Broad AI landscape

An interesting perspective on the US / China race to AI supremacy. FF has mentioned most of these points in the past but it's a good refreshing read. From the authors point of view, China's heavy hand and restrictions may be limiting their progress as well as the structure of their education system.

Additionally, this article points out how there is a lot of focus on the greater number of Chinese AI research articles and patents when compared to the US. However, this doesn't mean China is winning the AI war. That's the equivalent of thinking Intel has beaten Brainchip in the neuromorphic space because they have spent more time and money and have accumulated more patents. In scenarios like this quality beats quantity.



he global competition between the United States and China continues apace. Technology is rightly seen as providing unique leverage to win this geopolitical race. The U.S. long has been the global technology powerhouse, but not surprisingly, we have heard much about the Chinese government’s ambition to dominate high-tech industries such as 5G telecommunications, autonomous vehicles, blockchain, and semiconductor chips.

In this light, as a horizontal technology that can be applied across all sectors, artificial intelligence (AI) has become a strategic priority and the Chinese focus on superiority in this field is touted as something about which the U.S. should be concerned. Some have gone so far as to conclude that the West has already lost the AI race.

Don’t believe the hype. To be sure, the availability of large amounts of data is at the heart of AI success. It is tempting to think that less-democratic regimes that amass huge amounts of data about their citizens and have scant regard for privacy can develop better AI systems using that data. However, all other things being equal, better and higher quality AI systems emerge from countries with strong data privacy and data protection regulations because AI systems must undergo greater scrutiny during their development and deployment. An example of this can be seen in the United States regarding fair lending practices and consumer protection from credit bureaus. Further, the market for AI is global, and such high-quality AI systems find buyers in other countries as well.

Around the globe, Big Tech’s rising power has resulted in calls for more oversight. In a drastic move that stunned the industry and analysts alike, the Chinese government recently rewrote the rulebook for the country’s technology industry. In effect, China is vacating entire swaths of digital and creative industries, arenas that serve as training grounds and talent factories for other industries. This more restrictive approach may not bode well for China’s AI industry in the long term. China may find itself constrained on the extent of automation and AI in its manufacturing sector — labor-intensive manufacturing remains China’s main strength, and a high degree of automation can result in job losses, labor unrest, and instability.

Meanwhile, there is bipartisan support for AI in the United States. Former President Trump proposed increasing funding for AI development through the National Science Foundation. The National AI Initiative Act of 2020 signaled a sense of urgency and suggested that several federal agencies create a national strategy on artificial intelligence. The Biden administration has formed the Artificial Intelligence Research Resource Task Force to develop a roadmap to foment AI research and spark innovation nationwide. There is draft legislation, at both the state and federal level, to promote responsible use of AI and prevent its misuse.

Strong objections to the use of facial recognition and other AI systems by law enforcement in the U.S., raised by civil liberties advocates, have led some local authorities, such as the City of San Francisco, to ban such systems. To use a Silicon Valley phrase, these debates are “not a bug, but a feature.” They shine a light on the limitations of AI systems and help to set the “rules of the road” for proper use of AI. This will establish the U.S. as a global leader in AI regulation, once lawmakers and regulators do their work. China, meanwhile, has faced strong global criticism for using facial recognition software to monitor and surveil Uyghurs in its Xinjiang region. China has outlined a set of AI ethics principles, but the jury is still out on enforcement and how they function in practice.

The increasing number of AI research papers and patents by Chinese researchers is often cited as proof that China has caught up with the United States in this field. The increased focus is good for the Chinese AI ecosystem, and it will help them solve China-specific problems. But dominance in this emerging strategic industry is not guaranteed. The U.S. has several strategic advantages, including: the strengths of its higher education and research institutes, which attract the best STEM talent from across the world; the largest venture capital ecosystem; and the largest number of technology unicorns (start-ups with private valuations greater than $1 billion).

China is not overtaking the U.S. in artificial intelligence. The current evidence and trajectory paint a clear picture: The conditions for AI to flourish, such as incentives to experiment, freedom to pursue opportunities without restrictions, and the coming guardrails to prevent misuse, favor U.S. leadership. This is still the United States’s game to lose — though maybe both countries could win through collaboration. To solve planet-scale problems such as climate change, we are going to need AI solutions from both competitors.


James Cooper is professor of law and director of International Legal Studies at California Western School of Law in San Diego and a research fellow at Singapore University of Social Sciences.

Kashyap Kompella, a technology industry analyst, is CEO of RPA2AI, a global artificial intelligence advisory firm.
 
