BRN Discussion Ongoing

Esq.111

Fascinatingly Intuitive.
Hi Hop ,

Agree with what you have said.

Early voting can give the board a feel from the investors perspective , and hence withdraw or amend the points to be voted on before the AGM , if they so choose.

Additionally shareholders can amend their early vote , online , if something should happen prior to the AGM .

Everyone has different time horizons , and i feel for those that have been waiting , years , with baited breath .


Regards,
Esq.
 
  • Like
  • Love
Reactions: 22 users

HopalongPetrovski

I'm Spartacus!
Hi Hop ,

Agree with what you have said.

Early voting can give the board a feel from the investors perspective , and hence withdraw or amend the points to be voted on before the AGM , if they so choose.

Additionally shareholders can amend their early vote , online , if something should happen prior to the AGM .

Everyone has different time horizons , and i feel for those that have been waiting , years , with baited breath .


Regards,
Esq.
Wow, thanks Esq for that info.
Having only ever voted in person at the meetings I had no idea that shareholders could amend an early vote online.
Interesting........I wonder if our BOD avails themselves of this info?

As for time horizons, mine have definitely been uncomfortably stretched, but, with a bit more time in the market now and experiencing the rough and tumble of it all, would have to admit to them being naively short, initially.
I had the experience of early success which has perhaps coloured my expectational viewpoint unrealistically.
I have often had to learn "the hard way". 🤣
 
  • Like
  • Fire
  • Love
Reactions: 19 users

Sosimple

Regular
Hi SS,

Looks like Tianjic uses analog NN for audio. It also uses CNN (MAC-based) for video and something else for control.


A hybrid and scalable brain-inspired robotic platform | Scientific Reports (nature.com)


View attachment 61361

For each module, we opened independent data paths, which were distinct in information representation, frequency and throughput. In the visual module, each frame of video was resized to a 70 × 70 gray image and fed into a CNN as multi-bit values, enabling rich environmental spatial information to be maintained with limited computing resources (Fig. 3a). In contrast, the raw audio stream was transformed into binary spike trains in the auditory module. After end-point detection32,33, the key frequency features were obtained by taking the Mel Frequency Cepstral Coefficient (MFCC)34. A Gaussian population35 was used to encode each MFCC feature into spike trains as input to a three-layer fully-connected SNN (Fig. 3b). For motion control, sequential signals were first generated by functional cores to merge with the steering commands from other modules as the comprehensive target angle. Data from all sensors were then integrated via the MLP network for angle control (Fig. 3d). The designs and training of each network-based module are described in the Methods.

Tianjic uses different technologies to handle visual and audio. Akida can handle both.

They use 1-bit analog for audio which is very efficient but not so hot on accuracy. Akida can do 1-bit up to 8-bit inputs (256 values).

View attachment 61363

Tianjic needs a separate FPGA for pre-processing input signals. Akida has built-in preprocessing.

So, competition? - Yes.

Worried? - No,
Thanks Diogenes, your input is of great value to me and many others.
 
  • Like
  • Fire
  • Love
Reactions: 17 users

7für7

Top 20
It's all good friend.
I appreciate the cross pollination that happens here when different cultures, skill sets compulsions and languages mix and blend.
It's groovy. 🤣
I saw the movie several times but I guess I didn’t catch it Because it was synchronised in German language. 🙂↕️
 
  • Thinking
Reactions: 1 users

CHIPS

Regular
Good morning Dingo....

Sometimes it's hard to see past the forest because of all those trees !

Robot Ken is simply telling us that he's in France, most likely, Toulouse, where our brilliant software team has been working really
hard on his software makeup, as we all know the fastest process is the hardware phase, whereas the software is the much slower,
time consuming phase, the hint is, whom are we working in with in the robotics world.

But as usual, that's just my view...


Brain Mind GIF by University of California

Oh true, I forgot. Toulouse where Airbus is situated.
 
  • Like
  • Love
Reactions: 6 users

CHIPS

Regular
Whilst I’ve never been much of a fan of Ken the robot personally, there is a lot of hype around humanoid automatons in general.
They have a deep and rich place in our psyche and pander somewhat to our anthropocentric view of the world.
I hope that “if” this is a stratagem to introduce and popularise AKIDA 2 or even 3 that it's not too childlike and dumbed down.
In competition with the likes of Boston Dynamics that video doesn't cut the mustard in either messaging or production value in my opinion.

