CHIPS
Regular
He is a troll, don't feed him!
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
He is a troll, don't feed him!
HappyChippy
Member
Here found from a tear down findings post on a blog.
I read this with the below link which claims mediated having trialed an APU/NPU providing 400% power and performance efficiencies.
Article from late 2021.
Must be nothing.
MediaTek Announces Dimensity 9000: Supercharged Flagship SoC on 4nm
Attachments
Yeah evidently it all goes to his cavernous head
Yep, our work is done here. Let's kick on and resume focus on AKIDA. This could be a big week, perhaps with an announcement or two. You'd think with all of the neuromorphic news out there that investors must be learning about Brainchip - the worlds only commercially available neuromorphic chip. Bring it on
Fullmoonfever
Top 20
Agree, be nice to see some official Anns start to flow sooner than later.Yep, our work is done here. Let's kick on and resume focus on AKIDA. This could be a big week, perhaps with an announcement or two. You'd think with all of the neuromorphic news out there that investors must be learning about Brainchip - the worlds only commercially available neuromorphic chip. Bring it on
For those like me who appreciate John Wick....here's a shout out
@Diogenese
Patent from Accenture.
https://image-ppubs.uspto.gov/dirsearch-public/print/downloadPdf/20230290340
Patent from Accenture.
https://image-ppubs.uspto.gov/dirsearch-public/print/downloadPdf/20230290340
"We'll have more foundry partners, that's not the end of it"
TSMC, Global Foundries and Intel Foundry Services.
That's such a strong statement, we have proven that our technology is portable across a number of foundries already, customers
who may be checking us out have already been put at ease, proving to me at least, the company is ahead of the curve.
Despite our share price being so low, I can categorically say that I have never been so confident in our staff, company strategy and
technology in the last 8 years as I currently am...I have no fear at all, negative or otherwise, everything from our home webpage to
our presentations has improved, I can feel it and see it as I'm sure you can to.
IP Licenses are a bit like pulling teeth, it's not an easy journey, but once it's completed, the world is your oyster, our day is fast approaching.
Love Brainchip
TSMC, Global Foundries and Intel Foundry Services.
That's such a strong statement, we have proven that our technology is portable across a number of foundries already, customers
who may be checking us out have already been put at ease, proving to me at least, the company is ahead of the curve.
Despite our share price being so low, I can categorically say that I have never been so confident in our staff, company strategy and
technology in the last 8 years as I currently am...I have no fear at all, negative or otherwise, everything from our home webpage to
our presentations has improved, I can feel it and see it as I'm sure you can to.
IP Licenses are a bit like pulling teeth, it's not an easy journey, but once it's completed, the world is your oyster, our day is fast approaching.
Love Brainchip
robsmark
Regular
He's referring to his bank balance - courtesy of Brainchip!
Good to see BrainChip top of the list when searching neuromorphic on Arms website.
Attachments
Tothemoon24
Top 20
– There is a need for Neuromorphic Computing
#Artificial #intelligence (#AI) is revolutionizing industries and improving our lives, but it comes with a hidden downside:
massive energy consumption.
As AI systems become more complex, they require increasing computational power and energy to process vast amounts of data.
Training a single AI model can consume as much energy as a car does in its lifetime.
Data centers, which power AI systems, accounted for 1% of global electricity use in 2018.
This energy consumption has environmental implications and raises sustainability concerns. It also threatens resource shortages and hampers AI sector growth.
Tech giants like Google are investing in renewable energy and committing to powering data centers sustainably.
Researchers are developing energy-efficient algorithms and hardware, such as neuromorphic computing and quantum computing.
To continue benefiting from AI while mitigating its environmental impact, the industry must develop energy-efficient systems and adopt sustainable practices.
Balancing AI's power with environmental responsibility is crucial as we embrace its transformative potential.
The human brain consumes just 10-20 Watt.
A new technology of Artificial Intelligence is underway and ready to use:
#Neuromorphic #Computing
Hi MrRomper,
The US specification is not available on Espacenet just yet, but the EU version is.
The specification draws a distinction between "brain inspired hardware" and "SNNs which run on the neuromorphic hardware".
EP4243011A1 EFFICIENT SPEECH TO SPIKES CONVERSION PIPELINE FOR A SPIKING NEURAL NETWORK 20220307
[0005] Brain-inspired neuromorphic hardware and Spiking Neural Network models (SNNs) which run on the neuromorphic hardware can provide desirable benefits that make them well-suited for speech recognition for an edge device.
