Peter Sillan

[BrainChip @BrainChip_inc]

Most wearables don't die from sensing too much. They die from calling the cloud too often.

AKD1500 runs fall detection on-device instead. Link below.
https://t.co/avEqpKcGOS

[MIAM @MIAM_Schwan]

250 milliwatts. That's the whole story of the AKD1500 M.2 Card.
Compact M.2 2230 form factor.
AKD1500 Edge AI co-processor.
800 Effective GOPS (INT4).
No fan, no heatsink needed to run it.
https://t.co/A3RErN7f6k
@BrainChip_inc #Akida

[MIAM @MIAM_Schwan]

@elonmusk Before such a statement, I would recommend checking on BrainChip's #Akida.

[MIAM @MIAM_Schwan]

Neuromorphic solar edge AI for sustainable wildfire detection
https://t.co/wZmOPLH9yj
"BrainChip Akida enables 4,200 patrol hrs/year, 3× longer than CPU-based systems"

These days, this topic is more relevant than ever!
@BrainChip_inc

[Peter Sillan @sillan421]

@MIAM_Schwan @BrainChip_inc Good info but dropped the ball on the use of Akida in the Data center.

[Kevin D. Johnson | Heterogeneous Compute Ontology @computeontology]

Another important article I wrote earlier in July that I'm making available on X, the PDF is also linked at the bottom of the article for download if you prefer that format. https://t.co/J4SLh1MVGk

[MIAM @MIAM_Schwan]

They said neuromorphic was stuck at the edge. This put it in the data center.
https://t.co/islqJa3VEq
Seventy demonstrations in seven months, across twelve different silicon architectures.That is not enthusiasm.That is a body of work! @computeontology
@BrainChip_inc #Akida #IBM

[BrainChip @BrainChip_inc]

AKD1500 is now discoverable and design-ready inside Supplyframe's schematic workflow. What that means for eval time is worth a read.

https://t.co/zXi9XDUhEW

[Peter Sillan @sillan421]

@computeontology @BrainChip_inc Great work!! Kevin. A game changer.

[Kevin D. Johnson | Heterogeneous Compute Ontology @computeontology]

Demo #70, Akida via vLLM, Supersized. IBM Spectrum Symphony, GPFS, and Akida chips (and sims) work over Dallas, Pittsburgh, and DC. Policy defines where the workload runs, IBM Cloud or on-prem. Dallas fills→spills to Pittsburgh's 10 BrainChip Akida chips→then to DC.

[Kevin D. Johnson | Heterogeneous Compute Ontology @computeontology]

I think it would be fun to do more time-related demos, Symphony and TENNS-PLEIADES with GPFS and Akida. For now, I just mess around with ESP32 S3 V4 firmware on occasion when I have a free moment. @geerlingguy thought you might like this.

[MIAM @MIAM_Schwan]

I love "categorically different" 😁.
#Akida is superior to all other neuromorphic chips and the only one available in the USA.
@BrainChip_inc

[Kevin D. Johnson | Heterogeneous Compute Ontology @computeontology]

Akida learns a new class from one example. I built the shared learning substrate with an FPGA. 1,024 spikes on the train, zero on 1,679 empty frames.The fabric knows when. The chip knows what. Add more chips → different questions on the same event. One record. No merging.

[Kevin D. Johnson | Heterogeneous Compute Ontology @computeontology]

The power of this may not be immediately obvious. 50,000 facts is a substantial share of a field's working reference knowledge, answered in 131 ms on sub-watt chips. K3 doesn't know your data. It knows your field.

Every piece already runs: the Akida vLLM plugin, a Symphony semantic router, extractive RAG over your own corpus, content routed through the chips. All from previous demos I've done. What's left is wiring them together, your data and the field's data joined, going to the cloud only on the rare occasions neither can answer. That's Sovereign AI.

And 50,000 isn't a ceiling. The number came from early experimentation, what might conceivably fit on one chip. Symphony scales horizontally, so sharding across more chips takes it much higher. Domain knowledge gets very deep, very wide, or both.

[Kevin D. Johnson | Heterogeneous Compute Ontology @computeontology]

Milliwatts, not gigawatts. Take a look at BrainChip's Akida. Intelligence doesn't have to be expensive.

[Kevin D. Johnson | Heterogeneous Compute Ontology @computeontology]

Hey, everyone. Most of my demos and writing are on my website or on LinkedIn, but I've decided to add X to make my work more accessible to a wider audience. So, here we go. Feel free to follow and I'll likely follow in return.
https://t.co/GXhgy12mus

[Peter Sillan @sillan421]

@MIAM_Schwan @BrainChip_inc All this great news and people are selling at 14 cents on the ASX

[Peter Sillan @sillan421]

@DuckStockMan There is so much positive info regarding Brainchip and Neuromorphic AI it's unbelievable the stock is only 15cents on the ASX.

[Peter Sillan @sillan421]

@PaceBeyerBeware @UziCryptoo Texas has no state tax

[Peter Sillan @sillan421]

@JanudaX Should take a look at Brainchip holdings Akida Neuromorphic chip and IP.