BRN Discussion Ongoing

[cassip]

So Ford, VW, Intel, and Tesla have all attempted to bypass the intermediate steps to autonomous driving (AD) and fallen short - you have to crawl before you can walk.

It's not that the billions of dollars they invested has been totally wasted - you should always learn from your mistakes. But both Ford and Intel have recently sold their AD subsidiaries. The "recently" is significant in that it shows that these large companies are feeling the pinch now. As Sean intimated, it's a tough economic environment at the moment.

Remember that they were attempting AD without Valeo's SCALA. I haven't researched Mobileye's technology in depth, but this sample looks unpromising from a SNN point of view:

US2022222317A1 APPLYING A CONVOLUTION KERNEL ON INPUT DATA

View attachment 20521

[0028] Both application processor 180 and image processor 190 can include various types of processing devices. For example, either or both of application processor 180 and image processor 190 can include one or more microprocessors, preprocessors (such as image preprocessors), graphics processors, central processing units (CPUs), support circuits, digital signal processors, integrated circuits, memory, or any other types of devices suitable for running applications and for image processing and analysis. In some embodiments, application processor 180 or image processor 190 can include any type of single or multi-core processor, mobile device microcontroller, central processing unit, or other type of processor. Various processing devices can be used, for example including processors available from manufacturers (e.g., Intel®, AMD®, etc.), and can include various architectures (e.g., x86 processor, ARM®, etc.).

[0054] A convolution neural network includes an input layer, an output layer, as well as multiple hidden layers. The hidden layers of a CNN typically include a series of convolution layers that convolve with a multiplication or other dot product. The activation function is commonly a RELU layer, and is subsequently followed by additional convolutions such as pooling layers, fully connected layers and normalization layers, referred to as hidden layers because their inputs and outputs are masked by the activation function and final convolution. The final convolution, in turn, often involves backpropagation in order to more accurately weight the end product.

Argo AI seems likewise deficient:
US2022301099A1 SYSTEMS AND METHODS FOR GENERATING OBJECT DETECTION LABELS USING FOVEATED IMAGE MAGNIFICATION FOR AUTONOMOUS DRIVING

View attachment 20524


[0044] The machine learning model for generating a saliency map may be generated and/or trained using any now or hereafter known techniques such as, without limitation, kernel density estimation (KDE) and convolution neural network (CNN), both of which are differentiable and the parameters can be learned through the final task loss. In KDE, the system may use bounding box centers as the data points that have a bandwidth proportional to the square root of the area of the bounding box. In CNN, the system may represent the bounding boxes as an N×4 matrix, where N is a fixed maximum value for the number of bounding boxes. If there are less than N objects, the input may be zero-padded to this dimension. Once a model has been generated, the system may also apply the model to all bounding boxes in a training dataset to obtain a dataset-wide prior.

[0076] During operations, information is communicated from the sensors to the on-board computing device 720 . The on-board computing device 720 can (i) cause the sensor information to be communicated from the mobile platform to an external device (e.g., computing device 101 of FIG. 1) and/or (ii) use the sensor information to control operations of the mobile platform.

So, not only do they not have SCALA, they don't seem to be aware of Akida.

Could it be possible that both Argo and Mobileye had very good algorithms, but they were implementing them on 20th century processors?

That's like getting into the ring against Muhammad Ali with your shoe laces tied together.

If Mobileye is using its old tech, figures and outlook are all the more promising for BRNs future revenue possibilities. If they are aware of Akida (because of Intel) maybe they changed something in their development?
Good morning and good weekend to all
Regards Cassip

Intel's Mobileye sees $17 bln in assisted-driving product revenues by 2030

Intel Corp's self-driving tech unit Mobileye Global Inc said on Thursday it sees more than $17 billion in revenues for its advanced driver assistance systems (ADAS) products by 2030.

www.reuters.com

 

[HopalongPetrovski]

Was thinking in the slighltly less homicidal and more plutonic range.

I'm sure PF will be able to keep them in line.
And as long as you keep them satiated you'll have no trouble.
And yes, they are the plutonium range and have a battery half life of ten thousand years, so no problemo there either, old chap.
Might be best to avoid any cups of tea they offer though, wot!
Just keep them clean and well lubricated.

