BRN Discussion Ongoing

[butcherano]

Good summary....thanks for that @Diogenese....

 

[Jefwilto]

WOOHOO, ................................ PATIENT ANNOUNCEMENT

It was always coming.

But to think that the share price would magically just shoot up, well, you'd be wrong.

The markets understanding of the significance of this, and in fact, every granted patent that Peter and the team have won is never really appreciated.

The line is drawn in the sand, we can continue growing with no major roadblocks, while the mob have to fumble around looking for a new way forward.

Close on 20 years research comes with a price, and Peter being our founder, has well and truly earnt these Patents with the help of Anil and others in recent years, there was never going to be any shortcut, only years of almost obsessive research, determined never to give in.

Each and every one of these Patents carries a weight that none of us could truly ever understand.

AKD 2.0 IP We all await your arrival.

Tech x

Yes Tech,its always good to see another Patent granted,i do believe we are going to get another announcement of a Patent which will be granted soon i think,so we can look forward to another announcement soon i hope

 

[chapman89]

The International Centre For Neuromorphic Systems at Western Sydney Uni is holding a presentation on what neuromorphic computing is with demonstrations and you can also meet the team and ask more questions, this could be very interesting… will they be demonstrating with akida?

Register here, it’s also online if you can’t attend-

Wednesday, 2 November

www.westernsydney.edu.au

 

[Fact Finder]

I particularly like the following extract from the latest patent grant as it makes clear that Peter van der Made and Anil Mankar have an AKIDA event horizon that stretches as far and further than the eye can see:

“BACKGROUND
[0003] It has long been a goal for artificial neural networks to replicate the function of the biological neural network (the brain), with limited success. Brute force hardware approaches to the design of an artificial neural network have been cumbersome and inadequate, and fall far (short) of the desired goal of replicating the functionality of the human brain. Thus, a need exists for an approach that enables an autonomous, reconfigurable spiking neural network to be realized that can scale to very large networks, yet fit on a chip while rapidly making inferences from a wide variety of possible input data and/or sensor sources.”

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Dhm]

I've had a look at the USPTO examination report and BrainChip's response.

The argument centered mainly on Intel's US201900429209, and in particular the method of reconfiguring the nodes. Our attorneys pointed out that the Intel system uses a router, whereas our configuration is carried out using information stored in memory.

As Ella would say: " 'tain't what you do - it's the way that you do it."

On that basis, it's unlikely that our patent would cause Intel any embarrassment, not least because they have the earlier priority.

View attachment 18630

[0022] ... processor 100 includes a plurality of network elements 102 arranged in a grid network and coupled to each other with bi-directional links. However, an NoC in accordance with various embodiments of the present disclosure may be applied to any suitable network topologies (e.g., a hierarchical network or a ring network), sizes, bus widths, and processes. In the embodiment depicted, each network element 102 includes a router 104 and a core 108 (which in some embodiments may be a neuro-synaptic core block that implements neurons and/or synapses of a neural network), however in other embodiments, multiple cores from different network elements 102 may share a single router 104 . The routers 104 may be communicatively linked with one another in a network, such as a packet-switched network and/or a circuit-switched network, thus enabling communication between components (such as cores, storage elements, or other logic blocks) of the NoC that are connected to the routers. In the embodiment depicted, each router 104 is communicatively coupled to its own core 108 . In various embodiments, each router 104 may be communicatively coupled to multiple cores 108


BrainChip US11468299 = US2020143229A1 SPIKING NEURAL NETWORK

View attachment 18632

[0026] FIG. 17 illustrates configuration registers comprising a scan chain that define a configuration and connectivity of each spiking neuron circuit and each layer of spiking neuron circuits, according to some embodiments.

Note that configuration is a pre-SNN signal processing step, and thus does not imped the actual function of the SNN in identifying/classifying input data.

@Diogenese for those of us more neuromorphically challenged (at least me) can you explain why this patent is so important? Is it related to Akida 2000? Does it close a gap that a competitor may have otherwise taken advantage of?

 

[Newk R]

Good day. Now all we need is Mitchell to the Pies and my day will be complete.

