BRN Discussion Ongoing

Neuromorphia

fact collector
Context..Video about Renesas, Brainchips Customer.
 
Last edited:
  • Like
  • Fire
Reactions: 14 users
Following on from this, is it safe to assume that Apple are using us in their macs and perhaps tablets?

An article I was reading makes me believe the possibility.


The last paragraph from that article:
View attachment 23089


From the article that @Pmel just published:



View attachment 23090

Is my dot joining close to the money? Please let it be so.
Funny bring up Apple.

Found this yesterday but hadn't posted yet.






Erik Engheim

Erik Engheim
Follow
Jan 15, 2021
·
9 min read



The Secret Apple M1 Coprocessor​

Developer Dougall Johnson has through reverse engineering, uncovered a secret powerful coprocessor dubbed AMX: Apple Matrix coprocessor inside the M1 chip.​

1*VlTqk3Wqux8dx7MOdxhiCQ.png

Stories about the Apple Matrix coprocessor (AMX) are already out there. But not exactly discussed in a beginner friendly manner. And that is what I try to do here. Bring you the story buried under thick layers of technical jargon without treating you like an idiot.
To tell this story we need to clarify the basics such as what is a coprocessor? What is a matrix? And why should you even care about any of this?
More importantly why do none of the Apple slides talk about this coprocessor? Why is it seemingly a secret? If you have read about the Neural Engine inside the M1 System-on-a-Chip (SoC) you may be confused about what makes Apple’s Matrix coprocessor (AMX) is different.
Before we get to the big question, let me start with the basic concepts such as what a matrix and a coprocessor is.

What is a Matrix Anyway?​

A matrix is basically just a table of numbers. If you have worked with spreadsheets such as Microsoft Excel, you have basically worked with something very similar to matricies. The key difference is that in math such tables of numbers have a laundry list of operations they support and specific behavior. A matrix can come in different flavors as you see here. A matrix with such a row, is usually called a row vector. If one a column, we call it a column vector.
1*R6rU7A6pdLw9vrLrmx_bQA.png

We can add, subtract, scale and multiple matrices. Addition is pretty easy. You just add every element separately. Multiplication is a bit more involved. I am just showing the simple case here.
1*ma5F_5R-6Nnyx_QJUvnFxw.png

More in depth: Why Does Matrix Multiplication Work the Way it Does?
Using matrices to rotate and scale: Explaining Affine Rotation (this is pretty math geeky).

Why Do We Care About Matrices?​

The reason matrices are important is because they are heavily used in:
  • Image processing
  • Machine learning
  • Speech and handwriting recognition
  • Face recognition
  • Compression
  • Multimedia: audio and video
In particular machine learning which has been hot these last years. Just adding more cores to the CPU will not make this run fast enough as it is really demanding. You really need specialized hardware. Regular tasks such browsing the internet, writing email, word processing and spreadsheets has been running fast enough for years. It is for specialized tasks which we really need to boost the processing power.
1*V_JqQwU8HGIWjs6xP2IDxg.png

You could spend your silicon real-estate (transistors) on more CPU cores or by adding specialized hardware.
On any given chip, Apple has a max number of transistors to spend building different kinds of hardware. They could add more CPU cores but that really just speeds up regular tasks, which already run fast enough. Thus they have chosen to spend transistors to make specialized hardware to tackle image processing, video decoding and machine learning. This specialized hardware is the coprocessor and accelerators.
More talk about coprocessors and accelerators: Apple M1 foreshadows Rise of RISC-V.

How is Apple’s Matrix Coprocessor Different From the Neural Engine?​

If you have read about the Neural Engine, you will know that it also does matrix operations to help with machine learning tasks. So why do we need the Matrix coprocessor? Or are they actually just the same thing? Am I just confused? No, let me clarify how Apple’s Matrix Coprocessor differ from the Neural Engine and why we need both.
1*TV2WvfubVuR8I3gq3302EQ.png

The main processor (CPU), coprocessors and accelerators can usually exchange data over a shared data bus. The CPU usually controls memory access while an Accelerator such as a GPU often has its own dedicated memory.
I admit that in past stories I often use the term coprocessor and accelerator interchangeably but they are not the same. A GPU as found in your Nvidia graphics card and the Neural Engine are both a type of accelerator.
In both cases you have special areas of memory which the CPU has to fill up with data it wants processed as well as another part of memory which it fills up with a list of instructions that accelerator should perform. It is time consuming for a CPU to setup this kind of processing. There is a lot of coordination, filling in data, and then waiting to get results back.
Thus this only pays off for larger tasks. For smaller tasks the overhead will be too high.
1*S_E9rKaqa1Vp1rCJ5lpwHg.png