 

[Fact Finder]

I found the following paper however I cannot access it fully. It may be of no interest but when you open the link you will find an intriguing NASA reference and one of the authors has a recent link to Carnegie Mellon.

Anyway if you have access or can access who knows it might be profitable:

https://www.sciencedirect.com/science/article/abs/pii/S0167739X22004344

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Fact Finder]

The following paper is interesting in so far it talks about the final AKIDA technology destination but not actually AKIDA:



My opinion only DYOR
FF

AKIDA BALLISTA

 

[ndefries]

If Mobileye is using its old tech, figures and outlook are all the more promising for BRNs future revenue possibilities. If they are aware of Akida (because of Intel) maybe they changed something in their development?
Good morning and good weekend to all
Regards Cassip

Intel's Mobileye sees $17 bln in assisted-driving product revenues by 2030

Intel Corp's self-driving tech unit Mobileye Global Inc said on Thursday it sees more than $17 billion in revenues for its advanced driver assistance systems (ADAS) products by 2030.

www.reuters.com

I wonder if we will partner with MobileEye? Stranger things have happened. They are now listed with a value of 25b USD. I do wonder what Shareman thinks of that. If Akida can show you be better and more versatile you can see how much more growth our share price has.

 

[Rise from the ashes]

Easy one mate - Samsung!

Also year if you know but I'll still give you the subscription anyway if you don't know. Will sort it out later tonight just winding down with some beers.

 

[Food4 thought]

Here is a silly question kind of.

It is Brainchips stated ambition to make AKIDA technology ubiquitous.

Ubiquitous means: : existing or being everywhere at the same time

What if Brainchip only achieves 1% of ubiquity what will that mean in dollars and cents?

The Ai semiconductor market is in its ubiquity said by Cathy Wood to be worth trillions of dollars.

This silly question gives pause for thought I believe.

My opinion only DYOR
FF

AKIDA BALLISTA

That’s chook feed ……
A billion only

 

[Sirod69]

Markus Schäfer• Follower:inMitglied des Vorstands der Mercedes-Benz Group AG, Chief Technology Officer, Entwicklung & Einkauf
1 Std. • vor 1 Stunde

At Mercedes-Benz, technology is a key driver of desire. We are focused on making life easier by turning automated driving into reality, substantially improving in-car entertainment and giving our customers back precious time.

Here at #CES, we are introducing Automatic Lane Change to our Level 2 advanced driver-assistance systems in North America. DRIVE PILOT, our Level 3 system for conditionally automated driving, is already available in Germany and now we have applied for certification in the states of California and Nevada.

The Nevada Department of Motor Vehicles (DMV) has approved our application and they are now preparing the certificate of compliance that will be issued within the next two weeks. With that, Mercedes-Benz will be the first OEM to have a Level 3 system to offer in the US – and we are optimistic that California will follow soon.*

Our new Burmester audio system boasts 1,750 watts powering up to 31 speakers and eight exciters that transfer vibrations, creating an unmatched acoustic environment where you hear and feel every sound.

We’re also creating perfect harmony between the sound experience and premium content in collaboration with Apple Music, Universal Music Group and Dolby Laboratories. “Approved in a Mercedes-Benz” recordings make our vehicles the best place to listen to premium high-fidelity music. We recently added ZYNC to ensure the best quality streaming content is available directly in your Mercedes-Benz vehicle!

Finally, we announced yesterday our development of the Mercedes-Benz High Power Charging Network, to accelerate the adoption of electric mobility and drive the world toward a more sustainable future.

Follow us Mercedes-Benz AG and join us at CES (www.ces.tech) to see how we are transforming technology and desire.

 

[Getupthere]

 

[Sirod69]

................

A Few Words About AI...It Is Growing Up

We asked the CTO of an artificial intelligence company, Tim Llewellynn of NVISO SA to explain what these new AI applications are capable of (an example is OpenAI’s recently released ChatGPT).

Mr. Llewellynn advised, “Think of it as the equivalent of “compound interest” for AI development. As the generative AI gets betters - it not only reduces AI development costs but also leads to better AI results. E.g.: interest on interest. Every time you iterate the cycle, you get compounding interest. The more you iterate the greater the advantage you get. Everyone is still trying to get their head around the classical “simple interest” schematic for AI development so there is a first mover advantage here for smaller and nimble startups.