 

[Fact Finder]

Blind Freddie asked me to link this article with the above post. When I asked why he said it was obvious:

“MegaChips enters US for edge AI chips market​

POSTED ON MAY 2, 2022 UPDATED ON MAY 18, 2022

Osaka, Japan-based ASIC provider, MegaChips Corp. (Japan) has announced its entry into the US market. MegaChips is one of the world’s leading custom ASIC providers for consumer, telecom/network, industrial and automotive apps. Headquartered in Japan, it has offices in Silicon Valley and Taiwan. MegaChips is ISO9001 certified and ensures the highest levels of intellectual property security.

Douglas Fairbairn.
Right on time!
Looking at the announcement, has Megachips made a late entry into the global edge AI chips market? Douglas Fairbairn, Director of Business Development for MegaChips LSI USA Corp., said: “From our perspective, we’re not late for production, but right on time. There have been a number of companies entering this market from an IP point of view. However, those companies, in general, are still gaining traction and are not yet at the volume production stage, which is where we’re most interested in participating. We see a lot of market potential for volume ahead of us. There are still areas that might be a good early entry point for us to take advantage of beginning in 2022, and ramping up in 2023 and 2024.”



There are already a couple of areas where volume has been attained, such as the data center market, which is dominated by Nvidia and Intel. However, MegaChips is not addressing that market.

So, how is the AI chip industry positioned to address emerging cases for integrated processors, etc.? By integrated processors, things like RISC-V, or ARM, that are being embedded in other chips, or in SoCs. If that’s the case, the AI chip industry is very well positioned to take advantage of that emerging trend, which has been underway for some time.

What people have found is that the existing processors from ARM or RISC-V do not address the power performance requirements of the AI industry. There are some low-end cases that can be handled with software on these embedded processors. In general, people are looking for either accelerators to pair with those processors, or completely new processors, that would replace the embedded processor and AI into a much higher performance functionality.

In this case, MegaChips’ partner, BrainChip, is an example of an accelerator that would be combined with the existing embedded processors. In the case of its other IP partner, Quadric, they could be either used as an accelerator, or even supersede the need for an embedded processor.”

I just worked out why Blind Freddie was so dismissive of my question. It was because of this statement by Douglas Fairbairn Director of Business Development for MegaChips USA:

“What people have found is that the existing processors from ARM or RISC-V do not address the power performance requirements of the AI industry.”

It of course makes clear why ARM and SiFive both jumped virtually together to announce partnerships with Brainchip.

It was so they could “address the power performance requirements of the Ai industry.”

Now I get it.

Blind Freddie shakes his head and walks away.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Xhosa12345]

Good day. Now all we need is Mitchell to the Pies and my day will be complete.

dont worry about mitchell.....

 

[Fullmoonfever]

See a Prophesee tweet end Aug re the Nat Uni Singapore and their robotic skin project.

This is run by Benjamin Tee and suggest is part of the Tacniq product offering which mentioned being able to run on Akida.

Benjamin is a co-founder of Tacniq as a spinoff from NUS.

So a mention of Akida with Prophesee sensors via ACES way back when I first posted on this back in Feb.

Hopefully, can see some additional traction with our Prophesee partner confirmation mid year.

Tac-01 Sensors | TacniQ

Our Tac-01 sensors are ideal for a wide range of use cases, from raw data collection to higher order tactile interpretations.

tacniq.ai

We are developing sensory integrated artificial brain system that mimics biological neural networks, which can run on a power-efficient neuromorphic processor, such as Intel’s Loihi chip and Brainchip’s Akida neural processor. This novel system integrates ACES and vision sensors, equipping robots with the ability to draw accurate conclusions about the objects they are grasping based on the data captured by the sensors in real-time - while operating at a power level efficient enough to be deployed directly inside the robot.

stmnist | TEE Research Group

www.benjamintee.com

 

[Diogenese]

@Diogenese for those of us more neuromorphically challenged (at least me) can you explain why this patent is so important? Is it related to Akida 2000? Does it close a gap that a competitor may have otherwise taken advantage of?

Hi Dhm,

To determine the purpose of a patent, it is often useful to look at the Background of the Invention at the start of the specification ...

[0003] It has long been a goal for artificial neural networks to replicate the function of the biological neural network (the brain), with limited success. Brute force hardware approaches to the design of an artificial neural network have been cumbersome and inadequate, and fall far shown of the desired goal of replicating the functionality of the human brain. Thus, a need exists for an approach that enables an autonomous, reconfigurable spiking neural network to be realized that can scale to very large networks, yet fit on a chip while rapidly making inferences from a wide variety of possible input data and/or sensor sources.