And just what is in the packaged item on the table here??
An edge box??
View attachment 61351

Looks like wet wipes to me.
I find this all very exciting. They let us guess what is happening here.
I think that something big is coming up.

Excited Arrested Development GIF



Excited Coffee GIF
 
  • Haha
Reactions: 5 users

HopalongPetrovski

I'm Spartacus!
Looks like wet wipes to me.
I find this all very exciting. They let us guess what is happening here.
I think that something big is coming up.

Excited Arrested Development GIF



Excited Coffee GIF
I will let you read on and not spoil the big surprise. 🤣
 
  • Haha
Reactions: 4 users

CHIPS

Regular
  • Haha
  • Like
Reactions: 5 users
Hi Hop ,

Agree with what you have said.

Early voting can give the board a feel from the investors perspective , and hence withdraw or amend the points to be voted on before the AGM , if they so choose.

Additionally shareholders can amend their early vote , online , if something should happen prior to the AGM .

Everyone has different time horizons , and i feel for those that have been waiting , years , with baited breath .


Regards,
Esq.
 
  • Haha
Reactions: 4 users
Hi Hop ,

Agree with what you have said.

Early voting can give the board a feel from the investors perspective , and hence withdraw or amend the points to be voted on before the AGM , if they so choose.

Additionally shareholders can amend their early vote , online , if something should happen prior to the AGM .

Everyone has different time horizons , and i feel for those that have been waiting , years , with baited breath .


Regards,
Esq.
I’m going to hold off for now voting, just incase the company can change my mind beforehand.

1713864237941.gif
 
  • Haha
  • Like
Reactions: 9 users

MDhere

Regular
  • Haha
Reactions: 3 users

Learning

Learning to the Top 🕵‍♂️
@Diogenese,
This recent patents from Syntiant using co neuromorphic processor, is it anything of interest here?


Abstract

Provided herein is an integrated circuit including, in some embodiments, a special-purpose host processor, a neuromorphic co-processor, and a communications interface between the host processor and the co-processor configured to transmit information therebetween. The special-purpose host processor can be operable as a stand-alone processor. The neuromorphic co-processor may include an artificial neural network. The co-processor is configured to enhance special-purpose processing of the host processor through an artificial neural network. In such embodiments, the host processor is a pattern identifier processor configured to transmit one or more detected patterns to the co-processor over a communications interface. The co-processor is configured to transmit the recognized patterns to the host processor.


And the new release NDP250.


Moving Vision AI from the Cloud to the Device

The NDP250 has advanced image capabilities and is ideal for ultra-low power video applications for automotive security, battery-powered cameras and video doorbells. Running powerful always-on image recognition at under 30mW has several advantages, including:

A significant reduction in power consumption by processing data locally on devices, thereby extending battery life and enabling efficient resource utilization.

Lower latency since data doesn't need to travel back and forth to a remote server, resulting in faster response times crucial for real-time applications and greatly increases customer satisfaction.

Enhanced privacy by processing sensitive data locally, minimizing the need to transmit information over networks where it could be vulnerable to breaches or interception.

A notable reduction in cloud costs, sometimes as high as 90%, since less data needs to be transferred and processed in the cloud, leading to lower infrastructure expenses for businesses deploying edge AI solutions.

Equipped with an Arm Cortex M0 processor and a HiFi 3 DSP to support feature extraction and signal processing for image and voice enhancements, the NDP250’s integrated power management unit allows single power rail operation, where the integrated phase-locked Loop (PLL) provides further system cost and size optimization.

With the ability to process multiple heterogenous networks concurrently, the NDP250 also supports convolution neural networks including 1D, 2D and depth-wise, fully connected networks, and recurrent neural networks including LSTM (long short-term memory) and GRU (gated recurrent unit).

Other key features include:

Syntiant Core 3 neural network

Supports more than 6M neural parameters (in 8-bit mode)

Hardware acceleration over 30 GOPS

Dual 11-wire direct image interface

Dual PDM digital microphone interface

I2S serial interface with PCM

Quad-SPI and dual-I2C controller and target for multi-modal sensor fusion

Up to 120MHz internal operating frequency

Low power PLL for flexible clock input

Software Development Kit

Training Development Kit

120-ball 6.1mm x 5.1mm eWLB package (0.5mm pitch)

The NDP250 is sampling now.

