Potential Applications for Akida?

[clip]

AKIDA in angle grinders as a safety feature?

So with AKIDAs capability of vibration analysis, I was wondering if it could detect damage or improper use of angle grinders by its sound/vibration patterns.
Lets say the cutting disk is damaged and about to burst. Or someone is using the cutting disk for the grinding job or is trying to cut wood with it.
This will produce different sound patterns.

I skipped through this video



and was reminded, that there are many ways of inadvertently wrong usage.

Maybe AKIDA can shut the angle grinder off, when it detects misuse or damage.
Could be a helpful safety feature.

 

[Kachoo]

AKIDA in angle grinders as a safety feature?

So with AKIDAs capability of vibration analysis, I was wondering if it could detect damage or improper use of angle grinders by its sound/vibration patterns.
Lets say the cutting disk is damaged and about to burst. Or someone is using the cutting disk for the grinding job or is trying to cut wood with it.
This will produce different sound patterns.

I skipped through this video



and was reminded, that there are many ways of inadvertently wrong usage.

Maybe AKIDA can shut the angle grinder off, when it detects misuse or damage.
Could be a helpful safety feature.

I could have used Akida to direct my cuts with the jig saw so it be straighter lol

 

[clip]

Try this for your next project



Felder PCS

 

[stuart888]

Try this for your next project



Felder PCS

Wow. Very interesting. It makes a lot of safety sense. Great find.

Must be tons of industrial safety use cases for the Brainchip Sensor Smarts!

 

[clip]

Yes, hopefully.
Think about the car industry and the established safety standards (seatbelts, airbags, ESC ... ).
Same with industrial machines and power tools.
More and more safety features will be developed and will become mandatory.

You can have a look at Bosch power tool Sensor Technology:

Bosch power tools for trade and industry | Bosch Professional

Bosch Power Tools – Trade/Industry. The professional blue power tools from Bosch are engineered for excellence – meeting the highest standards in speed, precision and robustness for great, professional work results.

www.bosch-professional.com

They already have some products out there with multiple sensors for user safety.

But I have no idea what AKIDA could do with these sensors at this stage.

 

[clip]

Wow. Very interesting. It makes a lot of safety sense. Great find.

Must be tons of industrial safety use cases for the Brainchip Sensor Smarts!

View attachment 15585

Here is a video that explains how the technology works:




In addition, the technology is already a few years old, first invented by SawStop in 2000:
https://en.wikipedia.org/wiki/SawStop




And here is an in-build camera, vision-based solution by the company Altendorf:



Thats where AKIDA could possibly come into play.
This solution has two advantages: it works with gloves on your hands and it triggers the safety mechanism earlier. Your hand dont have to be a few millimeters in front of the sawblade to trigger.

This vision-based safety solution can surely be adopted to other types of machines like lathes, drill presses, bandsaws etc.

 

[clip]

Built Robotics



Their invention 'Exosystem' can be deployed on various, existing late-model excavators (Hitachi, Deere, CAT, Volvo, etc).

https://www.builtrobotics.com/

 

[clip]

Small update on the GoPro:

GoPro Hero11 Black: Will own a new Sensor & an old Design - SaveDelete

According to images & specs first emerged, the GoPro Hero11 Black will look virtually identical to the Hero10, having all the essential tweaks

savedelete.com

and

GoPro Hero 11 Black Mini to arrive as a smaller, likely cheaper alternative to the Hero 11 Black

While the GoPro Hero 11 Black looks set to be an incremental update on last year's Hero 10 Black, GoPro has now been touted to bring a new action camera to the market later this month as well. This camera, the Hero 11 Black Mini, appears to be a smaller, cheaper sibling to the Hero 11 Black.

www.notebookcheck.net

So with the GoPro mini without a touchscreen and only two(?) buttons, the voice-control feature will become even more practically.
And it is claimed, that the GoPro 11 will have smile and blink detection.
So it is imaginable, that both cameras could get a face or gesture control feature.

The GoPro Bones, which is similar to the GoPro mini, has QR code control. You can switch settings by scanning QR codes. Will be even more practical, when you can do that with hand gesture.

AKIDA could also be used for noise suppression, minimizing wind noises. The GoPro cameras have due to their compact size and waterproofness only modest audio quality.
It would also keep them ahead of competitor DJI.
Here is a demonstration of its drone feature active track:

 

[MADX]

I can definitely see Robotic Camera Operators popping up some day. Imagine having a robot on wheels following a footy game around the boundry or drones flying above following the play. Some cool stuff to look forward to!

How about Akidas recognising a brawl breaking out, or a streaker running on the field, and automatically switching the feed for viewers benefit

 

[MADX]

Some time ago I asked "Is it the case that where algorithms are using the sensor data to cause an action, Akida tech could cause more accurate actions because an algo. is limited to what the programmer has built into it and, for instance, could not cope with the unexpected?"
I did not get an answer & I now think that I did not ask clearly, so here goes......