  • Like
Reactions: 32 users
This article by is also relevant and interesting. It's a bit long so I've listed some of the highlights below. The chart below is an interesting take and shows US being slightly ahead of China technologically but not in the People (STEM graduates) space. This article indicates a need for the US to partner with other countries to get ahead (similar to how it previously partnered with South Korea for AI research)


To assess the winners and losers in the fulfillment of national AI aspirations, we first assembled a country-level dataset that contained a myriad of details on each country’s technology infrastructure, public and private investments, the number of AI-related patents and conference papers it has produced, and the number of technology-savvy STEM graduates within the country. We determined that the data could be grouped into two overarching factors: a technology-related factor and a people-related factor. We then arrayed each country by its relative achievement on those dimensions (see Figure 1).
1643979363743.png


As shown, the U.S. is positioned in the lower right quadrant—the Technology Prepared quadrant—which reflects the combination of its world-leading technology infrastructure (in the 95th percentile) and a relatively dismal people readiness (45th percentile). While the U.S. leads the world in the technology infrastructure dimension, a number of countries, among them India (92nd percentile), Singapore (88th percentile), Germany (85th percentile), China 72nd percentile) and Russia (48th percentile), score better on the people dimension.

In our analysis of the technology dimension of our scoring, we noted a very strong relationship between the size of a country’s economy and its grade on the technology infrastructure dimension. Simply put, countries with larger economies are more capable of investing substantial sums into the technology necessary to utilize AI.

As such, the U.S. has a people problem—not a spending or technology problem—and in the next section, we offer three options for the U.S. to achieve a position of prominence in AI.

OPTION 1: EXTRACT LESSONS FROM THE U.S. SPACE RACE FOR TALENT DEVELOPMENT

OPTION 2: TAKE A MULTI-NATIONAL CONSORTIUM APPROACH.

OPTION 3: CREATE ROBUST PARTNERSHIP WITH ONE OTHER COUNTRY


Regardless of the option(s) selected, there are four action items that the US needs to do immediately:
  • Action item 1: Educate the U.S. population on the future of AI. It appears that the population of the U.S. either views artificial intelligence as a futuristic utopia or an impending disaster. In actuality, both viewpoints do have a gem of truth in them. But without a realistic view of the world of AI, the populace is unlikely to understand the need for engagement, and the prospects that the industry sector offers. A focused educational campaign can present a balanced view of AI.
  • Action item 2: Create a sense of urgency. The reason that major initiatives like the space race and the Panama Canal succeeded was because the country had a well-articulated sense of urgency to drive both engagement and funding. While we applaud consideration of the Computer Science for All Act of 2021, it does not create the sense of urgency necessary to motivate the populace of the US to better engage. Without such urgency, the motivation to solve AI challenges is likely to wither and fade away much as the drive to adopt the metric system failed in the 1970s/1980s.
  • Action item 3: Raise the profile of STEM work and education. During the space race, astronauts and those working in aerospace were viewed as heroes by their contemporaries for embracing the challenges of space flight. A similar thing needs to be done with STEM workers to elevate their profile and encourage more bright students to study AI-related fields. Several firms have launched STEM-centric initiatives that are a great first step, particularly when these initiatives are focused on women who historically have not been as active in the STEM discipline as men.
  • Action item 4: Closely evaluate potential international partners. Geopolitical relationships are never certain and are subjected to both internal and external pressures. While we understand that these spats will never completely disappear, each spat has the possibility of stalling work and halting valuable momentum. However, the U.S. cannot afford the cost or the time to unilaterally try to solve the problem and must work with its allies to do so.
 
  • Like
Reactions: 25 users
An interesting perspective on the US / China race to AI supremacy. FF has mentioned most of these points in the past but it's a good refreshing read. From the authors point of view, China's heavy hand and restrictions may be limiting their progress as well as the structure of their education system.

Additionally, this article points out how there is a lot of focus on the greater number of Chinese AI research articles and patents when compared to the US. However, this doesn't mean China is winning the AI war. That's the equivalent of thinking Intel has beaten Brainchip in the neuromorphic space because they have spent more time and money and have accumulated more patents. In scenarios like this quality beats quantity.



he global competition between the United States and China continues apace. Technology is rightly seen as providing unique leverage to win this geopolitical race. The U.S. long has been the global technology powerhouse, but not surprisingly, we have heard much about the Chinese government’s ambition to dominate high-tech industries such as 5G telecommunications, autonomous vehicles, blockchain, and semiconductor chips.

In this light, as a horizontal technology that can be applied across all sectors, artificial intelligence (AI) has become a strategic priority and the Chinese focus on superiority in this field is touted as something about which the U.S. should be concerned. Some have gone so far as to conclude that the West has already lost the AI race.

Don’t believe the hype. To be sure, the availability of large amounts of data is at the heart of AI success. It is tempting to think that less-democratic regimes that amass huge amounts of data about their citizens and have scant regard for privacy can develop better AI systems using that data. However, all other things being equal, better and higher quality AI systems emerge from countries with strong data privacy and data protection regulations because AI systems must undergo greater scrutiny during their development and deployment. An example of this can be seen in the United States regarding fair lending practices and consumer protection from credit bureaus. Further, the market for AI is global, and such high-quality AI systems find buyers in other countries as well.

