MDhere
Top 20
Sounds like it was a good one, just like this pic from the Tata Daewoo, ZF . I've not been drinking but my head is spinning.Emptying as we speak ...
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Purple, that’s new to me. I looked up a review and found two blokes on twitter. They should be famous.That's a great man cave. It looks like a few of us here have great taste in beer... It's purple for me and then the red if I run out of purple, but the green is great too
View attachment 16918
The 8 bit floating point discussion relates to the type of data to be used to store the model libraries for use in AI/NNs - that is an information byte is a string of eight ones and zeros, eg, 10010011.
The first bit is the sign, 1 for minus, zero for plus.
The next 4 bits are the exponent (the power to which 2 is raised), adjusted to accommodate fractional numbers (negative exponents).
The last 3 bits are the mantissa (plus the implied 1 to the left of the 3 bits). [Thrifty lot, these engineers, aren't they?]*
As I was going down the stair,
I met a man(tissa) who wasn't there.
He wasn't there again today ...
BrainChip has developed a number of model libraries adapted to Akida's 1-to-4 bit format.
The CNN2SNN functionality of Akida is able to adapt CNN models to Akida format.
https://doc.brainchipinc.com/user_guide/akida_models.html
Overview
Brainchip akida_models package is a model zoo that offers a set of pre-built akida compatible models (e.g Mobilenet, VGG or AkidaNet), pretrained weights for those models and training scripts.See the model zoo API reference for a complete list of the available models.
akida_models also contains a set of quantization blocks and layer blocks that are used to define the above models.
Command-line interface for model creation
In addition to the programming API, the akida_models toolkit provides a command-line interface to instantiate and save models from the zoo.Instantiating models using the CLI makes use of the model definitions from the programming interface with default values. To quantize a given model, the CNN2SNN quantize CLI should be used.
The beauty of Akida is that it can learn on the job without having to reprogramme the whole dataset.
*FN: This thrift of Silicon real estate originates well before USB sticks were $1 per megabit.
PS: 8-bit FP model library is designed for use with CPU?GPU based processors - MetaTF can convert those model libraries for use in Akida.
Last edited:
This article talks about using AI to interpret whalespeak.
https://www.msn.com/en-au/news/worl...sedgntp&cvid=42702b474d8a45d481c709bbf261c37b
It would be great to use that to warn whales off from shark nets.
https://www.msn.com/en-au/news/worl...sedgntp&cvid=42702b474d8a45d481c709bbf261c37b
It would be great to use that to warn whales off from shark nets.
Stable Genius
Regular
Bigfella195
Member
Anyone have knowledge of Onsemi and their edge AI?
www.onsemi.com
Intelligent Power and Sensing Technologies | onsemi
Bigfella195
Member
BrainChip Holdings Ltd Appoints New Director
July 10, 2017
ALISO VIEJO, CA--(Marketwired - Jul 10, 2017) - BrainChip Holdings Ltd. (ASX: BRN)
- Appointment of highly regarded Silicon Valley executive, Emmanuel Hernandez, as Non-Executive Director.
- Brings more than 40 years of operating and board member experience including key roles with several of Silicon Valley's largest and most successful companies such as Cypress Semiconductor and ON Semiconductor.
Bigfella195
Member
What triggered my interest was Subaru announced that they would be implementing AI stereo cameras to their vehicle safety system in 2025. Onsemi has been Subaru’s safety system partner.
Bigfella195
Member
Akida is a Neuromorphic System-on-Chip (NSoC) that includes both Inference
(Deep Learning) and Autonomous Learning in a single low-power, low-latency
and low-cost integrated circuit. As a result, the Akida device provides a
competitive advantage in the advancement of Artificial Intelligence for “Edge”
applications.
Edge applications include use cases where data is acquired by a variety of
transducers and a benefit is derived by processing the information at the
transducer, rather than sending all data to an enterprise data center or cloud for
processing.
The primary applications for AI Edge are in vision systems, acoustic analysis,
cybersecurity and the industrial Internet-of-Things.
Akida Engagement Plan
•
1
The Company has established early discussions with vision sensor manufacturers,
automobile manufacturers for Advanced Driver Assistance Systems (ADAS) and
Autonomous Vehicles (AV).
Source Tractica 2018
Page | 2
For personal use only
•
•
•
It has also commenced discussions with Tier-One module suppliers to the
automotive industry in Europe and North America, Vision Guided Robotics,
Surveillance Camera, Smart Transducer and Vision Sensor manufacturers.
Original Equipment Manufacturers (OEMs), Module Manufacturers and Sensor
suppliers (some of whom the Company is already in contact with) in these
markets include:
o ADAS/AV – Valeo, Continental, Aptiv, General Motors, Ford, Micron,
Samsung
o Vision Guided Robotics - HiKvision, iRobot, IRIS, Samsung, Micron
o Drones – DJI, Safran, Boeing
o Surveillance cameras – HiKvision, Dauha, Axis
o Smart Transducers and Vision Sensors – Sensata, On Semiconductor,
Sony, ST Microelectronics, Samsung
(Deep Learning) and Autonomous Learning in a single low-power, low-latency
and low-cost integrated circuit. As a result, the Akida device provides a
competitive advantage in the advancement of Artificial Intelligence for “Edge”
applications.