[0026] The audio to spikes conversion pipeline 122 is configured to convert audio data into spikes for processing by the neuromorphic processor 124. A series of spikes output by the audio to spikes conversion pipeline 122 for audio data can be referred to as speech input that is provided as input to a SNN 126 executed by the neuromorphic processor 124. The audio to spikes conversion pipeline 122 can include multiple processing stages to convert the audio data into the spikes.
[0027] The neuromorphic processor 124 is configured to perform speech recognition tasks using a SNN 126. For example, the neuromorphic processor 124 can execute the SNN 126 to predict speech sounds represented by the spikes of the spike input output by the audio to spikes conversion pipeline 122. For example, the SNN 126 can process the spikes to generate speech recognition results that include predicted speech sounds, e.g., in the form of textual phonemes, triphones, full words, or other units of sound, corresponding to speech of the audio data.
However, it looks like learning is done off-chip.
[0030] One or both servers 110, 112 can include a hyperparameter adjuster 128, which can be implemented in hardware and/or software. In some implementations, the hyperparameter adjuster 128 can be part of the computing device 102. The hyperparameter adjuster 128 can evaluate SNNs and adjust hyperparameters of the audio to spikes conversion pipeline 122 based on the evaluation. Example hyperparameters of the audio to spikes conversion pipeline 122 are described with reference to FIG. 2 . For example, the hyperparameter adjuster 128 can perform hyperparameter sweeps using multiple sets of hyperparameter configurations and select a hyperparameter configuration based on the results of the sweeps. Each sweep can include applying a particular hyperparameter configuration to the audio to spikes conversion pipeline 122, training multiple SNNs using training data and spikes generated by the audio to spikes conversion pipeline 122 based on the training data, testing each of the multiple SNNs using testing data, measuring the accuracy of each SNN based on the test. The training data and/or the testing data can include labeled audio data with labels indicating predicted speech sounds, e.g., words or phonemes, corresponding to the audio data.
[0031] The hyperparameter adjuster 128 can adjust the hyperparameters of the audio to spikes conversion pipeline 122 based on the evaluation. For example, the hyperparameter adjuster 128 can identify the hyperparameter configuration for which the most accurate SNN was trained during the sweeps and apply that hyperparameter configuration to the audio to spikes conversion pipeline 122.
As an aside. remember Sean's Podcast with Jean-Luc Chatelain from Accenture from January:
Episode 23 – BrainChip Talks AI Innovation with Accenture’s Managing Director and Global Chief Technology Officer of Accenture Applied Intelligence Jean-Luc Chatelain - BrainChip
Episode 23: AI with Accenture’s Managing Dir. & Global CTO Accenture Applied Intelligence Jean-Luc Chatelain
Listen to an insightful podcast with Sean Heir and Jean-Luc Chatelain and gain valuable industry information.
brainchip.com
At the 12 minute mark, Sean asks an innocent question about transformers. If only he could have seen what was coming!!??
David holland
Regular
Old Lou and Peter in charge sent the company backwards , lucky we have a professional team now"We'll have more foundry partners, that's not the end of it"
TSMC, Global Foundries and Intel Foundry Services.
That's such a strong statement, we have proven that our technology is portable across a number of foundries already, customers
who may be checking us out have already been put at ease, proving to me at least, the company is ahead of the curve.
Despite our share price being so low, I can categorically say that I have never been so confident in our staff, company strategy and
technology in the last 8 years as I currently am...I have no fear at all, negative or otherwise, everything from our home webpage to
our presentations has improved, I can feel it and see it as I'm sure you can to.
IP Licenses are a bit like pulling teeth, it's not an easy journey, but once it's completed, the world is your oyster, our day is fast approaching.
Love Brainchip
Rise from the ashes:
I find it really strange that the big head has been responsible for at least 5 decent contributing posters to leave this forum.
And it keeps others that would be motivated to contribute out, I think. At least myself, because I know it would just bother me too much. I like to keep up here, and I’m very glad and thankful for all the great information many here share that I couldn’t find on my own. But the constant distractions and unnecessary and at times even hurtful comments also force me to stay away at times, even though I don’t even participate.
I find it really strange that the big head has been responsible for at least 5 decent contributing posters to leave this forum.
And it keeps others that would be motivated to contribute out, I think. At least myself, because I know it would just bother me too much. I like to keep up here, and I’m very glad and thankful for all the great information many here share that I couldn’t find on my own. But the constant distractions and unnecessary and at times even hurtful comments also force me to stay away at times, even though I don’t even participate.