Around the globe, Big Tech’s rising power has resulted in calls for more oversight. In a drastic move that stunned the industry and analysts alike, the Chinese government recently rewrote the rulebook for the country’s technology industry. In effect, China is vacating entire swaths of digital and creative industries, arenas that serve as training grounds and talent factories for other industries. This more restrictive approach may not bode well for China’s AI industry in the long term. China may find itself constrained on the extent of automation and AI in its manufacturing sector — labor-intensive manufacturing remains China’s main strength, and a high degree of automation can result in job losses, labor unrest, and instability.

Meanwhile, there is bipartisan support for AI in the United States. Former President Trump proposed increasing funding for AI development through the National Science Foundation. The National AI Initiative Act of 2020 signaled a sense of urgency and suggested that several federal agencies create a national strategy on artificial intelligence. The Biden administration has formed the Artificial Intelligence Research Resource Task Force to develop a roadmap to foment AI research and spark innovation nationwide. There is draft legislation, at both the state and federal level, to promote responsible use of AI and prevent its misuse.

Strong objections to the use of facial recognition and other AI systems by law enforcement in the U.S., raised by civil liberties advocates, have led some local authorities, such as the City of San Francisco, to ban such systems. To use a Silicon Valley phrase, these debates are “not a bug, but a feature.” They shine a light on the limitations of AI systems and help to set the “rules of the road” for proper use of AI. This will establish the U.S. as a global leader in AI regulation, once lawmakers and regulators do their work. China, meanwhile, has faced strong global criticism for using facial recognition software to monitor and surveil Uyghurs in its Xinjiang region. China has outlined a set of AI ethics principles, but the jury is still out on enforcement and how they function in practice.

The increasing number of AI research papers and patents by Chinese researchers is often cited as proof that China has caught up with the United States in this field. The increased focus is good for the Chinese AI ecosystem, and it will help them solve China-specific problems. But dominance in this emerging strategic industry is not guaranteed. The U.S. has several strategic advantages, including: the strengths of its higher education and research institutes, which attract the best STEM talent from across the world; the largest venture capital ecosystem; and the largest number of technology unicorns (start-ups with private valuations greater than $1 billion).

China is not overtaking the U.S. in artificial intelligence. The current evidence and trajectory paint a clear picture: The conditions for AI to flourish, such as incentives to experiment, freedom to pursue opportunities without restrictions, and the coming guardrails to prevent misuse, favor U.S. leadership. This is still the United States’s game to lose — though maybe both countries could win through collaboration. To solve planet-scale problems such as climate change, we are going to need AI solutions from both competitors.


James Cooper is professor of law and director of International Legal Studies at California Western School of Law in San Diego and a research fellow at Singapore University of Social Sciences.

Kashyap Kompella, a technology industry analyst, is CEO of RPA2AI, a global artificial intelligence advisory firm.

Thanks IDD.

Glad you came across. HC is a bin fire at the moment so it’s great to get some high quality informative posts going. I’ve always valued your input.

I am hopeful that in a few days things will settle down on this site and the FOCUS will return to BRAINCHIPS achievements which is what we are all here for: to learn about the company we have invested in.

To have pleasant and honest dialogue about BRAINCHIP in a forum that isn’t corrupted is what we all want.

Thanks again Zeebot for creating the forum. I hope everyone appreciates the opportunity for a reset and gets back to discussing BRAINCHIP so we can all enjoy the journey!

Cheers
 
  • Like
Reactions: 22 users
Not disagreeing by any stretch but shame some of it reads like a propaganda piece. I mean hilarious, the surveillance state that secretly surveilled everyone including their own citizens, foreign politicians, criticising another country.

Back to topic... posted recently a link to an older report by Deloitte's called "Global artificial intelligence industry whitepaper" provides some further context re China AI.

Can be downloaded here for free on signup:

Some snippets:
ChinaAI-World.png

3.2 China is in the highest demand on chip in the world yet relying heavily on imported high-end chips
AI framework can be broadly divided into three layers. The infrastructure layer includes core AI chip and big data, which is the base for sensing and cognitive
computing power at the technical layer. The application layer is at the top, providing autonomous driving, intelligent robot, smart security and virtual assistant services. As the core of the AI technical chain, AI chip is essential for AI algorithm processing, especially for deep neural networks. Today, the integrated circuit chips imported by China from the U.S. values over USD200 billion, far higher than crude oil imports.
China’s semiconductor industry is thriving at a double-digit rate. AI chip market sees active capital raising activities and an increasing number of M&As. A typical example is the global giant Xilinx’s acquisition of DeePhi Tech, a start-up specializing in machine learning, deep compression, pruning and system-level optimization for neural networks. Tech giants led by Alibaba, Baidu and Huawei have stepped into this battlefield. It is worth noting that Huawei has started AI chip competition in smartphone sector. And Chinese mainland is eating into Taiwan’s semiconductor market shares. Moreover, the growing Chinese mainland market will be a commercial channel for integrated circuit design industry and Chinese mainland companies will continue to invest in Taiwan’s semiconductor industry. However, Chinese semiconductor producers have significantly improved their competitiveness in recent years while the key parts are still largely imported from western countries with the self-sufficiency ratio lower than 20%. Chinese government is greatly concerned about this issue and released several favorable polices to support the growth of the semiconductor industry.
3.3 Chinese robot companies are growing fast with greater efforts in developing key parts and technologies domestically
Robot R&D and application is an important index for measuring the technological development level of one country and the future economic growth will be highly relevant with the robot industry to a large extent. As an important part of building advanced manufacturing industry, robots, both industrial robots for production activities in the industry sector and service robots participating in human’s daily life, are of great significance in seeking new growth engines. With strong support of funds and policies, China’s robot industry is growing rapidly and continues to top the world in terms of growth rate. The market exceeded USD8.74 billion in 201823 at the average growth rate of 29.7% from 2013 to 2018.
3.4 The U.S. has solid strengths in AI’s underlying technology while China is better in speech recognition technology
Natural language processing (NLP): China is catching up with the US NLP technology is able to change the way for human to interact with machines. Massive dark data that are undiscovered in business data sector cannot be applied by means of current technologies, including unstructured data such as short message, files, emails, videos, speeches and pictures. NLP technology will play an important role in commerce.
China is superior in speech recognition
Speech recognition technology can be widely applied in such scenarios as TVs, mobile phones, call centers and smart homes. In this aspect, the average recognition accuracy rate of main platforms, including Baidu, iFLYTEK and Sogou, reaches over 97%. Alibaba’s speech AI technology is superior to Google and is named by MIT as one of 10 Breakthrough Technologies 201924 . And this technology has been penetrated in many scenarios, such as express delivery, customer service and train ticket purchase. During the 2018 11.11 Global Shopping Festival, Ali Xiaomi handled 98% of customer enquiries from the whole platform, equivalent to one-day workload of 700,000 human customer service staff.
Computer vision: A large gap in basic algorithms
…From application layer perspective, there is little difference between China and the U.S. and even China is expected to overtake the U.S. in facial recognition technology. But there is a large gap in basic algorithms between the two countries. China has about 146 companies mostly engaged in the application, including HIKVISION while the U.S. has about 190 companies. China has 1,510 practitioners while the U.S. has more than 4,00025.
ChinaAD.png
 
Last edited:
  • Like
Reactions: 10 users
Not disagreeing by any stretch but shame some of it reads like a propaganda piece. I mean hilarious, the surveillance state that secretly surveilled everyone including their own citizens, foreign politicians, criticising another country.

Back to topic... posted recently a link to an older report by Deloitte's called "Global artificial intelligence industry whitepaper" provides some further context re China AI.

Can be downloaded here for free on signup:

Some snippets:
View attachment 158






View attachment 159
Hi everyone
I have made it clear I am not a fan of the CCP but the Chinese people are just people like us and fit in and take the line of least resistance.

The CCP refuse to sign up to any of the 2030 2050 targets and are comfortable throwing huge amounts of energy at brute force computing. The facial recognition systems in China are said to be more accurate but I pose this question at what energy cost. These systems are administered by the CCP as an instrument of security (control) and they control the sources of energy.

The engineer and data scientist working on these systems will only be concerned with accuracy because that is what the CCP want so energy efficiency and training time and cost are very, very secondary. They are fitting in and giving the customer what it wants.

Whereas the US side of this equation has engineers and data scientists prepared to engage in trade offs and accept failures to offer a diverse range of customers an overall better product.

This is what we see in the microcosm that is AKIDA technology where dropping to 1 to 4 bit activations trades off some accuracy to save on processing and power. This trade off on accuracy would not even be considered in a CCP style environment.

It is just plain old human nature. You see it in workforces of all types everywhere where you have a boss or supervisor not up to the job who imposes idiot processes and will not tolerate having the idiocy pointed out workers take the anything for a quite life approach just to keep the boss happy.

If Peter van der Made and Anil Mankar had been developing AKIDA for the CCP the likelihood would be that Studio would have been the outcome not AKIDA.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Fire
Reactions: 20 users
Not disagreeing by any stretch but shame some of it reads like a propaganda piece. I mean hilarious, the surveillance state that secretly surveilled everyone including their own citizens, foreign politicians, criticising another country.