Syntiant will be demonstrating the NDP250 and various hardware-agnostic deep learning vison models at Embedded World 2024 (Hall 2 - Booth 2-338), April 9-11 in Nuremberg, Germany. Contact info@syntiant.com to arrange a meeting or demo at Embedded World.


Anything to see here?

Learning 🪴
 
  • Like
  • Wow
  • Thinking
Reactions: 10 users

Guzzi62

Regular
@Diogenese,
This recent patents from Syntiant using co neuromorphic processor, is it anything of interest here?


Abstract

Provided herein is an integrated circuit including, in some embodiments, a special-purpose host processor, a neuromorphic co-processor, and a communications interface between the host processor and the co-processor configured to transmit information therebetween. The special-purpose host processor can be operable as a stand-alone processor. The neuromorphic co-processor may include an artificial neural network. The co-processor is configured to enhance special-purpose processing of the host processor through an artificial neural network. In such embodiments, the host processor is a pattern identifier processor configured to transmit one or more detected patterns to the co-processor over a communications interface. The co-processor is configured to transmit the recognized patterns to the host processor.


And the new release NDP250.


Moving Vision AI from the Cloud to the Device

The NDP250 has advanced image capabilities and is ideal for ultra-low power video applications for automotive security, battery-powered cameras and video doorbells. Running powerful always-on image recognition at under 30mW has several advantages, including:

A significant reduction in power consumption by processing data locally on devices, thereby extending battery life and enabling efficient resource utilization.

Lower latency since data doesn't need to travel back and forth to a remote server, resulting in faster response times crucial for real-time applications and greatly increases customer satisfaction.

Enhanced privacy by processing sensitive data locally, minimizing the need to transmit information over networks where it could be vulnerable to breaches or interception.

A notable reduction in cloud costs, sometimes as high as 90%, since less data needs to be transferred and processed in the cloud, leading to lower infrastructure expenses for businesses deploying edge AI solutions.

Equipped with an Arm Cortex M0 processor and a HiFi 3 DSP to support feature extraction and signal processing for image and voice enhancements, the NDP250’s integrated power management unit allows single power rail operation, where the integrated phase-locked Loop (PLL) provides further system cost and size optimization.

With the ability to process multiple heterogenous networks concurrently, the NDP250 also supports convolution neural networks including 1D, 2D and depth-wise, fully connected networks, and recurrent neural networks including LSTM (long short-term memory) and GRU (gated recurrent unit).

Other key features include:

Syntiant Core 3 neural network

Supports more than 6M neural parameters (in 8-bit mode)

Hardware acceleration over 30 GOPS

Dual 11-wire direct image interface

Dual PDM digital microphone interface

I2S serial interface with PCM

Quad-SPI and dual-I2C controller and target for multi-modal sensor fusion

Up to 120MHz internal operating frequency

Low power PLL for flexible clock input

Software Development Kit

Training Development Kit

120-ball 6.1mm x 5.1mm eWLB package (0.5mm pitch)

The NDP250 is sampling now.

Syntiant will be demonstrating the NDP250 and various hardware-agnostic deep learning vison models at Embedded World 2024 (Hall 2 - Booth 2-338), April 9-11 in Nuremberg, Germany. Contact info@syntiant.com to arrange a meeting or demo at Embedded World.


Anything to see here?

Learning 🪴
They surely seems to be a competitor (unless they are using Alkida).

Quote:

A New Kind of Processor for Deep Learning​

Our Neural Decision Processors™ enable customers to quickly and easily deploy deep-learning models on power-constrained devices that previously ran on cloud servers.
| Provide 100x the efficiency and 10/30x higher throughput when compared to existing low-power MCUs.
| At-memory compute greatly reduces power consumption and latency by eliminating any unnecessary data movement.
| Specially designed to run deep-learning models, directly processing neural network layers achieving unprecedented levels of efficiency, often greater than 80%.
| Multi-generational product offerings provide scalability for any edge workload.


They seems to be selling a lot we are involved in, which is quite worrying.

They already have mass production of different sensors as per link below, and no, I don't think you have to pay 1 Mill$ to get hold of them.


They also sells evaluation kits from Renesas and others for under 100$ a pop.


I really hope we are involved here.

Partners:

 
  • Thinking
  • Wow
  • Like
Reactions: 9 users

rgupta

Regular
They surely seems to be a competitor (unless they are using Alkida).

