BRN Discussion Ongoing

[itsol4605]

Hi Bravo,

The latest Qualcomm patent I found indicates they are using a hynrid analog compute-in-memory neuron arrangement:

US2023297335A1 Hybrid Compute-in-Memory 20220315

View attachment 71695


The capacitor C is charged when both the switches S1 & S2 are closed.

US2023115373A1 ACCUMULATOR FOR DIGITAL COMPUTATION-IN-MEMORY ARCHITECTURES 20211013

a method for in-memory computation. The method generally includes: accumulating, via each digital counter of a plurality of digital counters, output signals on a respective column of multiple columns of a memory, wherein a plurality of memory cells are on each of the multiple columns, the plurality of memory cells storing multiple bits representing weights of a neural network, wherein the plurality of memory cells of each of the multiple columns correspond to different word-lines of the memory; adding, via an adder circuit, output signals of the plurality of digital counters; and accumulating, via an accumulator, output signals of the adder circuit.

This slmple principle should be a patent?

 

[Diogenese]

This slmple principle should be a patent?

As Ella says:

" 'tain't what you do;
It's the way that you do it."

 

[McHale]

The point is that he might not be up to date on the latest developments. Neither regarding BrainChip, nor Intel, because he no longer works there. I can easily imagine that internally at Intel he was already acting like a know-it-all. He’s just an Intel lohi fanboy who wants to talk down BrainChip. Additionally, maybe he’s a shorter? Who knows. His profile say “investor” maybe he is investing in an other company trying to down ramp openly? DH

Does the bloke actually know what he is talking about, doesn't look like it.

 

[yogi]

Here we go

 

[Schwale]

What point is David Wyatt actually trying to make?

Is he suggesting that BrainChip's technology shouldn't be classified as spiking neural network (SNN) because it's not spiking in nature according to him?

Is he accusing BrainChip of false advertising?

Is he trying to tarnish BrainChip's reputation stating that " the lack of details and open technical information should be a red flag"?

Honestly, how can this type of denigrating communication on the world's largest professional network be considered appropriate?

Ignorance!

 

[7für7]

Does the bloke actually know what he is talking about, doesn't look like it.

I think he knows very well what he is talking about. That makes it more worse I would say

 

[Pom down under]

Surely an Ai company can work out a way to save $250k a year using AI.

It wouldn't be hard. Have an AI model generate an automated response. It's basically what we get from Tony in every response!

Who needs AI to work that out you just sack the CEO

 

[Pom down under]

Just got the key to our new house today so hopefully the Mrs now in a good mood let’s me buy a few more shares next week

 

[7für7]

Just got the key to our new house today so hopefully the Mrs now in a good mood let’s me buy a few more shares next week View attachment 71711

I hope she doesn’t make your life hell for buying a house instead of a bunch of new shares next week.

 

[TheDrooben]

Just got the key to our new house today so hopefully the Mrs now in a good mood let’s me buy a few more shares next week View attachment 71711

Congrats Pom. Larry bought more today so they should be cheaper next week for you.....

Happy as Larry

 

[Humble Genius]

BrainChip's Akida Pico Brings LLMs to the Edge | ipXchange

BrainChip's Akida Pico Brings LLMs to the Edge At embedded world Exhibition&Conference in Texas, BrainChip ’s Todd Vierra unveiled the revolutionary Akida Pico—a neural processor bringing large language models (LLMs) to the edge! This breakthrough eliminates the need for cloud processing...

www.linkedin.com

 

[Humble Genius]

BrainChip's Akida Pico is a new chip design for AI devices that operates at milliwatt-level power consumption. | STEAM Computer: A Division of STEAM Executive

BrainChip's Akida Pico is a new chip design for AI devices that operates at milliwatt-level power consumption. This neuromorphic computing chip mimics the brain's spiking nature, making it highly energy-efficient and suitable for power-constrained devices like smartwatches and mobile phones. The...

www.linkedin.com

 

[Pom down under]

I hope she doesn’t make your life hell for buying a house instead of a bunch of new shares next week.

I had to buy the house before I could buy anymore shares but she’ll probably still moan that I’ll be buying more shares. Can never win

 

[Diogenese]

This slmple principle should be a patent?

Sorry itso,

The drawing is just one of several, that one illustrating the analog multiplication element. I chose this to illustrate the analog part of the neuron where the multiplication is accumulated in the capacitor.

The digital part includes a row of flip-flops to count the spikes from the corresponding row as shown, for example, in:


US2023115373A1 ACCUMULATOR FOR DIGITAL COMPUTATION-IN-MEMORY ARCHITECTURES 20211013

The actual invention is defined in the claims of the patent, but these are are often in arcane patentese.

I haven't looked into these patents in detail, but it looks like Qualcomm are using single-bit, or possibly 2-bit, analog multipliers which, together with the digital summation, would largely circumvent the problems with manufacturing repeatability (variations in the size/spacing of the capacitor plates). The error effect of the variations escalates exponentially with the number of bits per neuron.

 

[BrainShit]

BrainChip's Akida Pico Brings LLMs to the Edge | ipXchange

BrainChip's Akida Pico Brings LLMs to the Edge At embedded world Exhibition&Conference in Texas, BrainChip ’s Todd Vierra unveiled the revolutionary Akida Pico—a neural processor bringing large language models (LLMs) to the edge! This breakthrough eliminates the need for cloud processing...

www.linkedin.com

View attachment 71713

Did I assume it right? There will be a Pico board "available" or is in planning?

 

[Pom down under]

Not sure if posted?

Neuromorphic NPU Sips Power to Handle Edge Machine-Learning Models

BrainChip’s Akida Pico neural processing unit, which leverages spiking neural networks, targets low-power IoT and edge-computing devices.

www.electronicdesign.com

 

[Pom down under]

Lowest-Power AI Acceleration Co-Processor

Low-power coprocessor enables wearable AI & sensor-integrated applications across consumer, healthcare, IoT, defense, & wake-up technologies

www.electronicsforu.com

 

[Pom down under]

Promising

 

[Humble Genius]

#supplychain #transportation | Hanns-Christian Hanebeck

#supplychain #transportation Why are supply chains so difficult to manage? First of all, I probably need to clarify what I mean by difficult. Essentially, you have to trust a lot and you have little control. When you think about it, supply chain management is very different from other fields...

www.linkedin.com

 

[Humble Genius]

Even searching the keywords "AKIDO Pico", comes up with more exposure for BRN. I'm actually quite surprised at the amount of exposure we are now getting..... hope its a good sign.

Neuromorphic NPU Sips Power to Handle Edge Machine-Learning Models | Electronic Design

BrainChip #Akido Pico neural processing unit (#NPU), which leverages spiking neural networks (#SNNs), targets low-power #IoT and #edgecomputing devices. This video explains how it works ⬇️

www.linkedin.com

#edgecomputing #ai #iot #neuralnetworks #innovation #techtrends | SuperDataScience

🌟 Exciting times in edge computing! BrainChip's Akido Pico neural processing unit (NPU) leverages spiking neural networks (SNNs) to achieve ultra-low power consumption for IoT devices. Operating at microwatt levels, it serves as a co-processor for low-power microcontrollers, enabling AI...

www.linkedin.com