Coprocessors unlike accelerator spy on the stream of instructions read from memory into the main processor. Accelerators in contrast don’t observe the instructions the CPU is pulling from memory.
This is where coprocessors are a benefit over accelerators.
Edit November 2nd 2021: More recent info on the AMX suggest the description I am giving below is not correct. There is no AMX per core, so it cannot spy on the instruction stream to a core. However the advantage relative to the Neural Engine is similar to what I describe. AMX access memory more like a CPU than a GPU or Neural Engine which are optimized for processing large but slow batches of data.
Coprocessors sit and spy on the stream of machine code instructions being fed from memory (or cache more specifically) into the CPU. Coprocessor are made to react to particular instructions they were made to process. The CPU meanwhile has been made to mostly ignore these instructions or help facilitate the handling of them by a coprocessor.
What we gain from this is that instructions carried out by the coprocessor can be placed inside your regular code. This is different from say a GPU. If you have done GPU programming you know that shader programs are placed into separate buffers of memory, and you have to explicitly transport these shader programs to the GPU. You cannot place GPU specific instruction inside your regular code. Thus for smaller workloads involving matrix processing AMX will be better than the Neural Engine.
What is the catch? You need to actually define the instructions in the instruction-set architecture (ISA) of your microprocessor. Thus you need much tighter integration with the CPU when using a coprocessor than when using an accelerator.
ARM Ltd. creators of the ARM instruction-set architecture (ISA) has long resisted adding custom instructions to their ISA. This is one of the advantages of RISC-V: What Is Innovative About RISC-V?
However due to pressure from customers ARM relented and announced in 2019 that they would allow extensions. EE Times reports:
The new instructions are interleaved with standard Arm instructions. To avoid software fragmentation and maintain a coherent software development environment, Arm expects customers to use the custom instructions mostly in called library functions.
This may help explain why AMX instructions are not described in official documentation. ARM Ltd. expects Apple to keep these kinds of instructions inside libraries provided by the customer (Apple in this case).

How is a Matrix Coprocessor Different From a SIMD Vector Engine?​

It is easy to confuse something like a matrix coprocessor with a SIMD vector engine, which you find inside most modern processors today including ARM processors. SIMD stands for Single Instruction Multiple Data.
1*zL0kk_7Vx5EeisutGLBf2g.png

Single Instruction Single Data (SISD) vs Single Instruction Multiple Data (SIMD)
SIMD is a way of getting higher performance when you need to perform the same operation on multiple elements. This is closely related to matrix operations. In fact SIMD instructions such as ARM’s Neon instructions or Intel x86 SSE or AVX are often used to speed up matrix multiplications.
Read more: RISC-V Vector Instructions vs ARM and x86 SIMD.
However a SIMD vector engine is part of a microprocessor core. Just like the ALU (Arithmetic Logic Unit) and FPU (Floating Point Unit) is part of the CPU. Inside the microprocessor there is an instruction decoder which will pick apart an instruction and decide what functional unit to activate (gray boxes).
1*Q_BtnKz3iYzYnuGQgSOo5A.png

Inside a CPU you got the ALU, FPU as well as SIMD vector engines (not shown) as separate parts activated by the instruction decoder. A coprocessor is external.
A coprocessor in contrast is external to a microprocessor core. In fact one of the early ones, Intel’s 8087 was a physically separate chip designed to speed up floating point calculations.
1*SEEZxqITUEJQJgSMxYhMbA.jpeg

Intel 8087. One of the early coprocessors used for performing floating point calculations.
Now you may wonder why anyone would want to complicate CPU design by having a separate chip like this which has to sniff on the data flowing from memory to the CPU, to see if anything is a floating point instruction.
The reason was simple, the original 8086 CPU in the first PCs contained 29,000 transistors. The 8087 in contrast was far more complex at 45,000 transistors. It was really hard to make anything with that many transistors. Combining these two chips into one would have been really hard and expensive.
But as manufacturing technology improved, it was not a problem to put floating point units (FPUs) inside the CPU. Thus FPUs replaced the floating point coprocessors.
Why the AMX is not simply a part of the Firestorm cores on the M1 is not clear to me. They are all on the same silicon die anyway. I can only offer some speculations. By being a coprocessor, it may be easier for the CPU to continue running in parallel. Apple may also have liked to keep non-standard ARM stuff outside of their ARM CPU cores.