Edge applications include use cases where data is acquired by a variety of
transducers and a benefit is derived by processing the information at the
transducer, rather than sending all data to an enterprise data center or cloud for
processing.
The primary applications for AI Edge are in vision systems, acoustic analysis,
cybersecurity and the industrial Internet-of-Things.
Akida Engagement Plan
•
1
The Company has established early discussions with vision sensor manufacturers,
automobile manufacturers for Advanced Driver Assistance Systems (ADAS) and
Autonomous Vehicles (AV).
Source Tractica 2018
Page | 2
For personal use only
•
•
•
It has also commenced discussions with Tier-One module suppliers to the
automotive industry in Europe and North America, Vision Guided Robotics,
Surveillance Camera, Smart Transducer and Vision Sensor manufacturers.
Original Equipment Manufacturers (OEMs), Module Manufacturers and Sensor
suppliers (some of whom the Company is already in contact with) in these
markets include:
o ADAS/AV – Valeo, Continental, Aptiv, General Motors, Ford, Micron,
Samsung
o Vision Guided Robotics - HiKvision, iRobot, IRIS, Samsung, Micron
o Drones – DJI, Safran, Boeing
o Surveillance cameras – HiKvision, Dauha, Axis
o Smart Transducers and Vision Sensors – Sensata, On Semiconductor,
Sony, ST Microelectronics, Samsung
I have been accumulating BRCHF throughout the past year through Ameritrade and am only charged $6.95USD/trade. And I now have enough shares to rank in the top 250 holders. Had there been a $50/trade fee my holdings, would I'm sure, be far less.
With the tech and overall market debacle, BrainChip being out of the investor mainstream as an Australian tech company generating essentially no revenue and not endeavoring to promote its wares and future to the investment public (run silent, run deep) and still outperforming the overall and tech market in recent months, I remain patient and optimistic and feel lucky that I discovered and acted while BrainChip is still in its infancy.
I appreciate your kindness,...I do try to make my posts enjoyable to read and attempt to state my points clearly and unambiguously.Hi @dippY22
I’ll preface my reply with stating I enjoy reading your posts and value your thoughts on this but I’m not sure what the issue is with our SP. It could be higher but given the macro events and where we are in our commercialisation journey it’s where we are currently valued. There has been satisfactory growth which beats the banks and my current super fund. It’s not exponential but it’s growth all the same!
I’ll also state that I am still waiting to buy a significant amount in December in my super so the SP remaining as it is works in my favour at the moment. I own a satisfactory amount at the moment but I’ve been nervously waiting to buy in and expecting a Murphy’s Law price rise to occur just before I can lock some more away!
I also understand shareholders are at different cycles in life, e.g. someone recently posted they were 80 years of age and obviously would like a steep price increase so they can enjoy some profits whilst they still can. That’s fair enough!
My opinion is that we are still in the infancy of commercialisation. Yes we have some potential big fish on the hook but we have yet to really see them landed yet. Therefore the public investor would still see us as a risk as we are only just starting to see revenues and not yet covering our running costs.
I am strongly of the opinion we are in Valeo Scala 3 Lidar which will eventually bring us strong revenue growth which is part of the basis of my investment decision. They are pivotal to our success at the moment.
I also note I am very disappointed Nanose hasn’t come through with it’s product as I first thought it would. It was my prime reason in investing in Brainchip when I first did 18+ months ago. I expected that to push the SP to astronomical heights but it hasn‘t eventuated and they have missed the Covid money train! I am thinking it had something to do with the ability to ”Clean“ the sensor between use to prevent contamination for the next use and therefore affect the results and/or possibly infect the next user? I thought the actual detection and analysis using Akita produced satisfactory results to enable TGA approval. I am still hopeful of a Nanose instrument of some kind to hit the market in the next few years.
I am glad that Brainchip are not selling a share price or trying to artificially inflate it with fluff announcements. I want them to focus on developing relationships, building the ecosystem and becoming a de-facto standard to increase our customer base and make IP sales. That’s what counts to me but I am comfortable waiting a few more years for that to occur.
Imagine where we would be without the Arm, SiFive, Megachips, Edge Impulse etc announcements?
As reported earlier this week NVIDIA and Intel are pushing a white paper which to my limited understanding will try to develop a standard which could affect Brainchips ability to be fully utilised as our ”Bits” are different to their “8 bit standard.” Is this an attempt to derail Brainchip’s success as VHS did with Beta. I am hopeful Brainchip are already too far down the product commercialisation path for that to occur!
So I see our SP as not really increasing significantly until we have product releases, such as MB and others to provide sustained revenue growth which might still be ”Lumpy” for the next few quarters. When/if Brainchip announce a patent regarding Akita 2.0 it might spice things up a bit so I’m hoping that will be around Xmas or January!
I am confident that if Brainchip continues to work on sales then eventually the SP will go up regardless of limited advertising to investors.
Eventually the financials will push the SP which to me is the way it should be!
I may be thinking differently come January when I am more heavily invested and get anxious for results. I will be interested to see how well I cope in that situation myself.
