BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
rgupta
Regular
I assume the govt and ASX has to review their priorities and policies to attract better tech companies on the asx.
e.g altassian is there on asx
Also didn't someone post not all that long ago that CommSuc said that BrainChip was one to watch! They clearly haven't been watching or have they? The plot thickens!
Fact Finder
Top 20
And I didn’t mention Todd Vierra’s comment in his CES 2024 interview about the Tape Out of AKIDA 2.0 which potentially allows Nandan Nayampally to hold up a real live fully functioning AKIDA 2.0 for his audience in May, 2024.
Perhaps even ask it about how well it thought he handled the presentation and whether there was anything else he might say before concluding.
My opinion only DYOR
Fact Finder
Perhaps even ask it about how well it thought he handled the presentation and whether there was anything else he might say before concluding.
My opinion only DYOR
Fact Finder
The problem is that Australia does not have the financial advisers and commentators who know anything about technology.
Give them a hole in the ground and they will start JORCing off to their heart's content, but give them a sniff of a transistor and their brain shuts down.
The November 2022 entry in Slide 15 (RadNeuro Maturation Timeline) refers to "Rad-tolerant FPGA neural cores developed with selected partners (eg, Brainchip)".
Slide 16 is a continuation of Slide 15, and gives the timeline for fabrication of a full-scale Rad-Tolerant Neuro-processor as 36 months possibly with September 2023 as the datum, which is a long time to hold one's breath.
If it is just a matter of increasing the architecture, then the tape-out is already there.
Berlinforever
Member
As longtimers, we are not concerned about recent price swings in any direction. We are not checking the stock price every 30 min. Right? Sure, we like to know about the catalysts that sparked this recent wave. My view: Akida 2.0 tape-out announcement imminent.
As a long term shareholder I have seen how the share price can spike. I will be watching it very closely this week.
Hi FF,
These are Mentium patents designed to compensate for manufacturing/temperature variability in analog neurons:
US11056169B2 Current comparator for submicron processes 20190717
BAYAT FARNOOD MERRIKH [US]; SULIMA JAROSLAW [US]; PREZIOSO MIRKO
A system for comparing currents is disclosed. The system may include a first current signal and a second current signal. The system may also include a subtractor that is configured to receive a plurality of current input signals and generate a single output current signal that is equal to a difference between the plurality of current input signals. The system may also include a current-to-voltage converter that is configured to receive the output current signal and convert it into an output voltage.
US10891080B1 Management of non-volatile memory arrays 20180604 20210112
The system may include a digital-to-analog converter configured to convert a digital signal to an analog signal. The system may include sample/hold circuits configured to receive and store the analog signal. The system may include an address controller configured to regulate which sample/hold circuits propagate the analog signal. The sample/hold circuits may be configured to feed the analog signal to devices of a memory array. The system may include an output circuit configured to program the devices by comparing currents of the devices to a target current. In response to one or more of the currents of the devices being within a threshold range, the output circuit may discontinue programming the corresponding devices. In response to one or more of the currents of the devices not being within the threshold range, the output circuit may continue programming the corresponding devices.
Depending on the sparsity regime they use, the current comparison could be quite intensive. This additional signal conditioning, plus the ADC conversion so the result can be further processed by a CPU, detracts from the notional advantage of analog over digital and increases latency.
Should we start a "Go Fund Me"?Gives quantitative meaning to the phrase "take it in the shorts"
Nvidia Stock Surge Causes $3 Billion Loss for Short Sellers
(Bloomberg) -- The surge in Nvidia Corp. shares on Thursday has left short sellers with about $3 billion in paper losses, according to an analysis by S3 Partners LLC, which called it an “AI generated nightmare” for bearish traders.Most Read from BloombergUS Unveils Biggest Sanctions Package on...finance.yahoo.com
Fact Finder
Top 20
Hi Diogenese
Something I find very strange is that a new wombat turned up on HC a couple of days ago after Tom & Jerry and another poster who shall remain nameless for reasons related to the Devil and Black Magic claiming that another company had replaced Brainchip AKIDA at NASA and was so sensational it was going to kill off Nvidia.
This poster was dismissed as another ratbag immediately. Then the same discussion commences over here on TSEx in a sensible fashion and you laid out all the intelligent reasons why such an assertion was not founded in fact.
What I thought stood out from the RadNeuro documents which as you know were previously posted months ago was the following paragraph:
“November 2022: Rad-tolerant FPGA neural cores developed with selected partners (e.g., Brainchip). Integration with radiation-tolerant configuration and synaptic memory. Radiation and functional testing through January 2023”
The significance I think comes from the actual wording and the fact that it was posted back in 2022 that Vorago (a selected partner by NASA for Brainchip) had submitted a radhard design for an AKIDA FPGA under their Phase 1 SBIR and sought funding in a Phase 11 to move on to actually produce a Hardsil Radhard AKIDA FPGA proof of concept. We have never had confirmation of this application being approved.