Quote:

A New Kind of Processor for Deep Learning​

Our Neural Decision Processors™ enable customers to quickly and easily deploy deep-learning models on power-constrained devices that previously ran on cloud servers.
| Provide 100x the efficiency and 10/30x higher throughput when compared to existing low-power MCUs.
| At-memory compute greatly reduces power consumption and latency by eliminating any unnecessary data movement.
| Specially designed to run deep-learning models, directly processing neural network layers achieving unprecedented levels of efficiency, often greater than 80%.
| Multi-generational product offerings provide scalability for any edge workload.


They seems to be selling a lot we are involved in, which is quite worrying.

They already have mass production of different sensors as per link below, and no, I don't think you have to pay 1 Mill$ to get hold of them.


They also sells evaluation kits from Renesas and others for under 100$ a pop.


I really hope we are involved here.

Partners:

Looks a good competitor look at their partner lists
Edge impulse, audrino, Bosch, renasas etc.
So let us wait and watch.
 
  • Like
  • Wow
Reactions: 3 users

Diogenese

Top 20
Geoffrey Hinton on how GenAI ate the world:


@Diogenese,
This recent patents from Syntiant using co neuromorphic processor, is it anything of interest here?


Abstract

Provided herein is an integrated circuit including, in some embodiments, a special-purpose host processor, a neuromorphic co-processor, and a communications interface between the host processor and the co-processor configured to transmit information therebetween. The special-purpose host processor can be operable as a stand-alone processor. The neuromorphic co-processor may include an artificial neural network. The co-processor is configured to enhance special-purpose processing of the host processor through an artificial neural network. In such embodiments, the host processor is a pattern identifier processor configured to transmit one or more detected patterns to the co-processor over a communications interface. The co-processor is configured to transmit the recognized patterns to the host processor.


And the new release NDP250.


Moving Vision AI from the Cloud to the Device

The NDP250 has advanced image capabilities and is ideal for ultra-low power video applications for automotive security, battery-powered cameras and video doorbells. Running powerful always-on image recognition at under 30mW has several advantages, including:

A significant reduction in power consumption by processing data locally on devices, thereby extending battery life and enabling efficient resource utilization.

Lower latency since data doesn't need to travel back and forth to a remote server, resulting in faster response times crucial for real-time applications and greatly increases customer satisfaction.

Enhanced privacy by processing sensitive data locally, minimizing the need to transmit information over networks where it could be vulnerable to breaches or interception.

A notable reduction in cloud costs, sometimes as high as 90%, since less data needs to be transferred and processed in the cloud, leading to lower infrastructure expenses for businesses deploying edge AI solutions.

Equipped with an Arm Cortex M0 processor and a HiFi 3 DSP to support feature extraction and signal processing for image and voice enhancements, the NDP250’s integrated power management unit allows single power rail operation, where the integrated phase-locked Loop (PLL) provides further system cost and size optimization.

With the ability to process multiple heterogenous networks concurrently, the NDP250 also supports convolution neural networks including 1D, 2D and depth-wise, fully connected networks, and recurrent neural networks including LSTM (long short-term memory) and GRU (gated recurrent unit).

Other key features include:

Syntiant Core 3 neural network

Supports more than 6M neural parameters (in 8-bit mode)

Hardware acceleration over 30 GOPS

Dual 11-wire direct image interface

Dual PDM digital microphone interface

I2S serial interface with PCM

Quad-SPI and dual-I2C controller and target for multi-modal sensor fusion

Up to 120MHz internal operating frequency

Low power PLL for flexible clock input

Software Development Kit

Training Development Kit

120-ball 6.1mm x 5.1mm eWLB package (0.5mm pitch)

The NDP250 is sampling now.

Syntiant will be demonstrating the NDP250 and various hardware-agnostic deep learning vison models at Embedded World 2024 (Hall 2 - Booth 2-338), April 9-11 in Nuremberg, Germany. Contact info@syntiant.com to arrange a meeting or demo at Embedded World.


Anything to see here?