I am thinking my investment will be settling at it’s peak range in about 2030 once is is ubiquitous in ADAS/EV cars, phones, laptops, large server centres (lookin at you Dell) and a plethora of Defence products. By then I expect dividends which will time in nicely for my retirement so I can live off of them and not touch my holdings which I will eventually leave to my children. That is my plan and I enjoy reading this forum and doing my own research whilst I wait patiently for that to occur!
Cheers all!
However, either you missed the gist of what I was trying to get across or I didn't say it clearly. My issue is with low volumes on the OTC markets for BRCHF shares and the ADR's in the United States, and I was replying to the end of J. Delekto's previous comment where he believed the word is getting out about the Brainchip story in the U.S.
My issue is not with the stock price. I get our price, and understand why it goes up, languishes or moves down. But I don't get our ( U.S.) traded volumes.
In short, and to wrap this up and put a bow on it, .... if the word is indeed getting out about Brainchip in America, then why were only a few thousand shares traded in BRCHF last Friday, and only 3 ADR shares, to cite the most recent "action" ? Those are extremely low trading volumes, but they are not unusual, and that tells me there is little to no interest in Brainchip, for the moment, in the USA. And yes it may also be in part because the retail tech trader has retreated to his or her bunker during the tech meltdown (and interest rate meltup), as FF mentioned in an earlier reply.
Oh, yeah,.....and my comment about the return on investment (ROI) at our current stock price of 60 cents +/- is not a criticism of the current price as I explained above, but rather that it represents a screaming buy compared to what the ROI on investing a similar amount in Intel, for example, will be in a few years, imo. Which one do I believe will have the better return in 3 - 5 years.....? Brainchip, of course....and by a significant margin. I just wish other American investors would join me and JDelekto (...and charles2) and BUY some shares for gosh sakes, .... because they're on sale.
My opinion only, and no influence has been intended.
Last edited:
JDelekto
Regular
So for me, "word getting out" doesn't necessarily mean people are going to run out and buy significant shares setting the market on fire. I'm not trying to over-inflate the response of investors in the US from the small number of people with whom I've shared the investment opportunity, I apologize if that's how it sounded.
I do know that three of my co-workers and a former co-worker have invested in the stock long-term. My daughter has also purchased shares. I have converted a considerable amount of my retirement savings into BrainChip shares.
Some of the people I've informed about BrainChip are not active investors; they are just as happy letting their retirement savings sit in a mutual fund rather than make the effort (as little as it may be) to invest it.
For some people with which I shared, there was skepticism because the stock was on the OTC markets. From my experience, it seems the first thing people think of when they hear OTC is penny stocks, thus a high-risk investment. I constantly have to remind them that OTCQX is not pink-sheets and that stocks listed on OTCQX have a higher standard they have to meet, which is that they cannot qualify as a penny stock.
For me, the word "getting out" is not just word-of-mouth through friends and family, but having BrainChip appear in articles to which US investors do have access. They are mentioned alongside other more well-known companies like IBM and Intel.
I share your frustration that the OTC markets are not moving the volumes of shares we would hope to see. I'm not that big a fan of the ADRs since there is a limited number of them, and people who purchase them do not own the underlying shares unless they are converted.
I prefer the actual stock, but the annoyance is that I get hit with a foreign transaction fee. Instead of being able to set aside savings every month to buy a small number of shares, I have to save for several months. Other brokers might charge less than Fidelity, but unfortunately, that is where my retirement account resides.
I do remember they were waiving the fee for a brief period, then mysteriously reinstated it. I got no satisfactory answer on that whole fiasco whatsoever. Fidelity said there was an imposed SEC alert (which I could not find), that made them DTC ineligible and said to check with the company. I sent an e-mail to investor relations, and the reply I received from Mark at Integrous Communications was: "There is no SEC imposed alert on us that we are aware of, so I don’t know what Fidelity is saying. DTC Eligibility has been slow - we hope to have it resolved soon."
I don't know if anyone from the company followed up, but I've heard nothing about it since. Yes, I'm still paying the fee, and yes, I'm still particularly miffed by it; I hope it wasn't something "made-up" to keep charging fees.
If BrainChip was indeed able to start planting the seeds at top-tier suppliers, and be used in one or more NASA and/or US government projects, their earnings will do all of the talking.
In the MegaChips announcement, it was stated that they would see an aggregate of $2mil in the current financial year (2021) up to the end of 2022, along with the "license fee and additional revenue opportunities". However, the last earnings statement was over $4mil. To me, that shows some progress. I'm looking forward to the Q3 financials, which I expect will be filed with the SEC near the end of October.
If they can keep this trend up, I think that investors will start taking more notice. I still think that it may be a few years before I expect to see a significant shift.
This is, of course, all my opinion.
thelittleshort
Top Bloke
Arijit Mukherjee on LinkedIn: #aimlsys #tutorial #tinyml
See you at Bangalore on 12th October 2022. I'll share the microsite of this tutorial soon. #AIMLSys #Tutorial #TinyML
thelittleshort
Top Bloke
Article pasted below Diana Deca’s comment roasting Loihi 
www.linkedin.com
The article:
INTERVIEW Every time a chipmaker or researcher announces an advancement in neuromorphics, it's inevitably the same story: a brain-like AI chip capable of stupendous performance-per-watt compared to traditional accelerators.