So yes AI applications directly translates into a) fewer developers required; b) higher quality outputs and c) faster development cycles.”

Still a long way to grow, but AI is growing up quickly.

AVFTTE 20230106 — Territory Funds Management

www.territoryfundsmanagement.com.au

 

[MadMayHam | 合氣道]

Finally... proof that ChatGPT has surpassed human (well, at least some) intelligence.

 

[Fact Finder]

Markus Schäfer• Follower:inMitglied des Vorstands der Mercedes-Benz Group AG, Chief Technology Officer, Entwicklung & Einkauf
1 Std. • vor 1 Stunde

At Mercedes-Benz, technology is a key driver of desire. We are focused on making life easier by turning automated driving into reality, substantially improving in-car entertainment and giving our customers back precious time.

Here at #CES, we are introducing Automatic Lane Change to our Level 2 advanced driver-assistance systems in North America. DRIVE PILOT, our Level 3 system for conditionally automated driving, is already available in Germany and now we have applied for certification in the states of California and Nevada.

The Nevada Department of Motor Vehicles (DMV) has approved our application and they are now preparing the certificate of compliance that will be issued within the next two weeks. With that, Mercedes-Benz will be the first OEM to have a Level 3 system to offer in the US – and we are optimistic that California will follow soon.*

Our new Burmester audio system boasts 1,750 watts powering up to 31 speakers and eight exciters that transfer vibrations, creating an unmatched acoustic environment where you hear and feel every sound.

We’re also creating perfect harmony between the sound experience and premium content in collaboration with Apple Music, Universal Music Group and Dolby Laboratories. “Approved in a Mercedes-Benz” recordings make our vehicles the best place to listen to premium high-fidelity music. We recently added ZYNC to ensure the best quality streaming content is available directly in your Mercedes-Benz vehicle!

Finally, we announced yesterday our development of the Mercedes-Benz High Power Charging Network, to accelerate the adoption of electric mobility and drive the world toward a more sustainable future.

Follow us Mercedes-Benz AG and join us at CES (www.ces.tech) to see how we are transforming technology and desire.

Anyone remember the competition Mercedes ran and Neuromorphic won and still we have not heard a word.

The reason Neuromorphic won was clearly because of TSEx.

I suggested at the time that this distorted what the Mercedes Benz focus groups had predicted and as a result Mercedes may solve the problem by delaying the Neuromorphic presentation till when they originally anticipated delivering it.

This seems to be playing out so if I was right and they expected it to come in third then I am thinking February, 2023 for it to be released.

My opinion only DYOR
FF

AKIDA BALLISTA

 

[MadMayHam | 合氣道]

FF I have access but no mention of Brainchip of Akida, it's more of an algorthmic model called Random Neural Network. I don't want to run afoul of copyright laws so am not going to link the document here.

"
Our approach is based on Locality Sensitive Hashing (LSH) [58], an algorithmic technique which uses hash functions to map input data points to buckets. These functions are designed to map similar inputs to the same bucket. LSH is typically used for data clustering [59] and nearest neighbor search [60] but can also be used for anomaly detection [61]. Since anomalous data points differ substantially from normal inputs, they will be mapped to different buckets. An anomaly score is then obtained by counting how many normal training samples are allocated to the same bucket as the test sample. Different families of hash functions such as p-stable hashing [62] and randomized trees [63] have been proposed before while other, more advanced methods use properties of the training data to select suitable hash functions [64].

A major benefit of LSH-based approaches for anomaly detection is that they have a limited computational and memory footprint."

I found the following paper however I cannot access it fully. It may be of no interest but when you open the link you will find an intriguing NASA reference and one of the authors has a recent link to Carnegie Mellon.

Anyway if you have access or can access who knows it might be profitable:

https://www.sciencedirect.com/science/article/abs/pii/S0167739X22004344

My opinion only DYOR
FF

AKIDA BALLISTA

 

[Fact Finder]

FF I have access but no mention of Brainchip of Akida, it's more of an algorthmic model called Random Neural Network. I don't want to run afoul of copyright laws so am not going to link the document here.