Last edited:
I feel like you need a hug. Maybe take some time out?
Look at your responses to those who are just trying to ask, pose a question or make a point.
Your responses appear to come across as harsh, out of touch/balance or are just plain rude. I used to love the green days just as much as the next poster…however you seem bewildered and want to take your frustrations out on people here. From my point of you…this is not ok.
Also you clog up a thread that I love to read, get a new perspective or move on @wilzy123 as I’ve had enough, and I’m sure many on here also have!
Rant over, I know I can put you on ignore but I’m reaching out for the sake of this thread and those you target on it @wilzy123 !
Love this, my post later on says much the same! @wilzy123 listen up!
He’s apparently a founding member of this forum? Can’t ignore this?
Accenture have a page on their website all about neuromorphic computing. But they only refer to Intel loihi throughout the document.Hi MrRomper,
The US specification is not available on Espacenet just yet, but the EU version is.
The specification draws a distinction between "brain inspired hardware" and "SNNs which run on the neuromorphic hardware".
EP4243011A1 EFFICIENT SPEECH TO SPIKES CONVERSION PIPELINE FOR A SPIKING NEURAL NETWORK 20220307
View attachment 44763
[0005] Brain-inspired neuromorphic hardware and Spiking Neural Network models (SNNs) which run on the neuromorphic hardware can provide desirable benefits that make them well-suited for speech recognition for an edge device.
[0026] The audio to spikes conversion pipeline 122 is configured to convert audio data into spikes for processing by the neuromorphic processor 124. A series of spikes output by the audio to spikes conversion pipeline 122 for audio data can be referred to as speech input that is provided as input to a SNN 126 executed by the neuromorphic processor 124. The audio to spikes conversion pipeline 122 can include multiple processing stages to convert the audio data into the spikes.
[0027] The neuromorphic processor 124 is configured to perform speech recognition tasks using a SNN 126. For example, the neuromorphic processor 124 can execute the SNN 126 to predict speech sounds represented by the spikes of the spike input output by the audio to spikes conversion pipeline 122. For example, the SNN 126 can process the spikes to generate speech recognition results that include predicted speech sounds, e.g., in the form of textual phonemes, triphones, full words, or other units of sound, corresponding to speech of the audio data.
However, it looks like learning is done off-chip.
[0030] One or both servers 110, 112 can include a hyperparameter adjuster 128, which can be implemented in hardware and/or software. In some implementations, the hyperparameter adjuster 128 can be part of the computing device 102. The hyperparameter adjuster 128 can evaluate SNNs and adjust hyperparameters of the audio to spikes conversion pipeline 122 based on the evaluation. Example hyperparameters of the audio to spikes conversion pipeline 122 are described with reference to FIG. 2 . For example, the hyperparameter adjuster 128 can perform hyperparameter sweeps using multiple sets of hyperparameter configurations and select a hyperparameter configuration based on the results of the sweeps. Each sweep can include applying a particular hyperparameter configuration to the audio to spikes conversion pipeline 122, training multiple SNNs using training data and spikes generated by the audio to spikes conversion pipeline 122 based on the training data, testing each of the multiple SNNs using testing data, measuring the accuracy of each SNN based on the test. The training data and/or the testing data can include labeled audio data with labels indicating predicted speech sounds, e.g., words or phonemes, corresponding to the audio data.
[0031] The hyperparameter adjuster 128 can adjust the hyperparameters of the audio to spikes conversion pipeline 122 based on the evaluation. For example, the hyperparameter adjuster 128 can identify the hyperparameter configuration for which the most accurate SNN was trained during the sweeps and apply that hyperparameter configuration to the audio to spikes conversion pipeline 122.
As an aside. remember Sean's Podcast with Jean-Luc Chatelain from Accenture from January:
Episode 23 – BrainChip Talks AI Innovation with Accenture’s Managing Director and Global Chief Technology Officer of Accenture Applied Intelligence Jean-Luc Chatelain - BrainChip
Episode 23: AI with Accenture’s Managing Dir. & Global CTO Accenture Applied Intelligence Jean-Luc Chatelain
Listen to an insightful podcast with Sean Heir and Jean-Luc Chatelain and gain valuable industry information.
At the 12 minute mark, Sean asks an innocent question about transformers. If only he could have seen what was coming!!??
Similar threads