Intuitively, the idea makes a lot of sense. Our brains are pretty good at making sense of the world, so why wouldn't a chip designed to work like them be good at it too?
Yet, after years of development and the backing of massive tech companies like IBM and Intel, these brain-like chips are still years away from making their way into consumer products.
That hasn't stopped the tech from grabbing headlines over the years. Neuromorphic chips up to 16 times more efficient; brain-like chips potentially poweringfuture supercomputers; Samsung wanting to reverse engineer the brain; IBM recreating a frog brain in silicon. You get the idea.
While the chips show promise, the reality is the field of neuromorphics is still in a very experimental stage, and faces many challenges that must be resolved before they are ready for prime time, explains Karl Freund, principal analyst at Cambrian AI Research, in an interview with The Register.
This may be one of the reasons many of the more promising neuromorphic processors have seemingly stalled.
IBM, for example, hasn't given an update on its True North neuromorphic chips, which are capable of simulating more than a million neurons, in more than four years. SpiNNaker, another promising spiking neural networking processor, received an €8 million (c $8.15 million) grant in 2019 to develop a second-gen chip based on the design. However, the company behind the chip, Dresden, Germany-based SpiNNcloud, is only now getting off the ground.
Intel's Loihi and Loihi 2 processors have come the closest to a commercial launch in so far as Intel has made development boards available to outside researchers alongside its Lava software development kit.
The Department of Energy's Sandia National Laboratories, for example, is exploring how these chips could be used to accelerate supercomputers. In a paper published in the journal Nature Electronics, researchers at Sandia demonstrated how Intel's Loihi chips "can solve more complex problems than those posed by artificial intelligence and may even earn a place in high-performance computing."
Yet, at least as of April, Intel has no plans to productize its Loihi chips anytime soon.
So what gives? Why is it these chips, which show such promise in the lab, haven't matured faster given the insatiable demand for AI/ML?
According to Freund, one of the biggest problems comes down to funding.
"I tried to connect some venture capitalists in both neuromorphic and analog [computing] and a fairly consistent response even before the current capital crunch was 'we don't invest in research'," he says. "Their take is pretty much the same as mine, which is most of the technologies, perhaps all, are still in the research phase."
As a result, progress in productizing neuromorphic computing has been limited to large companies with deep R&D budgets, he said.
But it's not just funding that's getting in the way. Freund argues the scope of the problem for neuromorphics has only gotten larger as the tech has grown more mature.
However, productizing such a chip means solving other problems, like how you get data in and out of the chip effectively.
This isn't a problem unique to neuromorphics by any means. It's one associated with quantum computing and even traditional accelerators, which have accumulated bottlenecks in recent generations due to the speed at which the data can be pre-and post-processed and/or ingested and egressed from the chip, Freund explained.
Finally, there's the issue of developing software that can take advantage of these accelerators.
"It's really going to take a whole community of researchers to solve the programmability problem of neuromorphic computing," Freund says.
"What they're finding is that platforms, like Nvidia Jetson Orin or some new novel platforms from startups, are solving the problem really quickly. So the need to do something super exotic is continuing to lessen as the state of the art in existing technologies evolves," Freund says. "If you look at what Qualcomm has done with their AI engine, you're talking milliwatts… and what it does when you take a photograph is astounding."
As a result, meaningful problems can be resolved in the power envelope required by existing digital technologies.
While neuromorphics may not be ready to replace traditional accelerators anytime soon, Freund believes the technology will eventually reach the mainstream.
"These things do take time to mature," he says, citing the rise of Arm processors in the datacenter as something that took more than 10 years to achieve. "And that was for CPUs; CPUs are easy compared to things like quantum and neuromorphic computing." ®
Bill Gonzalez on LinkedIn: What's holding back neuromorphic computing?
MATRIX - The UTSA AI Consortium for Human Well-Being recently hosted a presentation by Yulia Sandamirskaya with Intel Corporation that discussed some of the…
The article:
Brain-inspired chips promise ultra-efficient AI, so why aren’t they everywhere?
It's not because our AI overlords aren't keen on the idea
Tobias MannMon 12 Sep 2022 // 10:23 UTCINTERVIEW Every time a chipmaker or researcher announces an advancement in neuromorphics, it's inevitably the same story: a brain-like AI chip capable of stupendous performance-per-watt compared to traditional accelerators.
Intuitively, the idea makes a lot of sense. Our brains are pretty good at making sense of the world, so why wouldn't a chip designed to work like them be good at it too?
Yet, after years of development and the backing of massive tech companies like IBM and Intel, these brain-like chips are still years away from making their way into consumer products.
That hasn't stopped the tech from grabbing headlines over the years. Neuromorphic chips up to 16 times more efficient; brain-like chips potentially poweringfuture supercomputers; Samsung wanting to reverse engineer the brain; IBM recreating a frog brain in silicon. You get the idea.