Why Is the AMX a Secret?​

If AMX is not described in official documentation, how do we even know about it? Thanks to developer Dougall Johnson, who has done an amazing job reverse engineering the M1 to discover this coprocessor. His efforts are described here. For matrix related math operations Apple has special libraries or frameworks such as Accelerate, which is made up of:
  • vImage — higher level image processing, such as converting between formats, image manipulation.
  • BLAS — a sort of industry standard for linear algebra (what we call the math dealing with matricies and vectors).
  • BNNS — is used for running neural networks and training.
  • vDSP — digital signal processing. Fourier transformations, convolution. These are mathematical operations important in image processing or any signal really including audio.
  • LAPACK higher level linear algebra functions, e.g. for solving linear equations.
Dougall Johnson knew these libraries would use the AMX coprocessor to speed up their calculations. Thus he wrote special programs to analyze and observe what these programs did to discover the special undocumented AMX machine code instructions.
But why doesn’t Apple document this and let us use these instructions directly? As mentioned earlier, this is something ARM Ltd. would like to avoid. If custom instructions are widely used it could fragment the ARM ecosystem.
However more importantly, this is an advantage to Apple. By only letting their libraries use these special instructions Apple retains the freedom to radically change how this hardware works later. They could remove or add AMX instructions. Or they could let the Neural Engine do the job. Either way they make the job easier for developers. Developers only need to use the Accelerate framework and can ignore how Apple specifically speeds up matrix calculations.
This is one of the big advantages Apple has by being vertically integrated. By controlling both the hardware and the software, they can pull these kinds of tricks. So the next question is how big a deal is this? What does this buy Apple in terms of performance and capabilities?

What Are the Advantages of Apple’s Matrix Coprocessor?​

Nod Labs is a company that does machine interaction, intelligence and perception. Fast matrix operations are naturally in their interest. They have written a highly technical blog post of doing performance tests of AMX: Comparing Apple’s M1 matmul performance — AMX2 vs NEON.
What they are doing is comparing performance of doing similar code using AMX with doing it using the Neon instructions, which are officially supported by ARM. Neon is a type of SIMD instructions.
What Nod Labs found was that by using AMX they were able to get twice the performance of Neon instructions for matrix operations. It doesn’t mean AMX is better for everything, but at least for machine learning and high performance computing (HPC) type of work, we can expect that AMX gives an edge over the competition.

Summary​

The Apple Matrix Coprocessor looks like some rather impressive piece of hardware giving Apple’s ARM processor an edge in machine learning and HPC related tasks. Further investigation will give us a more complete picture and I can update this story with more details.
 
  • Like
  • Fire
  • Wow
Reactions: 25 users
  • Like
Reactions: 8 users

Interesting as SynQuacer is the system that Brainchip and Socionext announced as being available with AKIDA.

Just dots but very large overlapping dots much like the sheets of white paper that mean so much in the Chinese protest against the CCP.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Fire
  • Love
Reactions: 15 users
I have posted the video for this RZ/V2 previous page. Renesas has uploaded to YouTube
I'm not 100 percent convinced on this. Love to be wrong though.
 
  • Like
  • Love
Reactions: 4 users
Funny bring up Apple.

Found this yesterday but hadn't posted yet.






Erik Engheim
Erik Engheim
Follow
Jan 15, 2021
·
9 min read



The Secret Apple M1 Coprocessor​

Developer Dougall Johnson has through reverse engineering, uncovered a secret powerful coprocessor dubbed AMX: Apple Matrix coprocessor inside the M1 chip.​

1*VlTqk3Wqux8dx7MOdxhiCQ.png

Stories about the Apple Matrix coprocessor (AMX) are already out there. But not exactly discussed in a beginner friendly manner. And that is what I try to do here. Bring you the story buried under thick layers of technical jargon without treating you like an idiot.
To tell this story we need to clarify the basics such as what is a coprocessor? What is a matrix? And why should you even care about any of this?
More importantly why do none of the Apple slides talk about this coprocessor? Why is it seemingly a secret? If you have read about the Neural Engine inside the M1 System-on-a-Chip (SoC) you may be confused about what makes Apple’s Matrix coprocessor (AMX) is different.
Before we get to the big question, let me start with the basic concepts such as what a matrix and a coprocessor is.

What is a Matrix Anyway?​

A matrix is basically just a table of numbers. If you have worked with spreadsheets such as Microsoft Excel, you have basically worked with something very similar to matricies. The key difference is that in math such tables of numbers have a laundry list of operations they support and specific behavior. A matrix can come in different flavors as you see here. A matrix with such a row, is usually called a row vector. If one a column, we call it a column vector.
1*R6rU7A6pdLw9vrLrmx_bQA.png

We can add, subtract, scale and multiple matrices. Addition is pretty easy. You just add every element separately. Multiplication is a bit more involved. I am just showing the simple case here.
1*ma5F_5R-6Nnyx_QJUvnFxw.png

More in depth: Why Does Matrix Multiplication Work the Way it Does?
Using matrices to rotate and scale: Explaining Affine Rotation (this is pretty math geeky).