Back to topic... posted recently a link to an older report by Deloitte's called "Global artificial intelligence industry whitepaper" provides some further context re China AI.

Can be downloaded here for free on signup:

Some snippets:
View attachment 158






View attachment 159
Hi Rick,

Thanks for posting the Deloite article. It’s good to have it posted again. It helped in my decision making when I was first looking to buy BRAINCHIP. Not just that it was no. 2 out of 50 Or it’s 16000% possible price rise. DELOITES is one of the most respect firms of the ”Big Four” who provide financial advice services. But the fact the chip could also cross into various uses, e.g. edge, security, medical, autonomous vehicles etc. I thought that gave the company a greater chance of success in commercialisation.

The biggest risk to me was the chip failing at the production and testing stage, (which was a realisitic posibility). Once that was successful the company has been significantly de-risked. Now Brainchips only issue is to commericalise the worlds first neuromorphic chip which is significantly cheaper and better than competitors research chip. Pardon the pun but in my opinioin it is now a “No Brainer.”

I haven’t downloaded the article today but from memory it was produced in 2019 and I guesstimated the price at around 6c at the time, so a 16000% rise was going to put the share price at $9.60. I’d be pretty happy with that In the next couple of years. Obviously times have changed since the article was produced so as time moves forward so does it‘s accuracy. Pleasingly some posters have suggested the SP will be north of the $9.60 which I won’t disagree with.

Have a good weekend all!
 
  • Like
Reactions: 21 users
The two inventors of the SNN cybersecurity licenced by Brainchip together and independently just keep turning out papers and patents for all sorts of things from monitoring endangered fish supplies to monitoring forest fires and ways to secure communications all of which have that neuromorphic computing SNN vide going on but nothing concrete however there was some interesting stuff posted where NATO and the European Space Agency was concerned that could be connected back to NASA and Brainchip with Democratis University of Thrace thrown in because of their tie up with the Hellenic Defence Forces and NATO.

Given Rob Telson's comment in that podcast he did recently that they are working on vision with NASA and other stuff which he is not allowed to talk about it all seems quite plausible suffice to say using Rob Telson's words we can only see the tip of the iceberg where this NASA DARPA collaborations are concerned though the US Airforce project possibly gives a look in the back door at what is going on. Again pick up on Rob Telson these are exciting times.

Tony Dawe did say to me that over in Perth they are very alert to cyber infiltration as this is an ever present and realistic risk from an external power.

Lots of secret squirrel nuts being stored over there in my opinion but remember I am an anonymous poster on a really great site but still DYOR
FF

AKIDA BALLISTA
 
  • Like
Reactions: 20 users
Couple year old infographic but paints an easy to visualise AI Chip landscape.

At the time of its creation there were only 5 Neuromorphic start ups inclusive of BRN.

Haven't looked to see what inroads the others have made in the same timeframe that BRN has made significant industry connections.

AI_Chip_Landscape Mark Up.png
 
  • Like
  • Fire
Reactions: 18 users
For those wanting to really understand the semiconductor landscape, this slideshow (297 pages) is worth a wander through painting the picture for the relevant players, relationships and structures.

From Steve Blank - Lecturer. Gordian Knot Canter for National Security Innovation

BRN one mention page 97 as part of the TSMC stable.

 
  • Like
Reactions: 11 users
The U.S. DoD want to be the leaders rather than collaborators with china..
For all things Did and AI...
Interesting someone on HC posted and highlighted in Brainchip job description ' experience with DoD..



This is not the first time DoD was a requirement for Brainchip positions. I can say without giving away any secrets that there are a group of BRN employees who are in a silo that do not exchange details of what they are doing with the other employees regarding NASA.

This of course makes perfect sense in a situation where they are dealing in real time and not building a unit which you hope might one day deal in real time with these types of clients. As Rob Telson said there are things they are not allowed to take about.
My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Fire
Reactions: 15 users

Diogenese

Top 20
Couple year old infographic but paints an easy to visualise AI Chip landscape.

At the time of its creation there were only 5 Neuromorphic start ups inclusive of BRN.

Haven't looked to see what inroads the others have made in the same timeframe that BRN has made significant industry connections.

View attachment 393
Hi FMF,

This is a very useful resource.

The link at the bottom of the Startup Worldwide is:
https://basicmi.github.io/AI-Chip/

This site provides somewhat dated info on a whole gamut of AI companies, but it serves as a quick reference guide. (Who was that bolke over at the other place who kept trying to find competitors for Akida?)

Here is the info on BrainChip:

1644384474010.png
 
  • Like
Reactions: 6 users

Fulltank

Emerged
Hi everyone
I have made it clear I am not a fan of the CCP but the Chinese people are just people like us and fit in and take the line of least resistance.