While the chips show promise, the reality is the field of neuromorphics is still in a very experimental stage, and faces many challenges that must be resolved before they are ready for prime time, explains Karl Freund, principal analyst at Cambrian AI Research, in an interview with The Register.
This may be one of the reasons many of the more promising neuromorphic processors have seemingly stalled.
IBM, for example, hasn't given an update on its True North neuromorphic chips, which are capable of simulating more than a million neurons, in more than four years. SpiNNaker, another promising spiking neural networking processor, received an €8 million (c $8.15 million) grant in 2019 to develop a second-gen chip based on the design. However, the company behind the chip, Dresden, Germany-based SpiNNcloud, is only now getting off the ground.
Intel's Loihi and Loihi 2 processors have come the closest to a commercial launch in so far as Intel has made development boards available to outside researchers alongside its Lava software development kit.
The Department of Energy's Sandia National Laboratories, for example, is exploring how these chips could be used to accelerate supercomputers. In a paper published in the journal Nature Electronics, researchers at Sandia demonstrated how Intel's Loihi chips "can solve more complex problems than those posed by artificial intelligence and may even earn a place in high-performance computing."
Yet, at least as of April, Intel has no plans to productize its Loihi chips anytime soon.
What's the holdup?
So what gives? Why is it these chips, which show such promise in the lab, haven't matured faster given the insatiable demand for AI/ML?
According to Freund, one of the biggest problems comes down to funding.
"I tried to connect some venture capitalists in both neuromorphic and analog [computing] and a fairly consistent response even before the current capital crunch was 'we don't invest in research'," he says. "Their take is pretty much the same as mine, which is most of the technologies, perhaps all, are still in the research phase."
As a result, progress in productizing neuromorphic computing has been limited to large companies with deep R&D budgets, he said.
But it's not just funding that's getting in the way. Freund argues the scope of the problem for neuromorphics has only gotten larger as the tech has grown more mature.
- Neuromorphic chips 'up to 16 times more energy efficient' for deep learning
- Intel's neurochips could one day end up in PCs or a cloud service
- Brain-like neurochips good for supercomputers, not just AI, says Sandia
- Intel offers Loihi 2 to boffins: A 7nm chip with more than 1m programmable neurons
However, productizing such a chip means solving other problems, like how you get data in and out of the chip effectively.
This isn't a problem unique to neuromorphics by any means. It's one associated with quantum computing and even traditional accelerators, which have accumulated bottlenecks in recent generations due to the speed at which the data can be pre-and post-processed and/or ingested and egressed from the chip, Freund explained.
Finally, there's the issue of developing software that can take advantage of these accelerators.
"It's really going to take a whole community of researchers to solve the programmability problem of neuromorphic computing," Freund says.
Traditional accelerators are good enough
Perhaps the biggest reason that neuromorphic computers haven't taken over is that traditional accelerators are simply getting more powerful and more efficient quickly enough."What they're finding is that platforms, like Nvidia Jetson Orin or some new novel platforms from startups, are solving the problem really quickly. So the need to do something super exotic is continuing to lessen as the state of the art in existing technologies evolves," Freund says. "If you look at what Qualcomm has done with their AI engine, you're talking milliwatts… and what it does when you take a photograph is astounding."
As a result, meaningful problems can be resolved in the power envelope required by existing digital technologies.
While neuromorphics may not be ready to replace traditional accelerators anytime soon, Freund believes the technology will eventually reach the mainstream.
"These things do take time to mature," he says, citing the rise of Arm processors in the datacenter as something that took more than 10 years to achieve. "And that was for CPUs; CPUs are easy compared to things like quantum and neuromorphic computing." ®
Worker122
Regular
“Im interested in the 1000 eyes thoughts on this "8bit standard" document between Intel, Nvidia and Arm.
Is this where Intel and Nvidia know they can't match Akida performance technically, so make Akida "non compliant" buy creating a technical specification to suit themselves. Given the existing market acceptance, depth and customer base of Intel and Nvidia, personally, I think this is an attempt and ploy against Brainchip. aka VHS vs Beta.”
I sorta had that sort of “collusion” Theory a year or so ago in regards to our competitors. They can’t beat us, and along with a few yrs of RandD under their belt decide not to redesign, let’s run with what we got and hope our volume and good name in the marketplace pulls us through.
Anyhow, it’s their funeral in the long run. They cannot ignore us for much longer.
Happy to be a LTH.
cosors
👀
Article pasted below Diana Deca’s comment roasting Loihi
View attachment 16939
View attachment 16940
Bill Gonzalez on LinkedIn: What's holding back neuromorphic computing?
MATRIX - The UTSA AI Consortium for Human Well-Being recently hosted a presentation by Yulia Sandamirskaya with Intel Corporation that discussed some of the…
The article:
Brain-inspired chips promise ultra-efficient AI, so why aren’t they everywhere?
It's not because our AI overlords aren't keen on the idea
Tobias MannMon 12 Sep 2022 // 10:23 UTC
INTERVIEW Every time a chipmaker or researcher announces an advancement in neuromorphics, it's inevitably the same story: a brain-like AI chip capable of stupendous performance-per-watt compared to traditional accelerators.