Learning 🪴
Syntiant are very analoggy:

1713869666912.png


  • [0029] FIG. 2 provides a schematic illustrating an exemplary embodiment of an analog multiplier array in accordance with some embodiments;
  • [0030] FIG. 3 provides a schematic illustrating an exemplary embodiment of an analog multiplier array in accordance with some embodiments;

0051] Referring now to FIG. 2 , a schematic illustrating an analog multiplier array 200 is provided in accordance with some embodiments. Such an analog multiplier array can be based on a digital NOR flash array in that a core of the analog multiplier array can be similar to a core of the digital NOR flash array or the same as a core of the digital NOR flash array. That said, at least select and read-out circuitry of the analog multiplier array are different than a digital NOR array. For example, output current is routed as an analog signal to a next layer rather than over bit lines going to a sense-amp/comparator to be converted to a bit. Word-line analogs are driven by analog input signals rather than a digital address decoder. Furthermore, the analog multiplier array 200 can be used in neuromorphic ICs such as the neuromorphic IC 102. For example, a neural network can be disposed in the analog multiplier array 200 in a memory sector of a neuromorphic IC.
 
  • Like
  • Love
  • Wow
Reactions: 18 users

Learning

Learning to the Top 🕵‍♂️
Geoffrey Hinton on how GenAI ate the world:



Syntiant are very analoggy:

View attachment 61391

  • [0029] FIG. 2 provides a schematic illustrating an exemplary embodiment of an analog multiplier array in accordance with some embodiments;
  • [0030] FIG. 3 provides a schematic illustrating an exemplary embodiment of an analog multiplier array in accordance with some embodiments;

0051] Referring now to FIG. 2 , a schematic illustrating an analog multiplier array 200 is provided in accordance with some embodiments. Such an analog multiplier array can be based on a digital NOR flash array in that a core of the analog multiplier array can be similar to a core of the digital NOR flash array or the same as a core of the digital NOR flash array. That said, at least select and read-out circuitry of the analog multiplier array are different than a digital NOR array. For example, output current is routed as an analog signal to a next layer rather than over bit lines going to a sense-amp/comparator to be converted to a bit. Word-line analogs are driven by analog input signals rather than a digital address decoder. Furthermore, the analog multiplier array 200 can be used in neuromorphic ICs such as the neuromorphic IC 102. For example, a neural network can be disposed in the analog multiplier array 200 in a memory sector of a neuromorphic IC.

Thanks Dio.

Learning 🪴
 
  • Like
Reactions: 7 users

Guzzi62

Regular
Geoffrey Hinton on how GenAI ate the world:



Syntiant are very analoggy:

View attachment 61391

  • [0029] FIG. 2 provides a schematic illustrating an exemplary embodiment of an analog multiplier array in accordance with some embodiments;
  • [0030] FIG. 3 provides a schematic illustrating an exemplary embodiment of an analog multiplier array in accordance with some embodiments;

0051] Referring now to FIG. 2 , a schematic illustrating an analog multiplier array 200 is provided in accordance with some embodiments. Such an analog multiplier array can be based on a digital NOR flash array in that a core of the analog multiplier array can be similar to a core of the digital NOR flash array or the same as a core of the digital NOR flash array. That said, at least select and read-out circuitry of the analog multiplier array are different than a digital NOR array. For example, output current is routed as an analog signal to a next layer rather than over bit lines going to a sense-amp/comparator to be converted to a bit. Word-line analogs are driven by analog input signals rather than a digital address decoder. Furthermore, the analog multiplier array 200 can be used in neuromorphic ICs such as the neuromorphic IC 102. For example, a neural network can be disposed in the analog multiplier array 200 in a memory sector of a neuromorphic IC.

Sadly most seems to be fine using Analog.

If it's cheap/easy to implement and low on power, why using anything else?

If we are not getting any IP deals this year, it's game over I believe and Brainchip will remain a small company forever until someone buys it cheap.

Sorry if I sound negative but Sean soon needs to pull a rabbit or two up of the hat or his strategy is a failure.

I think he is aware of that and I am for that he stays for now but if no IP deals this year, he will have to go IMHO
 
  • Thinking
  • Like
  • Sad
Reactions: 11 users

wilzy123

Founding Member
  • Haha
  • Like
  • Fire
Reactions: 5 users

Diogenese

Top 20
Sadly most seems to be fine using Analog.

If it's cheap/easy to implement and low on power, why using anything else?

If we are not getting any IP deals this year, it's game over I believe and Brainchip will remain a small company forever until someone buys it cheap.

Sorry if I sound negative but Sean soon needs to pull a rabbit or two up of the hat or his strategy is a failure.