... but sometimes this only contains motherhood statements.


You can look at the claims, chiefly claim 1 to see what the invention is:

1 . A neuromorphic integrated circuit, comprising:
a spike converter configured to generate spikes from input data;
a reconfigurable neuron fabric comprising a neural processor comprising a plurality of spiking neuron circuits configured to perform a task based on the spikes and a neural network configuration;
a memory comprising the neural network configuration, wherein
the neural network configuration comprises a potential array and a plurality of synapses,
the neural network configuration defines connections between the plurality of spiking neuron circuits and the plurality of synapses,
the potential array comprising membrane potential values for the plurality of spiking neuron circuits, and
the plurality of synapses having corresponding synaptic weights; and
a processor configured to modify the neural network configuration based on a configuration file.

It is also worthwhile looking at other independent claims (claims that are not appended to another claim).

17 . A method, comprising:
receiving, at a spiking neuron circuit, a set of spike bits corresponding to a set synapses associated with the spiking neuron circuit;
applying, at the spiking neuron circuit, a first logical AND function to a first spike bit in the set of spike bits and a first synaptic weight of a first synapse in the set of synapses, wherein the first synapse corresponds to the first spike bit;
incrementing, at the spiking neuron circuit, a membrane potential value associated with the spiking neuron circuit based on the applying;
determining, at a neural processor, that the membrane potential value associated with the spiking neuron circuit reached a learning threshold value associated with the spiking neuron circuit; and
performing, at the neural processor, a Spike Time Dependent Plasticity (STDP) learning function based on the determination that the membrane potential value of the spiking neuron circuit reached the learning threshold value associated with the spiking neuron circuit.

So, per claim 1, this patent is for a circuit arrangement enabling a processor to configure the synaptic connexions between neurons, and to allocate weights to the synapses on the basis of a configuration file. The number of layers, the number of neurons in each layer, and the synaptic connexions between the neurons are defined in the configuration file.

The configuration file can be changed to suit the particular application. This gives Akida great flexibility as to the tasks it can be used to perform simply by changing the configuration file.

There can be different configuration files for camera data, DVS data, LiDaR data, radar data sound data, smell, touch, taste, ...

Claim 17 defines the learning process using STDP, so this is a crucially important patent. In fact, we seem to have gotten away with 2 inventions for the price of 1.

Long term readers may recall that I went through a stage of obsessively posting the first drawing from this patent because it is such an important patent:

AKIDA NPU

 

[Slade]

Hi Dhm,

To determine the purpose of a patent, it is often useful to look at the Background of the Invention at the start of the specification ...

[0003] It has long been a goal for artificial neural networks to replicate the function of the biological neural network (the brain), with limited success. Brute force hardware approaches to the design of an artificial neural network have been cumbersome and inadequate, and fall far shown of the desired goal of replicating the functionality of the human brain. Thus, a need exists for an approach that enables an autonomous, reconfigurable spiking neural network to be realized that can scale to very large networks, yet fit on a chip while rapidly making inferences from a wide variety of possible input data and/or sensor sources.

... but sometimes this only contains motherhood statements.


You can look at the claims, chiefly claim 1 to see what the invention is:

1 . A neuromorphic integrated circuit, comprising:
a spike converter configured to generate spikes from input data;
a reconfigurable neuron fabric comprising a neural processor comprising a plurality of spiking neuron circuits configured to perform a task based on the spikes and a neural network configuration;
a memory comprising the neural network configuration, wherein
the neural network configuration comprises a potential array and a plurality of synapses,
the neural network configuration defines connections between the plurality of spiking neuron circuits and the plurality of synapses,
the potential array comprising membrane potential values for the plurality of spiking neuron circuits, and
the plurality of synapses having corresponding synaptic weights; and
a processor configured to modify the neural network configuration based on a configuration file.

It is also worthwhile looking at other independent claims (claims that are not appended to another claim).