"
Our approach is based on Locality Sensitive Hashing (LSH) [58], an algorithmic technique which uses hash functions to map input data points to buckets. These functions are designed to map similar inputs to the same bucket. LSH is typically used for data clustering [59] and nearest neighbor search [60] but can also be used for anomaly detection [61]. Since anomalous data points differ substantially from normal inputs, they will be mapped to different buckets. An anomaly score is then obtained by counting how many normal training samples are allocated to the same bucket as the test sample. Different families of hash functions such as p-stable hashing [62] and randomized trees [63] have been proposed before while other, more advanced methods use properties of the training data to select suitable hash functions [64].

A major benefit of LSH-based approaches for anomaly detection is that they have a limited computational and memory footprint."

Thank you that’s great. Can you complete the sentence regarding NASA just so I/we know what this part relates too.

Regards
FF

AKIDA BALLISTA

 

[Jumpchooks]

New product at CES by Koehler. They should install a VOC sensor linked to Akida to detect disease & improve diet. Then Alexa would say something like "that's foul, go see the doctor" or "you have to eat more greens & drink less booze". Possibly standard feature in all dunnies in 10 years.

View attachment 26336

A toilet with Alexa inside

One minute we’re being told not to take Alexa into the bedroom, now we’re being encouraged to take Alexa into the toilet.

American manufacturer Kohler will show off a $16,900 toilet which has Alexa inside. So you can get the weather report while sitting on the throne.

The Numi 2.0 also comes with LED lights that illuminate the bowl, jets that can power wash all of your bits and bobs and a mist to help disguise smells.

And that perennial problem of people leaving the toilet seat up is a problem no more. The Numi can automatically lower the seat after every use so the interior of the bowl is banished from public view.

The Numi has been around as a concept for a while but the manufacturer says it is now ready to be installed.



Volatile organic compounds (VOCs) emitted from stool are the components of the smell of stool representing the end products of microbial activity and metabolism that can be used to diagnose disease. Despite the abundance of hydrogen, carbon dioxide, and methane that have already been identified in human flatus, the small portion of trace gases making up the VOCs emitted from stool include organic acids, alcohols, esters, heterocyclic compounds, aldehydes, ketones, and alkanes, among others. These are the gases that vary among individuals in sickness and in health, in dietary changes, and in gut microbial activity. Electronic nose devices are analytical and pattern recognition platforms that can utilize mass spectrometry or electrochemical sensors to detect these VOCs in gas samples. When paired with machine-learning and pattern recognition algorithms, this can identify patterns of VOCs, and thus patterns of smell, that can be used to identify disease states. In this review, we provide a clinical background of VOC identification, electronic nose development, and review gastroenterology applications toward diagnosing disease by the volatile headspace analysis of stool.

Diagnosing gastrointestinal illnesses using fecal headspace volatile organic compounds - PMC

Volatile organic compounds (VOCs) emitted from stool are the components of the smell of stool representing the end products of microbial activity and metabolism that can be used to diagnose disease. Despite the abundance of hydrogen, carbon dioxide, ...

www.ncbi.nlm.nih.gov

And,, with UV Lights installed , you get a Tan down under as well !!

 

[MadMayHam | 合氣道]

Looks like the only use of NASA is their dataset for failed bearings to see if the AI model can predict the failures.

Thank you that’s great. Can you complete the sentence regarding NASA just so I/we know what this part relates too.

Regards
FF

AKIDA BALLISTA

6. Evaluation on time series data​

In the previous sections, we focused on image data. Visual anomaly detection is an important task with many real-world applications in quality control and product safety. In addition, the visual nature of the data made it easier to intuitively understand the behavior of our model. There are however also many use cases where the input data is in the form of a time series. In this section, we show how our method can also be applied to this domain.

We experimented with the NASA Intelligent Maintenance Systems (IMS) bearing dataset [93]. It was collected on a test rig which consists of an electric motor that drives a shaft mounted with 4 double row Rexnord ZA-2115 bearings. For each bearing, accelerometer data was collected at a rate of 20 kHz. The entire lifetime of the bearing is recorded and the task is to detect when the bearing starts to degrade and ultimately fails. The dataset consists of three experiments with four failed bearings in total.