I think he is aware of that and I am for that he stays for now but if no IP deals this year, he will have to go IMHO
It's true that we have not seen any published details of engagements, but it may be that there are some behind the NDA wall.

The Syntiant market is PCB assemblers. There are many applications where analog will provide satisfactory performance. As we do not have a physical product in the market, we are not in that race.

Our strategy is a tougher nut to crack. We are targeting higher up the food chain - IC manufacturers - much longer lead times and higher barrier to entry. We keep hearing promising noises - they love us at the trade shows, but the lack of publicly disclosed engagements can be disheartening for us share holders.

That said, Akida is being used in the real world, albeit in niche applications.

Let's hope our end-user products, the Edge Box/server can significantly reduce our net cask outgoings soon. We should have news on the sales of these products in the near future.
 
  • Like
  • Love
  • Fire
Reactions: 36 users
Someone who's not behind a NDA appears to be ticking along and making some inroads with their development.



Tata Elxsi


MULTIMODAL AI AND NEUROMORPHIC AI: DETECTION, DIAGNOSIS, PROGNOSIS

Navigate​


The synergy of cutting-edge technologies like Multimodal and Neuromorphic AI signals a pivotal shift from reactive to proactive healthcare. This article explores captivating use cases, offering insights on the implementation of Responsible AI.
Join us as we navigate the frontier of healthcare, where the synergy of innovation and responsibility promises a revolution in patient care and well-being.

Current state of AI adoption in Healthcare​

Unlocking the full potential of AI in healthcare is an uncharted journey. From optimising drug combinations to spearheading clinical correlation and toxicity studies, AI is set to redefine every facet of the industry. Despite its transformative capabilities, AI remains a niche technology, requiring a nuanced understanding of its application in healthcare.
The tides are changing as the healthcare sector recognises the urgency for an interdisciplinary approach, marrying engineering with medical science. This paradigm shift signals an imminent era where AI’s vast capabilities will revolutionise diagnostics, patient treatment, protocols, drug development and delivery, and prescription practices over the next decade.
Join us as we navigate the frontier of healthcare, where the synergy of innovation and responsibility promises a revolution in patient care and well-being.

Multimodal AI and Neuromorphic Technology – A new era in Preventive Healthcare​

In the ever-evolving landscape of healthcare, the amalgamation of Multimodal AI and Neuromorphic Technology marks a pivotal moment—a shift from reactive medicine to a proactive, preventive healthcare paradigm. This synergy is not just a collaboration of cutting-edge technologies; it’s a gateway to a future where wellness takes centre stage.
Multimodal AI and Neuromorphic Technology A new era in Preventive Healthcare

These technologies hold promise to transform healthcare by enhancing diagnostics, enabling personalised medicine, predicting long-term prognosis and contributing to innovations in therapeutic interventions.
Let’s delve into compelling use cases and glimpse the future of preventive healthcare.
The tides are changing as the healthcare sector recognises the urgency for an interdisciplinary approach, marrying engineering with medical science. This paradigm shift signals an imminent era where AI’s vast capabilities will revolutionise diagnostics, patient treatment, protocols, drug development and delivery, and prescription practices over the next decade.
Join us as we navigate the frontier of healthcare, where the synergy of innovation and responsibility promises a revolution in patient care and well-being.

Defining Multimodal and Neuromorphic AI​

Multi-modal AI​

Multimodal AI refers to the artificial intelligence systems that process and analyse data from multiple modalities or sources. In healthcare, these modalities often include both visual and clinical data. Visual data may include medical images from scans, while clinical data encompasses patient records, parameters, and test reports. Multimodal AI integrates these diverse data types to provide a comprehensive understanding, draw meaningful insights and give suggestions based on data and image analytics.

Neuromorphic Technology​

The term “neuromorphic” comes from the combination of “neuro” (related to the nervous system) and “morphic” (related to form or structure). Neuromorphic technology is an innovative approach in computing that draws inspiration from the structure and function of a human brain. These are AI powered by brain-like computing architectures. It can help process larger amount of data with less computing power, memory and electric power consumption. Neuromorphic Technology utilises Artificial Neural Networks (ANN) and Spiking Neural Networks (SNN) to mimic the parallel processing, event-driven nature, and adaptability observed in biological brains.