Why Do We Care About Matrices?​

The reason matrices are important is because they are heavily used in:
  • Image processing
  • Machine learning
  • Speech and handwriting recognition
  • Face recognition
  • Compression
  • Multimedia: audio and video
In particular machine learning which has been hot these last years. Just adding more cores to the CPU will not make this run fast enough as it is really demanding. You really need specialized hardware. Regular tasks such browsing the internet, writing email, word processing and spreadsheets has been running fast enough for years. It is for specialized tasks which we really need to boost the processing power.
1*V_JqQwU8HGIWjs6xP2IDxg.png

You could spend your silicon real-estate (transistors) on more CPU cores or by adding specialized hardware.
On any given chip, Apple has a max number of transistors to spend building different kinds of hardware. They could add more CPU cores but that really just speeds up regular tasks, which already run fast enough. Thus they have chosen to spend transistors to make specialized hardware to tackle image processing, video decoding and machine learning. This specialized hardware is the coprocessor and accelerators.
More talk about coprocessors and accelerators: Apple M1 foreshadows Rise of RISC-V.

How is Apple’s Matrix Coprocessor Different From the Neural Engine?​

If you have read about the Neural Engine, you will know that it also does matrix operations to help with machine learning tasks. So why do we need the Matrix coprocessor? Or are they actually just the same thing? Am I just confused? No, let me clarify how Apple’s Matrix Coprocessor differ from the Neural Engine and why we need both.
1*TV2WvfubVuR8I3gq3302EQ.png

The main processor (CPU), coprocessors and accelerators can usually exchange data over a shared data bus. The CPU usually controls memory access while an Accelerator such as a GPU often has its own dedicated memory.
I admit that in past stories I often use the term coprocessor and accelerator interchangeably but they are not the same. A GPU as found in your Nvidia graphics card and the Neural Engine are both a type of accelerator.
In both cases you have special areas of memory which the CPU has to fill up with data it wants processed as well as another part of memory which it fills up with a list of instructions that accelerator should perform. It is time consuming for a CPU to setup this kind of processing. There is a lot of coordination, filling in data, and then waiting to get results back.
Thus this only pays off for larger tasks. For smaller tasks the overhead will be too high.
1*S_E9rKaqa1Vp1rCJ5lpwHg.png

Coprocessors unlike accelerator spy on the stream of instructions read from memory into the main processor. Accelerators in contrast don’t observe the instructions the CPU is pulling from memory.
This is where coprocessors are a benefit over accelerators.
Edit November 2nd 2021: More recent info on the AMX suggest the description I am giving below is not correct. There is no AMX per core, so it cannot spy on the instruction stream to a core. However the advantage relative to the Neural Engine is similar to what I describe. AMX access memory more like a CPU than a GPU or Neural Engine which are optimized for processing large but slow batches of data.
Coprocessors sit and spy on the stream of machine code instructions being fed from memory (or cache more specifically) into the CPU. Coprocessor are made to react to particular instructions they were made to process. The CPU meanwhile has been made to mostly ignore these instructions or help facilitate the handling of them by a coprocessor.
What we gain from this is that instructions carried out by the coprocessor can be placed inside your regular code. This is different from say a GPU. If you have done GPU programming you know that shader programs are placed into separate buffers of memory, and you have to explicitly transport these shader programs to the GPU. You cannot place GPU specific instruction inside your regular code. Thus for smaller workloads involving matrix processing AMX will be better than the Neural Engine.
What is the catch? You need to actually define the instructions in the instruction-set architecture (ISA) of your microprocessor. Thus you need much tighter integration with the CPU when using a coprocessor than when using an accelerator.
ARM Ltd. creators of the ARM instruction-set architecture (ISA) has long resisted adding custom instructions to their ISA. This is one of the advantages of RISC-V: What Is Innovative About RISC-V?
However due to pressure from customers ARM relented and announced in 2019 that they would allow extensions. EE Times reports:

This may help explain why AMX instructions are not described in official documentation. ARM Ltd. expects Apple to keep these kinds of instructions inside libraries provided by the customer (Apple in this case).

