toasty
Regular
I don't think theres any straws involved. Remember what happened with the first iteration of the product after it was given to EAP's???
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
BaconLover
Founding Member
EAPs would have the option imo.
toasty
Regular
Snap!!
mkm109
Regular
I think this is a sensible way to address the problem
BaconLover
Founding Member
Let's be positive.
Remember once a famous person said ''Let's be kind to each other''?
Take it easy guys.
jtardif999
Regular
You’re digging yourself a deep hole here… maybe just ease back on the throttle a little bring her in for a safe landing
. Me and many others here were sitting on 60-70% losses for nearly 2 years when the share price was languishing at around 6c. I believe we will all reap the benefit of this company’s (imo), expert timing and positioning in the industry over the next few years.I was hoping you'd show up.
Fullmoonfever
Top 20
Was just skimming a paper from last years NeurIPS conference.
The authors listed responses to some questions and one answer in particular to an official Reviewer question was good to see.
Hope they take note and we get some enquiries along this line of research as well.
openreview.net
Question.
5. What is computational burden of these Encoder? Since they are for implantable applications this would be an important metric. Latency for such computing these models. Definitely cannot use NVIDIA GPU for the implanted application. How would this scale
Answer.
We agree that further discussion on deployment is necessary, and have added a deployment section in the Discussion to address this (lines 299-303). To summarize, the encoder inference time for a single target MNIST image is 1.2 ms on a Nvidia 3090 GPU and 4.1ms on a single-core 2.1GHz CPU, and the saved model takes 5.3MB of memory. Future work could reduce these measures by employing techniques such as pruning, mixed floating point precision, search over smaller encoder architectures, or Tensorflow Lite conversion. Another practical solution may come in the form of Edge AI accelerators (e.g., Intel’s Neural Compute Stick) and deep learning-compatible neuromorphic hardware (e.g., BrainChip’s Akida Neuromorphic SoC), which can lead to orders of magnitude reductions in size, weight, and power compared to conventional CPUs or GPUs.
The authors listed responses to some questions and one answer in particular to an official Reviewer question was good to see.
Hope they take note and we get some enquiries along this line of research as well.
Hybrid Neural Autoencoders for Stimulus Encoding in Visual and...
We propose an encoder-decoder based stimulus encoding framework for sensory neuroprostheses and demonstrate its effectiveness for visual prostheses.
Hybrid Neural Autoencoders for Stimulus Encoding in Visual and Other Sensory Neuroprostheses
Question.
5. What is computational burden of these Encoder? Since they are for implantable applications this would be an important metric. Latency for such computing these models. Definitely cannot use NVIDIA GPU for the implanted application. How would this scale
Answer.
We agree that further discussion on deployment is necessary, and have added a deployment section in the Discussion to address this (lines 299-303). To summarize, the encoder inference time for a single target MNIST image is 1.2 ms on a Nvidia 3090 GPU and 4.1ms on a single-core 2.1GHz CPU, and the saved model takes 5.3MB of memory. Future work could reduce these measures by employing techniques such as pruning, mixed floating point precision, search over smaller encoder architectures, or Tensorflow Lite conversion. Another practical solution may come in the form of Edge AI accelerators (e.g., Intel’s Neural Compute Stick) and deep learning-compatible neuromorphic hardware (e.g., BrainChip’s Akida Neuromorphic SoC), which can lead to orders of magnitude reductions in size, weight, and power compared to conventional CPUs or GPUs.
Last edited:
Someone posted a NASA SBIR the other day in which they bagged Akida for having a superfluous CPU, while praising FD-SOI.
Deadpool
hyper-efficient Ai
You have a very Zen energy about you of late old hopRelax Pal. People are just talking and expressing themselves. Who said it has to be just all one’s and zero’s or dots which may or may not be relevant to anything? Sometimes some of us get a little carried away, but that too is just humans being human. When we are all just androids on steroids it gets pretty boring. Variety is the spice of life.
Rise from the ashes
Regular
Well ain't that a lovely sight 
Fullmoonfever
Top 20
Wadya do RiseWell ain't that a lovely sight
wilzy123
Founding Member
Rise from the ashes
Regular
Looks about right.
I may have been refreshing Dre thread too muchWadya do Rise
JK200SX
Regular
"however in TD's letter he states that the Akida1500 tape out was because of direct engagement with customers"
When the news of the AKD1500 came out the other week, I did find it strange and was of the opinion either the AKD500 or 2000 would be next. One thought that did come to me at the time was this - Brainchip has shown their AKD1000 to all of the NDA co's and other announced companies and we would assume they are all impressed with the technology. The fact that little revenue has come from all these excited companies is that they were aware of what Brainchip had available in the near future(at that time) and would only commit once they see the AKD1500 work for their application with the increased functionality. Once the AKD1500 is proven, the floodgates will open.
An analogy to this is why would someone want to buy the 2022 model of a vehicle, when they know that the 2023 model, just around the corner, offers additional functionality and performance?
Damo4
Regular
Haha agreed mate, just felt like adding in some low quality bait of my own.Relax Pal. People are just talking and expressing themselves. Who said it has to be just all one’s and zero’s or dots which may or may not be relevant to anything? Sometimes some of us get a little carried away, but that too is just humans being human. When we are all just androids on steroids it gets pretty boring. Variety is the spice of life.
In summary - GO BRN
Xhosa12345
Regular
Yep they have 4 strategies they are focusing on to grow - and we are mentioned in one of the strategies being Initiatives in AI field
But agree we will never know - we can only hope!
And its only been a little over a year since the deal was signed with megachips
Will have to see if we can reverse engineer the revenue figure in the upcoming annual report for the figures to end December
Seriously hoping the figure at the end of this calendar year will be a shite-load more
alwaysgreen
Top 20
From memory, most of Megachips revenue is from Nintendo. If we can make our way into their next console and controllers, we may end up in 80% of their products? That would be a nice earner as Nintendo sell about 20 million consoles a year.
Xhosa12345
Regular
Well ain't that a lovely sight
Maybe ole Shareman's 12,636 posts for 57 likes and 1 great analysis have finally crashed their server?
Similar threads