We applied our model in an online fashion to this dataset. Every sample is passed through the model which returns an anomaly score and updates its internal memory. We first transformed the data from the time domain to the frequency domain by applying a Fast Fourier Transform (FFT) to every window of 1 s (20.000 samples). The resulting 10.000 components are then passed through our random sparse anomaly detection model with two layers of 1000 neurons each and a sparsity rate of 90%. The model uses a learning rate of 0.05. To decide when to flag an input window as anomalous, we need to compare the returned anomaly score with a predefined threshold. For this, we calculated the mean (μ) and standard deviation (σ) of the predicted anomaly score of the first 10% of the dataset and set the threshold to (μ+5σ). The intuition behind this is that in the first part of the experiment, the bearing is still in a healthy state. We discard the first 100 predictions in this threshold calculation since the model returns a very high anomaly score for these samples as it is still learning to model the data. Table 8 shows the earliest detected anomalous samples for the four failed bearings, for different state-of-the-art methods. Our approach with sparse random neural networks performs similar or even sometimes slightly better than the existing approaches. A benefit of our approach however is that it does not require a separate training stage as it is trained in an online fashion.

 

[Shezza]

sGidday Fox,
Just between SpaceX and Tesla, Using Akida could be one thing that saves him.

LOL

Come on ladies/ gents,

First non subscriber to answer correctly wins a month subscription.
What company lent a camera to brainchip to use for gesture recognition and what year was it?

 

[Foxdog]

yeah... nah.... not really going to work. You can do weird things with infinity and zero, but not everything.
TGIF

Was worth having a crack at it tho

 

[Steve10]

Finally... proof that ChatGPT has surpassed human (well, at least some) intelligence.

View attachment 26344

It can write code for trading bots in pine script which can be loaded into tradingview. I had it write code for a BRN trading bot . 75% profitable & 3.632 profit factor for BRN.

 

[Fact Finder]

Looks like the only use of NASA is their dataset for failed bearings to see if the AI model can predict the failures.

6. Evaluation on time series data​

In the previous sections, we focused on image data. Visual anomaly detection is an important task with many real-world applications in quality control and product safety. In addition, the visual nature of the data made it easier to intuitively understand the behavior of our model. There are however also many use cases where the input data is in the form of a time series. In this section, we show how our method can also be applied to this domain.

We experimented with the NASA Intelligent Maintenance Systems (IMS) bearing dataset [93]. It was collected on a test rig which consists of an electric motor that drives a shaft mounted with 4 double row Rexnord ZA-2115 bearings. For each bearing, accelerometer data was collected at a rate of 20 kHz. The entire lifetime of the bearing is recorded and the task is to detect when the bearing starts to degrade and ultimately fails. The dataset consists of three experiments with four failed bearings in total.

We applied our model in an online fashion to this dataset. Every sample is passed through the model which returns an anomaly score and updates its internal memory. We first transformed the data from the time domain to the frequency domain by applying a Fast Fourier Transform (FFT) to every window of 1 s (20.000 samples). The resulting 10.000 components are then passed through our random sparse anomaly detection model with two layers of 1000 neurons each and a sparsity rate of 90%. The model uses a learning rate of 0.05. To decide when to flag an input window as anomalous, we need to compare the returned anomaly score with a predefined threshold. For this, we calculated the mean (μ) and standard deviation (σ) of the predicted anomaly score of the first 10% of the dataset and set the threshold to (μ+5σ). The intuition behind this is that in the first part of the experiment, the bearing is still in a healthy state. We discard the first 100 predictions in this threshold calculation since the model returns a very high anomaly score for these samples as it is still learning to model the data. Table 8 shows the earliest detected anomalous samples for the four failed bearings, for different state-of-the-art methods. Our approach with sparse random neural networks performs similar or even sometimes slightly better than the existing approaches. A benefit of our approach however is that it does not require a separate training stage as it is trained in an online fashion.

Just brilliant. Many thanks.

Brainchip-Peter van der Made published performance figures for AKIDA processing a bearing fault data set. Pretty sure it ended up on website under automotive vibration monitoring and my vague memory was a state of the art 99% but will need to check.

My opinion only DYOR
FF

AKIDA BALLISTA