How is a Matrix Coprocessor Different From a SIMD Vector Engine?​

It is easy to confuse something like a matrix coprocessor with a SIMD vector engine, which you find inside most modern processors today including ARM processors. SIMD stands for Single Instruction Multiple Data.
1*zL0kk_7Vx5EeisutGLBf2g.png

Single Instruction Single Data (SISD) vs Single Instruction Multiple Data (SIMD)
SIMD is a way of getting higher performance when you need to perform the same operation on multiple elements. This is closely related to matrix operations. In fact SIMD instructions such as ARM’s Neon instructions or Intel x86 SSE or AVX are often used to speed up matrix multiplications.
Read more: RISC-V Vector Instructions vs ARM and x86 SIMD.
However a SIMD vector engine is part of a microprocessor core. Just like the ALU (Arithmetic Logic Unit) and FPU (Floating Point Unit) is part of the CPU. Inside the microprocessor there is an instruction decoder which will pick apart an instruction and decide what functional unit to activate (gray boxes).
1*Q_BtnKz3iYzYnuGQgSOo5A.png

Inside a CPU you got the ALU, FPU as well as SIMD vector engines (not shown) as separate parts activated by the instruction decoder. A coprocessor is external.
A coprocessor in contrast is external to a microprocessor core. In fact one of the early ones, Intel’s 8087 was a physically separate chip designed to speed up floating point calculations.
1*SEEZxqITUEJQJgSMxYhMbA.jpeg

Intel 8087. One of the early coprocessors used for performing floating point calculations.
Now you may wonder why anyone would want to complicate CPU design by having a separate chip like this which has to sniff on the data flowing from memory to the CPU, to see if anything is a floating point instruction.
The reason was simple, the original 8086 CPU in the first PCs contained 29,000 transistors. The 8087 in contrast was far more complex at 45,000 transistors. It was really hard to make anything with that many transistors. Combining these two chips into one would have been really hard and expensive.
But as manufacturing technology improved, it was not a problem to put floating point units (FPUs) inside the CPU. Thus FPUs replaced the floating point coprocessors.
Why the AMX is not simply a part of the Firestorm cores on the M1 is not clear to me. They are all on the same silicon die anyway. I can only offer some speculations. By being a coprocessor, it may be easier for the CPU to continue running in parallel. Apple may also have liked to keep non-standard ARM stuff outside of their ARM CPU cores.

Why Is the AMX a Secret?​

If AMX is not described in official documentation, how do we even know about it? Thanks to developer Dougall Johnson, who has done an amazing job reverse engineering the M1 to discover this coprocessor. His efforts are described here. For matrix related math operations Apple has special libraries or frameworks such as Accelerate, which is made up of:
  • vImage — higher level image processing, such as converting between formats, image manipulation.
  • BLAS — a sort of industry standard for linear algebra (what we call the math dealing with matricies and vectors).
  • BNNS — is used for running neural networks and training.
  • vDSP — digital signal processing. Fourier transformations, convolution. These are mathematical operations important in image processing or any signal really including audio.
  • LAPACK higher level linear algebra functions, e.g. for solving linear equations.
Dougall Johnson knew these libraries would use the AMX coprocessor to speed up their calculations. Thus he wrote special programs to analyze and observe what these programs did to discover the special undocumented AMX machine code instructions.
But why doesn’t Apple document this and let us use these instructions directly? As mentioned earlier, this is something ARM Ltd. would like to avoid. If custom instructions are widely used it could fragment the ARM ecosystem.
However more importantly, this is an advantage to Apple. By only letting their libraries use these special instructions Apple retains the freedom to radically change how this hardware works later. They could remove or add AMX instructions. Or they could let the Neural Engine do the job. Either way they make the job easier for developers. Developers only need to use the Accelerate framework and can ignore how Apple specifically speeds up matrix calculations.
This is one of the big advantages Apple has by being vertically integrated. By controlling both the hardware and the software, they can pull these kinds of tricks. So the next question is how big a deal is this? What does this buy Apple in terms of performance and capabilities?

What Are the Advantages of Apple’s Matrix Coprocessor?​

Nod Labs is a company that does machine interaction, intelligence and perception. Fast matrix operations are naturally in their interest. They have written a highly technical blog post of doing performance tests of AMX: Comparing Apple’s M1 matmul performance — AMX2 vs NEON.
What they are doing is comparing performance of doing similar code using AMX with doing it using the Neon instructions, which are officially supported by ARM. Neon is a type of SIMD instructions.
What Nod Labs found was that by using AMX they were able to get twice the performance of Neon instructions for matrix operations. It doesn’t mean AMX is better for everything, but at least for machine learning and high performance computing (HPC) type of work, we can expect that AMX gives an edge over the competition.

Summary​

The Apple Matrix Coprocessor looks like some rather impressive piece of hardware giving Apple’s ARM processor an edge in machine learning and HPC related tasks. Further investigation will give us a more complete picture and I can update this story with more details.

Not related but I think apple spat the dummy.
They have deleted all their tweets?
 
  • Wow
  • Sad
  • Like
Reactions: 5 users
Was having a check in on MB jobs.

This one advertised beginning Nov for India MB.

Doesn't mention Akida but that would probs be difficult to request that experience as I don't expect as yet, that many out there with full hands on experience.

