator Program which shares technical knowledge, promotes leading-edge discoveries and positions students to be next-generation technology innovators”

We also know that Synopsys is offering us in their Data Centre AI thanks to the Synopsys investor presentation back a few months ago, well it’s suspected they are as they were using a photo of Brainchip accelerator, which is hard to find even if you Google it, not to mention it’s a copyright issue if they weren’t working together and Synopsys has taken it upon themselves to just use somebody else’s photo of a product that’s not theirs.

Is this a coincidence? I think not.

39698564-33F2-4D3C-8D62-2FBB0C7C8DBE.jpeg
 
Last edited:
  • Like
  • Fire
  • Love
Reactions: 59 users

Dhm

Regular
His in retirement as everybody knows and maybe 🤔 the Mrs wanted a new kitchen! Or suffer the consequences of continuous headaches from the non stop 🛑 yapping. 😂
He’s chairman
 
  • Like
  • Haha
Reactions: 5 users

Sirod69

bavarian girl ;-)
The idea of #neuromorphic computing has now really caught up and we want to see how far we can go in this path. Let me invite you to the TCS Forum: Thought Leadership Conversations organised by Tata Consultancy Services on this topic - "Neuromorphic Computing for Transformation of the Industrial Future" where leaders from the academia and industry will take us on an enthralling journey through this evolving concept.

Our distinguished guests are Dr Mary Doborje, Dr. Zohreh Doborjeh, Dr. Priyadarshini Panda, Dr. Manan Suri, Anupam Singhal, Dr. Arpan Pal, and Vincent Ofrecio.

Ple
1668060198386.png
ase join us on 16th November to get a glimpse of the future. The registration link is - https://lnkd.in/gueZGy6u
 
  • Like
  • Fire
  • Love
Reactions: 24 users

HopalongPetrovski

I'm Spartacus!
  • Like
  • Haha
  • Fire
Reactions: 6 users
https://cdn-api.markitdigital.com/a...access_token=83ff96335c2d45a094df02a206a39ff4

What do we make of this? Certainly a justified reason.Only 125k shares out of a whole lot more.
Hi again @Dhm

I am glad you asked this question as I was just discussing this privately with @stan9614.

It is VERY important to remember what has gone before in Annual Reports and presentations to shareholders and for newer shareholders to go back and read these documents.

Brainchip advised shareholders that to be commercially successful they needed to employ the best people.

Brainchip advised shareholders that it did not have the cash to offer attractive salaries to entice the best people so it could only get the best people by offering a salary made up of cash and shares.

So todays announcement of the sale of 125,000 shares to pay tax is as significant as your employer paying you your salary and handing you a payslip on which all the deductions are noted one of which being the tax paid out of the companies cash reserves to the Australian Taxation Office on your behalf.

In other words in my opinion as these shares are part of his agreed salary the sale of 125,000 to pay tax on his salary is a non event and of no significance whatsoever.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Love
  • Fire
Reactions: 42 users
Top Bottom