The CCP refuse to sign up to any of the 2030 2050 targets and are comfortable throwing huge amounts of energy at brute force computing. The facial recognition systems in China are said to be more accurate but I pose this question at what energy cost. These systems are administered by the CCP as an instrument of security (control) and they control the sources of energy.

The engineer and data scientist working on these systems will only be concerned with accuracy because that is what the CCP want so energy efficiency and training time and cost are very, very secondary. They are fitting in and giving the customer what it wants.

Whereas the US side of this equation has engineers and data scientists prepared to engage in trade offs and accept failures to offer a diverse range of customers an overall better product.

This is what we see in the microcosm that is AKIDA technology where dropping to 1 to 4 bit activations trades off some accuracy to save on processing and power. This trade off on accuracy would not even be considered in a CCP style environment.

It is just plain old human nature. You see it in workforces of all types everywhere where you have a boss or supervisor not up to the job who imposes idiot processes and will not tolerate having the idiocy pointed out workers take the anything for a quite life approach just to keep the boss happy.

If Peter van der Made and Anil Mankar had been developing AKIDA for the CCP the likelihood would be that Studio would have been the outcome not AKIDA.

My opinion only DYOR
FF

AKIDA BALLISTA
 
This article by Andrew Ng (one of the pioneers of machine learning) is very relevant to Brainchip's ambitions.
In it he predicts a shift in industry focus from a small number of huge machine learning models to many smaller ones. He also talks about the need to empower everyday employees to be able to manage the machine learning systems at their own workplaces because there simply aren't enough machine learning engineers. It's worth mentioning he started his own company Landing AI to assist with similar issues in the computer vision space. I've highlighted a few key snippets.


ANDREW NG HAS SERIOUS STREET CRED in artificial intelligence. He pioneered the use of graphics processing units (GPUs) to train deep learning models in the late 2000s with his students at Stanford University, cofounded Google Brain in 2011, and then served for three years as chief scientist for Baidu, where he helped build the Chinese tech giant’s AI group. So when he says he has identified the next big shift in artificial intelligence, people listen. And that’s what he told IEEE Spectrum in an exclusive Q&A.

You often talk about companies or institutions that have only a small amount of data to work with. How can data-centric AI help them?

Ng: You hear a lot about vision systems built with millions of images—I once built a face recognition system using 350 million images. Architectures built for hundreds of millions of images don’t work with only 50 images. But it turns out, if you have 50 really good examples, you can build something valuable, like a defect-inspection system. In many industries where giant data sets simply don’t exist, I think the focus has to shift from big data to good data. Having 50 thoughtfully engineered examples can be sufficient to explain to the neural network what you want it to learn.

How do you deal with changing needs? If products change or lighting conditions change in the factory, can the model keep up?

Ng: It varies by manufacturer. There is data drift in many contexts. But there are some manufacturers that have been running the same manufacturing line for 20 years now with few changes, so they don’t expect changes in the next five years. Those stable environments make things easier. For other manufacturers, we provide tools to flag when there’s a significant data-drift issue. I find it really important to empower manufacturing customers to correct data, retrain, and update the model. Because if something changes and it’s 3 a.m. in the United States, I want them to be able to adapt their learning algorithm right away to maintain operations.

In the consumer software Internet, we could train a handful of machine-learning models to serve a billion users. In manufacturing, you might have 10,000 manufacturers building 10,000 custom AI models. The challenge is, how do you do that without Landing AI having to hire 10,000 machine learning specialists?

So you’re saying that to make it scale, you have to empower customers to do a lot of the training and other work.

Ng: Yes, exactly! This is an industry-wide problem in AI, not just in manufacturing. Look at health care. Every hospital has its own slightly different format for electronic health records. How can every hospital train its own custom AI model? Expecting every hospital’s IT personnel to invent new neural-network architectures is unrealistic. The only way out of this dilemma is to build tools that empower the customers to build their own models by giving them tools to engineer the data and express their domain knowledge. That’s what Landing AI is executing in computer vision, and the field of AI needs other teams to execute this in other domains.

Is there anything else you think it’s important for people to understand about the work you’re doing or the data-centric AI movement?

Ng: In the last decade, the biggest shift in AI was a shift to deep learning. I think it’s quite possible that in this decade the biggest shift will be to data-centric AI. With the maturity of today’s neural network architectures, I think for a lot of the practical applications the bottleneck will be whether we can efficiently get the data we need to develop systems that work well. The data-centric AI movement has tremendous energy and momentum across the whole community. I hope more researchers and developers will jump in and work on it.
 
  • Like
  • Fire
Reactions: 15 users
Hi Frogstar

Having the enormous ego I do and having posted about Renesas being engaged in both Health and Defence applications over in that other place when they were announced on 23 December, 2020 I am sure every shareholder can recite the exact wording of what you just posted. LOL.