Is based around NVIDIA from what I read.

But what caught my eye was the job title.

Now, I'm wondering if that is just an internal naming convention or if relates back to Momenta AI out of China.

Musing on overlaps, diff scenarios of where we fit still without any formal updates and US / China tech issues.

Article below on some strategic investors in Momenta inc MB and Bosch.



Vendor/Privacy Policy
Career Job Detail

Mercedes Benz Logo

November 2022
Mercedes-Benz Research and Development India Private Limited

MBOS - ADAS – Software Development​

tasks

Our team:
We build best in class software for Autonomous driving functions for Mercedes Benz passenger cars. We are a team of talented engineers working in a global network to develop the next generation of Driver Assistance and Automated Driving features used in all Mercedes Benz cars sold globally.
About the role:
Be part of an agile software development team developing AI & ML based software for ADAS and Automated Driving features
Job Title:
MB - Momenta ADAS & Automated Driving – Software Development

Job Category:
Internal
Department/Group:
ICA
Job Code:

Years of Experience
1-4 Years(T9)
Opportunity No:

Responsibilities
- Contribute to the R&D of cutting-edge ADAS/AD functions
- AI and ML based function development of ADAS/AD features (e.g. highway pilot) with data driven and rule based approaches.
- Implement ADAS/AD functions from the system/software specification
- Practical usage of mathematical, physical and logical knowledge
- Develop functions from concept until production ready maturity level
- Implement algorithms on embedded hardware
- Test and validate algorithms in simulation and real-car driving
- Analyze and solve problem reports coming from field tests
- Perform effective root cause analysis for problems reported during vehicle testing
- Create tests for the feature on various levels e.g. Unit Tests, SIL, HIL
- Ensure release readiness by creating the required documentation for the customer
- Provide fast and effective software implementation for proof of concepts, when required


Required Skills: Technical
- Over 1-4 years of experience in software development or equivalent in the automotive industry, preferably in the AD/ADAS domain
- Strong C/C++ skills to design, build, maintain efficient and reliable code
- Strong with data driven algorithm development using Machine learning, Deep learning and AI
- Familiarity with hardware acceleration platforms (Nvidia TensorRT, Intel OpenVino)
- Familiarity with some common ML frameworks (TensorFlow, Onxx, Pytorch, etc)
- Understanding of ML workflow: preparing the data, implementing and training ML models, evaluating results, deploying inference on different platforms
- Experience making performant ML pipelines/ inference servers
- Knowledge of different ML models and how to train/benchmark their performance
- Familiar with Python programming
- Familiar with Linux environment
- Professional experience with classic and/or adaptive AUTOSAR is a strong plus
- Strong skills in Automates/State-Machines
- Knowledge in the ADAS field of different SAE Automated Driving levels and in ISO 26262 will be an advantage
- Exposure to configuration management tools like GIT and requirement management tools like DOORS, design tools like Rhapsody, Enterprise Architect
- Good understanding of the SW build environment and build process – compilation, linking, preprocessing etc.
- Hands-on experience in working with Static Quality check tools PRQA, ASTREE, Coverity
- Capability of testing on target (both, HiL and test vehicles)

Required Skills: Non-Technical
- Ability to work in a team environment with open mindedness.
- Curious, team-work, self-motivated
- Excellent oral and written communication skills
- Strong Problem solving, logical thinking and analytical skills.
- Ability to communicate and discuss ideas effectively - verbally and through presentations
- Deep interest in technology
- Excellent organizational, time management, prioritization and multi-tasking skills




Autonomous Driving Startup Momenta Raises Another $500M

In the wake of a $300 million investment from General Motors in September, Momenta, an autonomous driving solution provider from China, announced today an additional $500 million added to its Series C round.

Source: TechCrunch | Published on November 5, 2021


The new injection brings the total of the startup’s Series C to over $1 billion. Momenta adapts what it calls a two-legged strategy of supplying advanced driver assistance systems (ADAS) to auto OEMs like GM and Tier 1 suppliers like Bosch, while conducting R&D on truly unmanned vehicles, that is, Level 4 driving.

The startup has assembled a list of heavyweight strategic investors, including China’s state-owned SAIC Motor, GM, Toyota, Mercedes Benz, and Bosch. Singapore’s sovereign fund Temasek and Jack Ma’s Yunfeng Capital are among its institutional investors.

Momenta often speaks of how its alliance with car manufacturers differentiates itself from its peers, which have chosen a more cash-intensive route of developing in-house robotaxi fleets. Instead, it counts on gleaning data from a network of mass-produced vehicles powered by its solutions. Pony.ai and WeRide are among its closest rivals and have also raised a significant amount of capital.