Intuitively, the idea makes a lot of sense. Our brains are pretty good at making sense of the world, so why wouldn't a chip designed to work like them be good at it too?
Yet, after years of development and the backing of massive tech companies like IBM and Intel, these brain-like chips are still years away from making their way into consumer products.
That hasn't stopped the tech from grabbing headlines over the years. Neuromorphic chips up to 16 times more efficient; brain-like chips potentially poweringfuture supercomputers; Samsung wanting to reverse engineer the brain; IBM recreating a frog brain in silicon. You get the idea.
While the chips show promise, the reality is the field of neuromorphics is still in a very experimental stage, and faces many challenges that must be resolved before they are ready for prime time, explains Karl Freund, principal analyst at Cambrian AI Research, in an interview with The Register.
This may be one of the reasons many of the more promising neuromorphic processors have seemingly stalled.
IBM, for example, hasn't given an update on its True North neuromorphic chips, which are capable of simulating more than a million neurons, in more than four years. SpiNNaker, another promising spiking neural networking processor, received an €8 million (c $8.15 million) grant in 2019 to develop a second-gen chip based on the design. However, the company behind the chip, Dresden, Germany-based SpiNNcloud, is only now getting off the ground.
Intel's Loihi and Loihi 2 processors have come the closest to a commercial launch in so far as Intel has made development boards available to outside researchers alongside its Lava software development kit.
The Department of Energy's Sandia National Laboratories, for example, is exploring how these chips could be used to accelerate supercomputers. In a paper published in the journal Nature Electronics, researchers at Sandia demonstrated how Intel's Loihi chips "can solve more complex problems than those posed by artificial intelligence and may even earn a place in high-performance computing."
Yet, at least as of April, Intel has no plans to productize its Loihi chips anytime soon.
What's the holdup?
So what gives? Why is it these chips, which show such promise in the lab, haven't matured faster given the insatiable demand for AI/ML?
According to Freund, one of the biggest problems comes down to funding.
"I tried to connect some venture capitalists in both neuromorphic and analog [computing] and a fairly consistent response even before the current capital crunch was 'we don't invest in research'," he says. "Their take is pretty much the same as mine, which is most of the technologies, perhaps all, are still in the research phase."
As a result, progress in productizing neuromorphic computing has been limited to large companies with deep R&D budgets, he said.
But it's not just funding that's getting in the way. Freund argues the scope of the problem for neuromorphics has only gotten larger as the tech has grown more mature.
With the first neuromorphic test chips, scientists were primarily focused on getting to a point where they could do useful work, he explains.
- Neuromorphic chips 'up to 16 times more energy efficient' for deep learning
- Intel's neurochips could one day end up in PCs or a cloud service
- Brain-like neurochips good for supercomputers, not just AI, says Sandia
- Intel offers Loihi 2 to boffins: A 7nm chip with more than 1m programmable neurons
However, productizing such a chip means solving other problems, like how you get data in and out of the chip effectively.
This isn't a problem unique to neuromorphics by any means. It's one associated with quantum computing and even traditional accelerators, which have accumulated bottlenecks in recent generations due to the speed at which the data can be pre-and post-processed and/or ingested and egressed from the chip, Freund explained.
Finally, there's the issue of developing software that can take advantage of these accelerators.
"It's really going to take a whole community of researchers to solve the programmability problem of neuromorphic computing," Freund says.
Traditional accelerators are good enough
Perhaps the biggest reason that neuromorphic computers haven't taken over is that traditional accelerators are simply getting more powerful and more efficient quickly enough.
"What they're finding is that platforms, like Nvidia Jetson Orin or some new novel platforms from startups, are solving the problem really quickly. So the need to do something super exotic is continuing to lessen as the state of the art in existing technologies evolves," Freund says. "If you look at what Qualcomm has done with their AI engine, you're talking milliwatts… and what it does when you take a photograph is astounding."
As a result, meaningful problems can be resolved in the power envelope required by existing digital technologies.
While neuromorphics may not be ready to replace traditional accelerators anytime soon, Freund believes the technology will eventually reach the mainstream.
"These things do take time to mature," he says, citing the rise of Arm processors in the datacenter as something that took more than 10 years to achieve. "And that was for CPUs; CPUs are easy compared to things like quantum and neuromorphic computing." ®
Which one commercial?
Dozzaman1977
Regular
What a pathetic article by Tobias mann, not worth the paper it is written onArticle pasted below Diana Deca’s comment roasting Loihi
View attachment 16939
View attachment 16940
Bill Gonzalez on LinkedIn: What's holding back neuromorphic computing?
MATRIX - The UTSA AI Consortium for Human Well-Being recently hosted a presentation by Yulia Sandamirskaya with Intel Corporation that discussed some of the…
The article:
Brain-inspired chips promise ultra-efficient AI, so why aren’t they everywhere?
It's not because our AI overlords aren't keen on the idea
Tobias MannMon 12 Sep 2022 // 10:23 UTC
INTERVIEW Every time a chipmaker or researcher announces an advancement in neuromorphics, it's inevitably the same story: a brain-like AI chip capable of stupendous performance-per-watt compared to traditional accelerators.