In reality I suspect that because of the way that other place operated and how the trolls used to descend and criticise the former CEO's speech, tired appearance, un-ironed shirts, lack of a tie, lack of jackets, hair cuts, the lighting of the room, its paint colour and any other thing they could lay their grubby little hands on for days and weeks on end when he presented a significant price sensitive announcement such as Renesas was and remains in circumstances where Renesas was largely unknown here in Australia that you are right to suspect that it is not common knowledge. Then when they finished with these matters they then got into questioning why he had not disclosed the quantum of the licence fee being paid. Then they got into saying that because nothing had happened Renesas had decided it was a dud and Brainchip were not releasing the fact that it was a failed deal. It went on and on and on.

On more than one occasion at the other place I encouraged others to dig deep into Renesas's web site because Renesas was and is a major coup for the company. When I tried to post something like the linked article you have tonight it would be moderated quickly as off topic. If I only posted a link then those who were not logged in as members could not open the link and I suspect given the reputation of HC many were loath to open links even if they could.

As it is not a very long document I would suggest that you copy and past the main body of the document here so even those who do not like to open links can read the salient points and as you explore all the areas Renesas covers you post similar links and extracts giving the flavour of what is going on at Renesas.

I am sure as you are obviously exploring Renesas you will like myself have started to wonder why its purchase of an IP licence (Brainchip's first IP sale) and that it is soon to market an MCU containing AKIDA is not capturing the imagination of the whole investor market here in Australia. It has been quite sometime since I read the numbers but I think Renesas regularly sells 40 million MCU's each financial year. This number should be checked before telling anyone as it might have been larger.

Great research generously shared. Many thanks.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Fire
Reactions: 13 users

Brubaker

Regular
All good FF, follow the money. Who spends the most us dod 😎 unfortunately. But it is highly beneficial what it is. Brain chip would not survive if it didn't go down this path imo.. door bells aren't gonna ring it.
Does it matter really at the end of the day who buys now? I got barred ages ago on HC. My least frequented post was BRN as I felt my money was in good hands. I had a short speak in the charts forum , oversight from hc imagine. Lol.
Threw a few dots in there before I got banned possibly from get rich 0 report ? Anyway I give no fucks except not going all in. Still happy my average is 16c. Imo nothing will make a differrence except for time. I hope the markets/ banks don't collapse as for me I see that as our riches may turn to ditches...
On sharing details I'm more a headline guy. Maybe meditating might favour me to delve into annunciating the detail 🤔 i think youre and others best at that... i like throwing in the basil.
Much 💙 and appreciatition FF
Anyway, good to end a Friday in the green......cheers all
 
  • Like
Reactions: 4 users
In trying to educate myself about Ai I came across this site that talks about Ai adoption. It highlights to me that it's still in its early stages of growth, it also appears that a lot of it is cloud based and cybersecurity fears may be holding back adoption as well as probably in house technical expertise to navigate the use of it.
This is where I see how good Akida will be being able to work both on the edge and in the cloud as well as being superior with cybersecurity, yes?
I'm looking forward to reading more about Brainchip progress with their software arm..🤓
I'm guessing Ai will be the norm in the coming years to be competitive.

There's some quick short vids in the link too as well as they reports in the menu.

Remember as Anil Mankar regularly points out 99.9% of those claiming to be doing Artificial Intelligence have shanghaied this term and have redefined it to no longer mean Artificial Intelligence but deep training.

They are not in anyway shape or form being intelligent in the true sense of this word.

This shanghaiing of the term Artificial Intelligence means that you need to now speak about Artificial General Intelligence to describe what Brainchip is developing.

They are on the road to an artificial brain that can learn and extrapolate from limited data. The demonstrations of AKD1000 doing hand gesture recognition in the following presentation is what Brainchip is developing in its most primitive form. You will see Peter van der Made teach AKD1000 a hand holding up 1 finger, then 2 fingers, then 3 fingers then asks it to learn and shows it a hand holding up 5 fingers and AKD1000 extrapolates from the earlier teaching that it has been showing a hand with 5 fingers and identifies it correctly. It has never seen the image before. It has never been trained with other than the three different images of 1,2 & 3 fingers yet it identifies 5 fingers correctly. Not only this but each image is moving about as the hand is being waved in front of the DVS camera.


Peter van der Made speaks about how a child learns and how a child might see two or three cats or images of a cat and there after will identify cats in all different circumstances and of all different breeds. AKIDA1000 is doing a very primitive form of this type of intelligent learning.

AKD2000 will advance this primitive form by adding LSTM and as a result will learn to extrapolate from a known event a series of further events which might follow. Peter van der Made talks about a ball rolling out across the path of a vehicle from between parked cars and how a child might be following. He refers to a plastic bag blowing across the road in front of a car and the car having the ability to recognise that it is a plastic bag and that it does not need to be avoided as a collision with the plastic bag will not be of consequence.