17 . A method, comprising:
receiving, at a spiking neuron circuit, a set of spike bits corresponding to a set synapses associated with the spiking neuron circuit;
applying, at the spiking neuron circuit, a first logical AND function to a first spike bit in the set of spike bits and a first synaptic weight of a first synapse in the set of synapses, wherein the first synapse corresponds to the first spike bit;
incrementing, at the spiking neuron circuit, a membrane potential value associated with the spiking neuron circuit based on the applying;
determining, at a neural processor, that the membrane potential value associated with the spiking neuron circuit reached a learning threshold value associated with the spiking neuron circuit; and
performing, at the neural processor, a Spike Time Dependent Plasticity (STDP) learning function based on the determination that the membrane potential value of the spiking neuron circuit reached the learning threshold value associated with the spiking neuron circuit.

So, per claim 1, this patent is for a circuit arrangement enabling a processor to configure the synaptic connexions between neurons, and to allocate weights to the synapses on the basis of a configuration file. The number of layers, the number of neurons in each layer, and the synaptic connexions between the neurons are defined in the configuration file.

The configuration file can be changed to suit the particular application. This gives Akida great flexibility as to the tasks it can be used to perform simply by changing the configuration file.

There can be different configuration files for camera data, DVS data, LiDaR data, radar data sound data, smell, touch, taste, ...

Claim 17 defines the learning process using STDP, so this is a crucially important patent. In fact, we seem to have gotten away with 2 inventions for the price of 1.

Very helpful @Diogenese and really appreciated. With your detailed explanation I can see how significant this new patent is.

 

[Newk R]

dont worry about mitchell.....
View attachment 18638

Yeah Baby!!!

 

[Slade]

On that note I guess there was never any doubt. The patent before this one was not announced by the BrainChip team. This one has been announced because they know its a game changer.

 

[Fullmoonfever]

Not liking look of Nov...Boo

C'mon BRN, make sure you buck that trend this year

Damn liking Dec & Jan though

 

[Fact Finder]

On that note I guess there was never any doubt. The patent before this one was not announced by the BrainChip team. This one has been announced because they know its a game changer.

They may well say it was not important but they completely ignored my argument that it was important because it in my opinion would allow Hey Mercedes to securely speak to your Samsung Smarthome as you wiz along the autobahn at 280 kph running late for an appointment when you realise you left the iron on and forgot to lock the back door.

This incredibly important use case was just ignored in the same way a parent ignores a child who repeats something they were not supposed to hear the adults talking about.

Instead they relied upon the opinion of their patent lawyers that it was not scientifically revolutionary or first of its kind so nothing to see here kind of argument so move on. I have never liked having my head patted.

Anyway so much happening what’s one more exciting potential relationship anyway.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Slade]

They may well say it was not important but they completely ignored my argument that it was important because it in my opinion would allow Hey Mercedes to securely speak to your Samsung Smarthome as you wiz along the autobahn at 280 kph running late for an appointment when you realise you left the iron on and forgot to lock the back door.

This incredibly important use case was just ignored in the same way a parent ignores a child who repeats something they were not supposed to hear the adults talking about.

Instead they relied upon the opinion of their patent lawyers that it was not scientifically revolutionary or first of its kind so nothing to see here kind of argument so move on. I have never liked having my head patted.

Anyway so much happening what’s one more exciting potential relationship anyway.

My opinion only DYOR
FF

AKIDA BALLISTA

Damn those patent lawyers.

 

[Fullmoonfever]

From a Clarivate presentation earlier this year.

Can't recall if I posted this here or the other place.

Looks like data up to maybe 2020/21.

Hopefully we will start to see an accelerated climb in strength & age as the new patents due mentioned this year by BRN, start to come online.

https://www.morganlewis.com/-/media/files/publication/outside-publication/white-paper/identifying-gaps-in-the-ai-hardware-patent-landscape-to-grow-market-share-clarivate.pdf

 

[Dhm]

Hi Dhm,

To determine the purpose of a patent, it is often useful to look at the Background of the Invention at the start of the specification ...

[0003] It has long been a goal for artificial neural networks to replicate the function of the biological neural network (the brain), with limited success. Brute force hardware approaches to the design of an artificial neural network have been cumbersome and inadequate, and fall far shown of the desired goal of replicating the functionality of the human brain. Thus, a need exists for an approach that enables an autonomous, reconfigurable spiking neural network to be realized that can scale to very large networks, yet fit on a chip while rapidly making inferences from a wide variety of possible input data and/or sensor sources.