Defining Multimodal and Neuromorphic AI

Multimodal Inputs​

  • Medical Images
  • Lab Reports
  • Clinical History
  • Patient Demographic Information

Fusion Module​

Cross-modal attention mechanism to dynamically weigh the importance of text, video, and other parameters for calculating index to decide priority of selection.

Inference Outputs​

Inference Results for
  • Diagnostic
  • Prognostic
  • Lifestyle Recommendation
  • Disease Prediction

Use Cases of Multimodal and Neuromorphic AI​

Early Screening & Disease Detection​

Multimodal AI​

  • Integrates visual and clinical data for holistic analysis.
  • Advanced –image recognition for early detection.
  • Comprehensive patient profiling.

Neuromorphic Technology​

  • Efficient pattern recognition for subtle disease indicators.
  • Event-driven processing for real-time detection. This is crucial for detecting anomalies or irregularities that may be early signs of diseases.
  • Continuous monitoring for dynamic changes. This continuous surveillance is especially valuable for conditions with varying symptoms.

Diagnosis​

Multimodal AI​

  • Integrated diagnostic insights from diverse data.
  • Cross-verification for reliability.
  • Tailored treatment plans based on nuanced understanding.
  • Continuous updates based on latest findings reported in the subject.

Neuromorphic Technology​

  • Large and Efficient data processing with minimal energy consumption. This efficiency contributes to faster and more accurate diagnoses.
  • Allows integration of more complex algorithms on wearable devices; makes diagnostics more real-time and helps timely interventions
  • Implantable devices can be made AI enabled with Neuromorphic computing, due to the low computing requirement and power consumption, making the diagnosis and management more precise and real-time.
  • Adaptive intelligence for dynamic adjustments. This adaptability enhances the precision of diagnostic processes. This event-driven processing aligns with the dynamic nature of healthcare data allowing for more accurate and timely diagnoses.
  • SNN for real-time response
    and accuracy.

Prognosis​

Multimodal AI​

  • Research Advancements: Facilitates discovery of new insights, contributing to medical advancements and innovations.
  • Personalised Prognostic Models: Considering both visual and clinical data, these models account for individual variations, and correlate with prior case records and provide more accurate predictions of disease outcomes.
  • Dynamic Adaptability: The adaptability of multimodal AI to changing data patterns ensures that prognostic models can dynamically adjust based on evolving patient conditions and improve prognosis predictions

Neuromorphic Technology​

  • Analysis of longitudinal data for predicting disease progression.
  • Dynamic adaptability in prognostic models that can adjust to changing data patterns. This adaptability improves prognosis prediction accuracy for evolving patient conditions.
  • Personalised prognostic insights based on individual variations can help in more accurate predictions tailored to individual patient profiles.

Tata Elxsi Use Case

Disease Detection and Diagnosis

Utilising Neuromorphic Technology, we’ve achieved significant advancements in the analysis of medical images on low-computing embedded platforms, enabling on-field diagnostics of ultrasound images.
This innovative approach provides critical diagnostic information for musculoskeletal injuries, including tissue damage extent, recovery progress, and healing time predictions, all with enhanced efficiency and device portability making it ideal for applications such as sports medicine.

Applications of Multimodal and Neuromorphic AI​

Multimodal AI​

  • Comprehensive Patient Analysis
  • Diagnostic Accuracy
  • Mental Health and Behavioural Analysis
  • Lifestyle Reviews and Recommendations
  • Management of Chronic Diseases like Diabetes/HT/Cardiac Diseases with continous monitoring and personalised medications
  • Diagnosis/Management and Prognosis of various types of cancers, Digital Drug Trials, Effective Pandemic Surveillance and staging, Gene Therapy and Genomics
  • Recommendations for interventions and Prioritisation of therapeutic resources and modalities

Neuromorphic Technology​

  • Implants, Wearables Devices
  • Processing Large Data
  • Medical Imaging Analysis
  • Drug Discovery and Personalised Medicine
  • Robotic Surgery Assistance
  • Neurological Disorder Understanding
  • Patient Care and Rehabilitation
  • Predictive Analytics for Healthcare Management
  • Energy Efficient Remote Monitoring

The Synergy of MLOps and Advanced AI​

The transformative impact of MLOps across operational efficiency, data management, patient outcomes, and the overall quality of care is unmistakable. In the quest for advancing healthcare, the convergence of Machine Learning Operations (MLOps) with Multimodal and Neuromorphic AI has emerged as a game-changer. These technologies can help in seamless deployment, continuous monitoring, and collaborative development across various stakeholders in the healthcare ecosystem.
While Advanced AI Technologies offer the potential for improvising the use cases, the application of MLOps can be instrumental in strengthening and regulating these advancements. It achieves this by bringing in streamlined AI development processes, dataset management, continuous monitoring of model accuracy across different versions, ensuring the deployment of thoroughly vetted versions for clinical use. Additonally, MLOps frameworks enable and learn from deviations, further enhancing their efficacy in healthcare applications.