Then AKD3000 will have a functioning artificial cortical column which is the ultimate goal and when Brainchip will all going to plan have achieved something very close to true Artificial General Intelligence. If you look at the date that Emeritus Professor Alan Harvey joined the SAB at Brainchip it was just before that, that Peter van der Made mentioned they had a working artificial cortical column on the bench at the Perth Research Centre. As Peter van der Made has said, "We are only just getting started."

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Fire
Reactions: 15 users

Diogenese

Top 20
Hi Frogstar

Having the enormous ego I do and having posted about Renesas being engaged in both Health and Defence applications over in that other place when they were announced on 23 December, 2020 I am sure every shareholder can recite the exact wording of what you just posted. LOL.

In reality I suspect that because of the way that other place operated and how the trolls used to descend and criticise the former CEO's speech, tired appearance, un-ironed shirts, lack of a tie, lack of jackets, hair cuts, the lighting of the room, its paint colour and any other thing they could lay their grubby little hands on for days and weeks on end when he presented a significant price sensitive announcement such as Renesas was and remains in circumstances where Renesas was largely unknown here in Australia that you are right to suspect that it is not common knowledge. Then when they finished with these matters they then got into questioning why he had not disclosed the quantum of the licence fee being paid. Then they got into saying that because nothing had happened Renesas had decided it was a dud and Brainchip were not releasing the fact that it was a failed deal. It went on and on and on.

On more than one occasion at the other place I encouraged others to dig deep into Renesas's web site because Renesas was and is a major coup for the company. When I tried to post something like the linked article you have tonight it would be moderated quickly as off topic. If I only posted a link then those who were not logged in as members could not open the link and I suspect given the reputation of HC many were loath to open links even if they could.

As it is not a very long document I would suggest that you copy and past the main body of the document here so even those who do not like to open links can read the salient points and as you explore all the areas Renesas covers you post similar links and extracts giving the flavour of what is going on at Renesas.

I am sure as you are obviously exploring Renesas you will like myself have started to wonder why its purchase of an IP licence (Brainchip's first IP sale) and that it is soon to market an MCU containing AKIDA is not capturing the imagination of the whole investor market here in Australia. It has been quite sometime since I read the numbers but I think Renesas regularly sells 40 million MCU's each financial year. This number should be checked before telling anyone as it might have been larger.

Great research generously shared. Many thanks.

My opinion only DYOR
FF

AKIDA BALLISTA


" ... I encouraged others to dig deep into Renesas's web site ..."

... and so began a happy correspondence between ManChild , the ogre and Ella.
That's pretty clever to be able to extrapolate like that..🤓
I just had a quick look at this to get an understanding of what a cortical column is/does... that's mad if they can make that happen 😳 😎
Thanks FF for your insights 🙏
Hi Frogstar,

This is a fascinating paper, which I have skimmed through, so I don't have the fine detail, but my understanding is that one of the things the authors are proposing that, in order to recognize an object, the sensor needs to move relative to the object to take "samples" from different aspects or different points on the object. They also propose that the sensor information needs to contain both object information and location information.

Most people are familiar with the X-Y coordinate graph in which the position of a point is defined by its distance from the origin (X = 0, Y = 0) along the X axis and the Y axis. There is an alternative coordinate system, the radial coordinate system, where the position of a point is defined by the angle above the X axis and the distance along a line between the origin of the graph and the point.

Thinking about Lidar, Lidar provides radial coordinate information. The reflected light indicates that there is an object there, and the pixel location on the light detector on which the reflected light is focused by a lens provides the angle information of the bit of the object from which the spot of Lidar light was reflected. [A light ray entering a lens and passing through the centre of the lens onto the pixel array traces a straight line from the object to the corresponding pixel, whereas rays parallel to that centre ray and which are off centre relative to the lens are focused to the same pixel so ray tracing is merely drawing a line from a point on the object through the centre of the lens to the pixel array].

The Lidar transmitter scans the light beam across the object in a manner similar to the way the electron beam was scanned across the screen of the old vacuum tube TVs, and the light is reflected at slightly different angles from the various points on the object via the lens to the pixel array, so Lidar provides the relative movement across the object to which the authors refer.

Now what was it that LdN said about Lidar?

The authors then discuss the role of what I will refer to as "cross-talk" within and between cortical columns in facilitating learning, but I haven't really looked at how they achieve the necessary cross-talk in their physical model. That said, I guess they would cross-connect a few of the outputs from neurons in different columns to achieve this.

This may or may not be analogous to or somewhat similar to the moving window used in CNNs which Akida is so adept at converting to SNNs.
 
  • Like
  • Fire
Reactions: 9 users
Top Bottom