... but sometimes this only contains motherhood statements.


You can look at the claims, chiefly claim 1 to see what the invention is:

1 . A neuromorphic integrated circuit, comprising:
a spike converter configured to generate spikes from input data;
a reconfigurable neuron fabric comprising a neural processor comprising a plurality of spiking neuron circuits configured to perform a task based on the spikes and a neural network configuration;
a memory comprising the neural network configuration, wherein
the neural network configuration comprises a potential array and a plurality of synapses,
the neural network configuration defines connections between the plurality of spiking neuron circuits and the plurality of synapses,
the potential array comprising membrane potential values for the plurality of spiking neuron circuits, and
the plurality of synapses having corresponding synaptic weights; and
a processor configured to modify the neural network configuration based on a configuration file.

It is also worthwhile looking at other independent claims (claims that are not appended to another claim).

17 . A method, comprising:
receiving, at a spiking neuron circuit, a set of spike bits corresponding to a set synapses associated with the spiking neuron circuit;
applying, at the spiking neuron circuit, a first logical AND function to a first spike bit in the set of spike bits and a first synaptic weight of a first synapse in the set of synapses, wherein the first synapse corresponds to the first spike bit;
incrementing, at the spiking neuron circuit, a membrane potential value associated with the spiking neuron circuit based on the applying;
determining, at a neural processor, that the membrane potential value associated with the spiking neuron circuit reached a learning threshold value associated with the spiking neuron circuit; and
performing, at the neural processor, a Spike Time Dependent Plasticity (STDP) learning function based on the determination that the membrane potential value of the spiking neuron circuit reached the learning threshold value associated with the spiking neuron circuit.

So, per claim 1, this patent is for a circuit arrangement enabling a processor to configure the synaptic connexions between neurons, and to allocate weights to the synapses on the basis of a configuration file. The number of layers, the number of neurons in each layer, and the synaptic connexions between the neurons are defined in the configuration file.

The configuration file can be changed to suit the particular application. This gives Akida great flexibility as to the tasks it can be used to perform simply by changing the configuration file.

There can be different configuration files for camera data, DVS data, LiDaR data, radar data sound data, smell, touch, taste, ...

Claim 17 defines the learning process using STDP, so this is a crucially important patent. In fact, we seem to have gotten away with 2 inventions for the price of 1.

Long term readers may recall that I went through a stage of obsessively posting the first drawing from this patent because it is such an important patent:

View attachment 18639

@Diogenese thank you. I wish I could give you 100 Likes! Anyway I will symbolically give you a man hug instead!

 

[Cardpro]

Once BrainChip ship goes on the full speed, it will soon transform to a rocket and fly off to space. The growth will be exponential!!!

 

[Dhm]

I just worked out why Blind Freddie was so dismissive of my question. It was because of this statement by Douglas Fairbairn Director of Business Development for MegaChips USA:

“What people have found is that the existing processors from ARM or RISC-V do not address the power performance requirements of the AI industry.”

It of course makes clear why ARM and SiFive both jumped virtually together to announce partnerships with Brainchip.

It was so they could “address the power performance requirements of the Ai industry.”

Now I get it.

Blind Freddie shakes his head and walks away.

My opinion only DYOR
FF

AKIDA BALLISTA

Billie Field's Blind Freddie Knew That:



Apologies to Billie, but I have altered the lyrics to suit our new love affair;

Standing on the outside, I don't know where I'm going to
But I do know-ow just one thing Intel, and that is it's over with you

I've been very lonely, I did not think I could go on
I was caught, with Loihi, and dreams I should have won

Blind Freddy knew that, a blind man could see
I was in love with you, but Intel you weren't in love with me

Now I want to change it, it’s Brainchip, and I know why
We all know, that you’re the best Akida, it helps if I cry

Blind Freddy knew that, a blind man could see
I am not in love with Intel (not in love with you)
And you weren't in love with me

Now it’s so perfect (so perfect)
It used to be so crap
In lovin' you (Akida it’s so perfect lovin' you)
Now it’s so perfect, Brainchip’s so perfect
In lovin' you Akida (now it’s so perfect)

So now I'm standing on the outside, I know now where I'm going to
And I do know just one thing, Akida is so much better than you