(It might jog your memory if I mention the LinkedIn statement of the Vorago CEO that AKIDA was going to permit the Rover to achieve NASA’s target speed of 20kph.)
Anyway going back to the extracted paragraph you will see there is a full stop after the bracketed word “(Brainchip).”
What I think is that this is acknowledging the AKIDA FPGA has already been made radhard by Brainchip’s selected partner Vorago.
Further remember the Numem Phase 1 SBIR that was to prove out if the SRAM memory of AKIDA could be replaced by or supplemented with RRam. I know you commented on the fact that the Nemum Phase 1 had closed and that we would probably not hear any more.
Well I am putting all this together with the next sentence in the above extract “Integration with radiation-tolerant configuration and synaptic memory” to suggest this NASA, Brainchip, Vorago, Numem puzzle has been put together for us in this paragraph.
What do you think?
My opinion only DYOR
Fact Finder
Last edited:
Humble Genius
Regular
When I read your post yesterday, it made me think of the surveillance setup I did earlier last week to determine what was eating my figs. It took me 1.2Terabytes of footage to finally figure it out. Applying the AKIDA technology could have potentially reduced the data to a tiny miniscule of what I had to get to the end result.
Heres the vid, in case you're wondering whats been eating my figs........
(and in case anyone asks why i didn't use motion detection, it is because it would've been difficult to setup considering I didn't know what I was looking for, ie, a bird, mouse, possum, etc.. I could've run inmto a situation of recording false positives such as rustling leaves, and potentially not picking up on the said animal)
Yes. Knowing NASA's proclivity for ReRAM because of its inherent RadHard characteristics, I had previously speculated that a combination of Akida with ReRAM as a backup memory storing models and configuration data and other mission critical data would be on the cards with NASA.
I haven't studied the Mentium patents, but I think the current-matching circuit would only work with 1-bit weights/activations, or at least it gets more complex with multi-bits. As we know Akida was upgraded to 4 bits by popular request, so there is a need for greater accuracy than 1-bit in NNs, so I think this should weigh in our favour when NASA has to decide which NN to adopt.
Something which Mentium says is interesting:
https://mentium.tech/#tecnology
- 100x the speed and 50x efficiency of current state-of-the-art solutions, without compressing DNN models
So this looks like sparsity is not high on their agenda. They are being profligate with the benefits of analog. When you add in the current-correction, it's not surprising that their improvement over SOTA is so modest.
FN: SOTA in the NASA world was said to be the Google TensorFlow Edge processor in 2021:
https://ntrs.nasa.gov/api/citations...GCD APR - RadNeuroSept2021AmesReview.pptx.pdf
Slide 18
Google's TPU uses ... shock... horror ... SOFTWARE!!!!
https://en.wikipedia.org/wiki/Tensor_Processing_Unit
Tensor Processing Unit (TPU) is an AI accelerator application-specific integrated circuit (ASIC) developed by Google for neural network machine learning, using Google's own TensorFlow software.[1] Google began using TPUs internally in 2015, and in 2018 made them available for third-party use, both as part of its cloud infrastructure and by offering a smaller version of the chip for sale.
Comparison to CPUs and GPUs[edit]
Compared to a graphics processing unit, TPUs are designed for a high volume of low precision computation (e.g. as little as 8-bit precision)[2] with more input/output operations per joule, without hardware for rasterisation/texture mapping.[3] The TPU ASICs are mounted in a heatsink assembly, which can fit in a hard drive slot within a data center rack, according to Norman Jouppi.[4]That's a pretty low bar.
When I read your post yesterday, it made me think of the surveillance setup I did earlier last week to determine what was eating my figs. It took me 1.2Terabytes of footage to finally figure it out. Applying the AKIDA technology could have potentially reduced the data to a tiny miniscule of what I had to get to the end result.
Heres the vid, in case you're wondering whats been eating my figs........
(and in case anyone asks why i didn't use motion detection, it is because it would've been difficult to setup considering I didn't know what I was looking for, ie, a bird, mouse, possum, etc.. I could've run inmto a situation of recording false positives such as rustling leaves, and potentially not picking up on the said animal)
Must have been a bit of a BATtle, for you HG.
Nice post, and quite a bit, a different type of surveillance in my understanding. You figured that out well, and so did I
Hotty...
gltah...
Dracula!When I read your post yesterday, it made me think of the surveillance setup I did earlier last week to determine what was eating my figs. It took me 1.2Terabytes of footage to finally figure it out. Applying the AKIDA technology could have potentially reduced the data to a tiny miniscule of what I had to get to the end result.
Heres the vid, in case you're wondering whats been eating my figs........
(and in case anyone asks why i didn't use motion detection, it is because it would've been difficult to setup considering I didn't know what I was looking for, ie, a bird, mouse, possum, etc.. I could've run inmto a situation of recording false positives such as rustling leaves, and potentially not picking up on the said animal)
I have a similar problem with flying foxes in my mango tree.
Another "Go Fund Me"?
Similar threads