Use Cases​


Disease Detection​

Disease Prediction and Prevention
Real World Application – Early Detection of Chronic Disease, Infectious Disease Monitoring
Healthcare Fraud Detection
Real World Application – Claims Analysis, Identify Theft Prevention
Medical Imaging Analysis
Real World Application – Early Cancer Detection, Neurological Disorders Diagnosis
Genomic Research
Real World Application – Cancer Genomics, Rare Genetic Diseases

Diagnosis​

Drug Discovery and Development
Real World Application – Protein Folding Prediction, Drug Toxicity Prediction
Personalised Medicine
Real World Application – Oncology and Targeted Therapies, Chronic Disease Management
Healthcare Resource Management
Real World Application – Emergency Room Digitalised Management, Pharmaceutical Supply Chain Management

Prognosis​

Prediction of Clinical Outcome
Real World Application – Prediction of recovery time and quality of life, Adverse Effects, Short term and long term impact
Remote Patient Monitoring
Real World Application – Chronic Disease Management, Post Surgery Monitoring

Responsible AI – Navigating Ethical Frontiers​

In the realm of AI, addressing bias stands as a pivotal ethical imperative, particularly in fields like medical analysis where the demand for precision is ethically, legally, and morally paramount. As AI practitioners, our commitment to responsible AI requires rigorous testing using diverse, unbiased anonymised datasets, continual monitoring to mitigate biases, and a steadfast dedication to achieving fair outcomes across diverse patient populations.
Ethical AI

Moreover, the ethical considerations extend to the strategic utilisation of data. The foundation of responsible AI in healthcare is laid upon a robust ethical framework that guides the entire lifecycle of Neuromorphic and Multimodal AI systems. Stakeholders must unwaveringly adhere to established ethical principles, ensuring transparency, fairness, and accountability at every stage of AI implementation.
When we delve into the realm of Gen AI, the potential for malpractice looms. Consider scenarios where a patient, with normal renal function manageable through medication, undergoes a renal scan. Unscrupulous use of Gen AI could manipulate images, creating false lesions, leading to unnecessary surgeries or even nephrectomy which can benefit illegal organ trade
Thus, the imperative lies in defining strong ethical boundaries, implementing robust audits, and establishing legal frameworks to prevent data manipulation and ensure the highest standards of integrity.
In embracing these technological advancements responsibly, we are not just witnessing the future of healthcare; we are actively shaping it. The era of proactive, preventive healthcare beckons, promising a future where wellness is at the forefront of the industry’s evolution.
The shift from traditional, one-size-fits-all medical practices, prone to misinterpretation and diagnostic errors, to AI-enhanced methodologies, heralds a new era of precision and personalised care. AI’s capability to analyse a broad spectrum of patient data—ranging from genetic backgrounds to lifestyle factors—promises a departure from misdiagnoses and introduces tailored therapeutic interventions.
AI and multimodal technologies enable a holistic view of the patient’s health, integrating diverse data points. While, Neuromorphic computing advances the portability of medical devices, including wearables and implants, transforming them into intelligent systems capable of adapting to varying conditions.
As thought leaders in the healthcare industry, our commitment to responsibly integrate these technologies paves the way for a future where healthcare is not only reactive, but anticipatory, personalised, and universally accessible.

Author
Anup SS, Practice Head, AI and ML, Tata Elxsi
Anup S S
Practice Head, Artificial Intelligence, Tata Elxsi
Anup S.S. is a visionary in leveraging Artificial Intelligence, Machine Learning and Deep Learning. Leading breakthrough AI projects in healthcare, Anup’s strategic insight and innovation ignite client success, unlocking AI’s full potential.
 
Last edited:
  • Like
  • Love
  • Fire
Reactions: 66 users
Top Bottom