In the case of the GM deal, Momenta’s solution, which uses a mix of consumer-grade millimeter-wave radars and high-definition cameras, will be used in GM’s vehicles sold in China rather than the United States. The startup recently opened its first overseas office in Stuttgart to be closer to its German partners, which may imply the footprint of its technology could extend beyond its home market.
 
  • Like
  • Fire
  • Thinking
Reactions: 11 users
Anybody looked into this yet?

 
  • Like
  • Love
Reactions: 6 users

Neuromorphia

fact collector
Last edited:
  • Like
  • Fire
Reactions: 13 users
Anybody looked into this yet?

Yes a few weeks ago. It will accelerate their ability to undertake research but does not have any implications for AKIDA/Brainchip apart from this.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Love
Reactions: 7 users
Yes a few weeks ago. It will accelerate their ability to undertake research but does not have any implications for AKIDA/Brainchip apart from this.

My opinion only DYOR
FF

AKIDA BALLISTA
Cheers for that. I'll still have a further dig anyway just in case 🤓
 
  • Like
Reactions: 2 users

TheFunkMachine

seeds have the potential to become trees.
Don't know if this was posted previously when Anils one was?

Is from the recent TinyML forum end of Sept.

Watch from around the 9.15 min mark when at the end of his piece the moderator asks Christoph about the Neuromorphic Hardware / Chips etc roles could play with event-based vision.

Starts answer / comment with.....but then gets a bit cagey on architecture...wonder if just on their stuff or us as well :unsure:

View attachment 23060

This is the vid presso:

Neuromorphic Event-based Vision
Christoph POSCH, CTO, PROPHESEE

Abstract (English)
Neuromorphic Event-based (EB) vision is an emerging paradigm of acquisition and processing of visual information that takes inspiration from the functioning of the human vision system, trying to recreate its visual information acquisition and processing operations on VLSI silicon chips. In contrast to conventional image sensors, EB sensors do not use one common sampling rate (=frame rate) for all pixels, but each pixel defines the timing of its own sampling points in response to its visual input by reacting to changes of the amount of incident light. The highly efficient way of acquiring sparse data, the high temporal resolution and the robustness to uncontrolled lighting conditions are characteristics of the event sensing process that make EB vision attractive for numerous applications in industrial, surveillance, IoT, AR/VR, automotive. This short presentation will give an introduction to EB sensing technology and highlight a few exemplary use cases.


Don't know if this was posted previously when Anils one was?

Is from the recent TinyML forum end of Sept.

Watch from around the 9.15 min mark when at the end of his piece the moderator asks Christoph about the Neuromorphic Hardware / Chips etc roles could play with event-based vision.

Starts answer / comment with.....but then gets a bit cagey on architecture...wonder if just on their stuff or us as well :unsure:

View attachment 23060

This is the vid presso:

Neuromorphic Event-based Vision
Christoph POSCH, CTO, PROPHESEE

Abstract (English)
Neuromorphic Event-based (EB) vision is an emerging paradigm of acquisition and processing of visual information that takes inspiration from the functioning of the human vision system, trying to recreate its visual information acquisition and processing operations on VLSI silicon chips. In contrast to conventional image sensors, EB sensors do not use one common sampling rate (=frame rate) for all pixels, but each pixel defines the timing of its own sampling points in response to its visual input by reacting to changes of the amount of incident light. The highly efficient way of acquiring sparse data, the high temporal resolution and the robustness to uncontrolled lighting conditions are characteristics of the event sensing process that make EB vision attractive for numerous applications in industrial, surveillance, IoT, AR/VR, automotive. This short presentation will give an introduction to EB sensing technology and highlight a few exemplary use cases.



@ 9:15 she asks him about the use of neuromorphic chips in regards to event based vision systems. He states that he think SNN are a good fit, but then after that he said something strange considering our previous conversation with Prophesee on our podcast where they basically said that Brainchip is the key to their success (paraphrased)

He said: “but I’m not sure if the optimal processing architecture has been identified yet” Very strange thing to say when Luca Verre spoke so highly of Brainchip and Akida and the recent partnership etc. am I reading too much into this?
 
  • Like
  • Love
  • Wow
Reactions: 9 users
@ 9:15 she asks him about the use of neuromorphic chips in regards to event based vision systems. He states that he think SNN are a good fit, but then after that he said something strange considering our previous conversation with Prophesee on our podcast where they basically said that Brainchip is the key to their success (paraphrased)

He said: “but I’m not sure if the optimal processing architecture has been identified yet” Very strange thing to say when Luca Verre spoke so highly of Brainchip and Akida and the recent partnership etc. am I reading too much into this?
Agreed.

I didn't want to bias the post preferring to let others judge themselves.