Intuitively, the idea makes a lot of sense. Our brains are pretty good at making sense of the world, so why wouldn't a chip designed to work like them be good at it too?
Yet, after years of development and the backing of massive tech companies like IBM and Intel, these brain-like chips are still years away from making their way into consumer products.
That hasn't stopped the tech from grabbing headlines over the years. Neuromorphic chips up to 16 times more efficient; brain-like chips potentially poweringfuture supercomputers; Samsung wanting to reverse engineer the brain; IBM recreating a frog brain in silicon. You get the idea.
While the chips show promise, the reality is the field of neuromorphics is still in a very experimental stage, and faces many challenges that must be resolved before they are ready for prime time, explains Karl Freund, principal analyst at Cambrian AI Research, in an interview with The Register.
This may be one of the reasons many of the more promising neuromorphic processors have seemingly stalled.
IBM, for example, hasn't given an update on its True North neuromorphic chips, which are capable of simulating more than a million neurons, in more than four years. SpiNNaker, another promising spiking neural networking processor, received an €8 million (c $8.15 million) grant in 2019 to develop a second-gen chip based on the design. However, the company behind the chip, Dresden, Germany-based SpiNNcloud, is only now getting off the ground.
Intel's Loihi and Loihi 2 processors have come the closest to a commercial launch in so far as Intel has made development boards available to outside researchers alongside its Lava software development kit.
The Department of Energy's Sandia National Laboratories, for example, is exploring how these chips could be used to accelerate supercomputers. In a paper published in the journal Nature Electronics, researchers at Sandia demonstrated how Intel's Loihi chips "can solve more complex problems than those posed by artificial intelligence and may even earn a place in high-performance computing."
Yet, at least as of April, Intel has no plans to productize its Loihi chips anytime soon.
What's the holdup?
So what gives? Why is it these chips, which show such promise in the lab, haven't matured faster given the insatiable demand for AI/ML?
According to Freund, one of the biggest problems comes down to funding.
"I tried to connect some venture capitalists in both neuromorphic and analog [computing] and a fairly consistent response even before the current capital crunch was 'we don't invest in research'," he says. "Their take is pretty much the same as mine, which is most of the technologies, perhaps all, are still in the research phase."
As a result, progress in productizing neuromorphic computing has been limited to large companies with deep R&D budgets, he said.
But it's not just funding that's getting in the way. Freund argues the scope of the problem for neuromorphics has only gotten larger as the tech has grown more mature.
With the first neuromorphic test chips, scientists were primarily focused on getting to a point where they could do useful work, he explains.
- Neuromorphic chips 'up to 16 times more energy efficient' for deep learning
- Intel's neurochips could one day end up in PCs or a cloud service
- Brain-like neurochips good for supercomputers, not just AI, says Sandia
- Intel offers Loihi 2 to boffins: A 7nm chip with more than 1m programmable neurons
However, productizing such a chip means solving other problems, like how you get data in and out of the chip effectively.
This isn't a problem unique to neuromorphics by any means. It's one associated with quantum computing and even traditional accelerators, which have accumulated bottlenecks in recent generations due to the speed at which the data can be pre-and post-processed and/or ingested and egressed from the chip, Freund explained.
Finally, there's the issue of developing software that can take advantage of these accelerators.
"It's really going to take a whole community of researchers to solve the programmability problem of neuromorphic computing," Freund says.
Traditional accelerators are good enough
Perhaps the biggest reason that neuromorphic computers haven't taken over is that traditional accelerators are simply getting more powerful and more efficient quickly enough.
"What they're finding is that platforms, like Nvidia Jetson Orin or some new novel platforms from startups, are solving the problem really quickly. So the need to do something super exotic is continuing to lessen as the state of the art in existing technologies evolves," Freund says. "If you look at what Qualcomm has done with their AI engine, you're talking milliwatts… and what it does when you take a photograph is astounding."
As a result, meaningful problems can be resolved in the power envelope required by existing digital technologies.
While neuromorphics may not be ready to replace traditional accelerators anytime soon, Freund believes the technology will eventually reach the mainstream.
"These things do take time to mature," he says, citing the rise of Arm processors in the datacenter as something that took more than 10 years to achieve. "And that was for CPUs; CPUs are easy compared to things like quantum and neuromorphic computing." ®
Dhm
Regular
I had to dive in as well as Diana Deca.Article pasted below Diana Deca’s comment roasting Loihi
View attachment 16939
View attachment 16940
Bill Gonzalez on LinkedIn: What's holding back neuromorphic computing?
MATRIX - The UTSA AI Consortium for Human Well-Being recently hosted a presentation by Yulia Sandamirskaya with Intel Corporation that discussed some of the…
The article:
Brain-inspired chips promise ultra-efficient AI, so why aren’t they everywhere?
It's not because our AI overlords aren't keen on the idea
Tobias MannMon 12 Sep 2022 // 10:23 UTC
INTERVIEW Every time a chipmaker or researcher announces an advancement in neuromorphics, it's inevitably the same story: a brain-like AI chip capable of stupendous performance-per-watt compared to traditional accelerators.
Intuitively, the idea makes a lot of sense. Our brains are pretty good at making sense of the world, so why wouldn't a chip designed to work like them be good at it too?