I worked on the premise that maybe saying that to not give too much away or provide a form of confirmation of where they're at with Akida even though Luca spoke quite glowingly.

That and they probs still running tests against the options they have and don't have complete results yet.
 
  • Like
Reactions: 8 users

Slade

Top 20
Stay patient Chippers. 2023 is going to be a fun year.
 
  • Like
  • Love
  • Fire
Reactions: 21 users

alwaysgreen

Top 20
This is an example of what the next generation of mass production vehicles will be coming out with.

Akida can do facial recognition, voice commands and gestures all at the same time using very low power.


Akida can do this easy.
I'm not trying to be a dick but if you post an article/video etc, it would be good to give it some context. So many posters do it and its annoying
 
  • Like
  • Love
  • Fire
Reactions: 12 users
I'm not trying to be a dick but if you post an article/video etc, it would be good to give it some context. So many posters do it and its annoying
I put my hand up to that. Will make a conscious effort from now on.
 
  • Like
  • Love
Reactions: 5 users

SERA2g

Founding Member
I'm not trying to be a dick but if you post an article/video etc, it would be good to give it some context. So many posters do it and its annoying
I agree with you on this one. It is annoying to see video's or articles posted by someone but the poster provides literally no context or opinion. I never read or watch the links when that's the case.

If it wasn't worth the posters time to to provide some information then I assume it's not worth my time to watch/read.

In this case though, @Getupthere posted the video then posted a separate comment providing the context, so go easy on the fella!
 
  • Like
  • Love
  • Fire
Reactions: 14 users

alwaysgreen

Top 20
I agree with you on this one. It is annoying to see video's or articles posted by someone but the poster provides literally no context or opinion. I never read or watch the links when that's the case.

If it wasn't worth the posters time to to provide some information then I assume it's not worth my time to watch/read.

In this case though, @Getupthere posted the video then posted a separate comment providing the context, so go easy on the fella!
Only because I asked if it was akida :) All good though, it's no big deal but it's just as you say, if they aren't going to read it before posting and add some context, it's likely a useless article that isn't worth reading.
 
  • Like
  • Love
  • Haha
Reactions: 5 users

Learning

Learning to the Top 🕵‍♂️

Akida in action with Renesas

Thanks for sharing Bennysmadness,

The video shows said Renesas is using their DPR-AI ACCELERATORS.
Screenshot_20221129_105308_Chrome.jpg


However, here is my non engineer background thinking out loud.
As we know from the Brainchip & Arm Podcast.
Arm is using Akida neuromorphic architecture for very specific task.
And we know Renesas only licensed two nodes of Akida.
So could it be, Renesas had integrated Akida into Renesas DPR-AI ACCELERATORS and as it not the full Akida architecture, there for, Renesas can advertise its as their in house AI accelerators.

Learning
PS, if I get 🤣 emoji from our tech members, then I know I am wrong in my thought. 😂😂😂
 
  • Like
  • Thinking
  • Fire
Reactions: 17 users
@ 9:15 she asks him about the use of neuromorphic chips in regards to event based vision systems. He states that he think SNN are a good fit, but then after that he said something strange considering our previous conversation with Prophesee on our podcast where they basically said that Brainchip is the key to their success (paraphrased)

He said: “but I’m not sure if the optimal processing architecture has been identified yet” Very strange thing to say when Luca Verre spoke so highly of Brainchip and Akida and the recent partnership etc. am I reading too much into this?
I would say yes to your question you are reading too much into this. Simple logic based on known facts.

AKIDA is the only commercial SNN available as IP. AKIDA IP is so advanced it is science fiction. So if for example AKIDA was producing 96% accuracy with Prophesee's event based vision sensor it is a long way ahead of anything out there and yet is 4% short of 100% accuracy. This missing 4% allows for a statement regarding whether they have found the optimal solution without telling an untruth.

The other thing however is that if AKIDA is producing 96% today just like it was first producing 94% with NaNose this level of accuracy will increase over time. In Nanose it increased to on various reports 96% to 98% accuracy which was over a year ago now so who knows what has been achieved since then. So with incremental learning and fine tuning 96% today could be 98% tomorrow or 100%. At 100% there would be no room to suggest it was not optimal.

On the speculative side of the equation do not forget that both Prophesee and AKIDA are working with NASA and DARPA and some pretty secret research is being undertaken across these two bodies with respect to hypersonic missile tracking. If this old technophobe can work out that event based vision and AKIDA provide the avenue for finding and tracking hypersonic missiles highly intelligent techies working for foreign governments would have worked it out and so there would be a lot of pressure on both companies to be circumspect about what they say at this stage.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Love
  • Fire
Reactions: 37 users
Top Bottom