Yet, after years of development and the backing of massive tech companies like IBM and Intel, these brain-like chips are still years away from making their way into consumer products.
That hasn't stopped the tech from grabbing headlines over the years. Neuromorphic chips up to 16 times more efficient; brain-like chips potentially poweringfuture supercomputers; Samsung wanting to reverse engineer the brain; IBM recreating a frog brain in silicon. You get the idea.
While the chips show promise, the reality is the field of neuromorphics is still in a very experimental stage, and faces many challenges that must be resolved before they are ready for prime time, explains Karl Freund, principal analyst at Cambrian AI Research, in an interview with The Register.
This may be one of the reasons many of the more promising neuromorphic processors have seemingly stalled.
IBM, for example, hasn't given an update on its True North neuromorphic chips, which are capable of simulating more than a million neurons, in more than four years. SpiNNaker, another promising spiking neural networking processor, received an €8 million (c $8.15 million) grant in 2019 to develop a second-gen chip based on the design. However, the company behind the chip, Dresden, Germany-based SpiNNcloud, is only now getting off the ground.
Intel's Loihi and Loihi 2 processors have come the closest to a commercial launch in so far as Intel has made development boards available to outside researchers alongside its Lava software development kit.
The Department of Energy's Sandia National Laboratories, for example, is exploring how these chips could be used to accelerate supercomputers. In a paper published in the journal Nature Electronics, researchers at Sandia demonstrated how Intel's Loihi chips "can solve more complex problems than those posed by artificial intelligence and may even earn a place in high-performance computing."
Yet, at least as of April, Intel has no plans to productize its Loihi chips anytime soon.
What's the holdup?
So what gives? Why is it these chips, which show such promise in the lab, haven't matured faster given the insatiable demand for AI/ML?
According to Freund, one of the biggest problems comes down to funding.
"I tried to connect some venture capitalists in both neuromorphic and analog [computing] and a fairly consistent response even before the current capital crunch was 'we don't invest in research'," he says. "Their take is pretty much the same as mine, which is most of the technologies, perhaps all, are still in the research phase."
As a result, progress in productizing neuromorphic computing has been limited to large companies with deep R&D budgets, he said.
But it's not just funding that's getting in the way. Freund argues the scope of the problem for neuromorphics has only gotten larger as the tech has grown more mature.
With the first neuromorphic test chips, scientists were primarily focused on getting to a point where they could do useful work, he explains.
- Neuromorphic chips 'up to 16 times more energy efficient' for deep learning
- Intel's neurochips could one day end up in PCs or a cloud service
- Brain-like neurochips good for supercomputers, not just AI, says Sandia
- Intel offers Loihi 2 to boffins: A 7nm chip with more than 1m programmable neurons
However, productizing such a chip means solving other problems, like how you get data in and out of the chip effectively.
This isn't a problem unique to neuromorphics by any means. It's one associated with quantum computing and even traditional accelerators, which have accumulated bottlenecks in recent generations due to the speed at which the data can be pre-and post-processed and/or ingested and egressed from the chip, Freund explained.
Finally, there's the issue of developing software that can take advantage of these accelerators.
"It's really going to take a whole community of researchers to solve the programmability problem of neuromorphic computing," Freund says.
Traditional accelerators are good enough
Perhaps the biggest reason that neuromorphic computers haven't taken over is that traditional accelerators are simply getting more powerful and more efficient quickly enough.
"What they're finding is that platforms, like Nvidia Jetson Orin or some new novel platforms from startups, are solving the problem really quickly. So the need to do something super exotic is continuing to lessen as the state of the art in existing technologies evolves," Freund says. "If you look at what Qualcomm has done with their AI engine, you're talking milliwatts… and what it does when you take a photograph is astounding."
As a result, meaningful problems can be resolved in the power envelope required by existing digital technologies.
While neuromorphics may not be ready to replace traditional accelerators anytime soon, Freund believes the technology will eventually reach the mainstream.
"These things do take time to mature," he says, citing the rise of Arm processors in the datacenter as something that took more than 10 years to achieve. "And that was for CPUs; CPUs are easy compared to things like quantum and neuromorphic computing." ®
H.E. Pennypacker
Regular
A recap of of the 2022 AI Hardware Summit Conference by Dan McCreary.
Dan is an engineer at Optum.
Unfortunately Brainchip did not get a mention in his recap but his takeaway from his visit to the Brainchip booth from his LinkedIn has been included
“I did stop by the Brainchip booth. It is an impressive edge AI chip that could be used in our home-health monitoring systems. I liked the combination of high-precision and low-power anomaly detection. We would need it packaged into a healthcare solution to be useful in our evaluation labs.”
Dan is an engineer at Optum.
Unfortunately Brainchip did not get a mention in his recap but his takeaway from his visit to the Brainchip booth from his LinkedIn has been included
“I did stop by the Brainchip booth. It is an impressive edge AI chip that could be used in our home-health monitoring systems. I liked the combination of high-precision and low-power anomaly detection. We would need it packaged into a healthcare solution to be useful in our evaluation labs.”
Attachments
Similar threads