You have a device to perform certain actions. The actions are triggered by sensors. An algorithm dictates what action is performed.
The algo. is limited to what the programmer has built into it so an unforeseen inappropriate action may be performed. Can Akida tech detect the inappropriate action and prevent future wrong actions?

I have a strong feeling that I'm showing ignorance by asking this, but too bad.

 

[buena suerte :-)]

Some time ago I asked "Is it the case that where algorithms are using the sensor data to cause an action, Akida tech could cause more accurate actions because an algo. is limited to what the programmer has built into it and, for instance, could not cope with the unexpected?"
I did not get an answer & I now think that I did not ask clearly, so here goes......

You have a device to perform certain actions. The actions are triggered by sensors. An algorithm dictates what action is performed.
The algo. is limited to what the programmer has built into it so an unforeseen inappropriate action may be performed. Can Akida tech detect the inappropriate action and prevent future wrong actions?

I have a strong feeling that I'm showing ignorance by asking this, but too bad.

Maybe our resident retired engineer @Diogenese could assist?

 

[Diogenese]

Some time ago I asked "Is it the case that where algorithms are using the sensor data to cause an action, Akida tech could cause more accurate actions because an algo. is limited to what the programmer has built into it and, for instance, could not cope with the unexpected?"
I did not get an answer & I now think that I did not ask clearly, so here goes......

You have a device to perform certain actions. The actions are triggered by sensors. An algorithm dictates what action is performed.
The algo. is limited to what the programmer has built into it so an unforeseen inappropriate action may be performed. Can Akida tech detect the inappropriate action and prevent future wrong actions?

I have a strong feeling that I'm showing ignorance by asking this, but too bad.

Hi MADX,

Akida 1 does not directly cause an action. It's primary job is to identify/classify the sensor/video signals, eg, by identifying the objects detected by a LiDaR system in a car. The LiDaR system can determine the distance to the objects. There will be some form of processor which responds to Akida's output to cause the action, and that is where your issue may arise. Of course, the other possibility is that Akida mis-classifies the sensor data - in some tests Akida has an accuracy of about 96% which is very good, but not perfect. That is why you need at least 2 different types of sensors, and cross-your fingers that the 4% errors do not coincide.

Akida 2 will be able to classify sensor data and determine additional characteristics such as direction of motion of detected objects.

 

[buena suerte :-)]

Hi MADX,

Akida 1 does not directly cause an action. It's primary job is to identify/classify the sensor/video signals, eg, by identifying the objects detected by a LiDaR system in a car. The LiDaR system can determine the distance to the objects. There will be some form of processor which responds to Akida's output to cause the action, and that is where your issue may arise. Of course, the other possibility is that Akida mis-classifies the sensor data - in some tests Akida has an accuracy of about 96% which is very good, but not perfect. That is why you need at least 2 different types of sensors, and cross-your fingers that the 4% errors do not coincide.

Akida 2 will be able to classify sensor data and determine additional characteristics such as direction of motion of detected objects.

Awesome... thanks Dodgy knees

 

[MADX]

Hi MADX,

Akida 1 does not directly cause an action. It's primary job is to identify/classify the sensor/video signals, eg, by identifying the objects detected by a LiDaR system in a car. The LiDaR system can determine the distance to the objects. There will be some form of processor which responds to Akida's output to cause the action, and that is where your issue may arise. Of course, the other possibility is that Akida mis-classifies the sensor data - in some tests Akida has an accuracy of about 96% which is very good, but not perfect. That is why you need at least 2 different types of sensors, and cross-your fingers that the 4% errors do not coincide.

Akida 2 will be able to classify sensor data and determine additional characteristics such as direction of motion of detected objects.

Thanks Dio. So the arrangement is sensor(s) > sensor signals > Akida classifies the sensor signals > a signal judged to be appropriate by Akida is sent to a processor > the processor is triggered to use measured data (e.g. distance) to make calculation(s) to produce e.g. a certain voltage > the voltage causes a device to do something?
Can the Akida check the accuracy of the final action and learn from it for improvement?

I'm hoping the penny will eventually drop to clear my muddled thinking.

 

[Diogenese]

Thanks Dio. So the arrangement is sensor(s) > sensor signals > Akida classifies the sensor signals > a signal judged to be appropriate by Akida is sent to a processor > the processor is triggered to use measured data (e.g. distance) to make calculation(s) to produce e.g. a certain voltage > the voltage causes a device to do something?
Can the Akida check the accuracy of the final action and learn from it for improvement?

I'm hoping the penny will eventually drop to clear my muddled thinking.

Now that's an interesting thought - feedback from the processor's response back to Akida.

Short answer is "no". The feedback would need to be to the processor's parameters which it used to make the change.

But the sensor would detect the change and this would be fed to Akida. So maybe the long answer is sort of "yes".

Akida can learn from the sensor data, so if a new shape is detected as a dog, it can be added to the Akida model library.

Also if data is collected in the cloud from all similar vehicles (for example) this can be used to update the Akida model libraries of all the
connected vehicles, so all can learn from one. I think this is called "federated learning".

 

[MADX]

Now that's an interesting thought - feedback from the processor's response back to Akida.

Short answer is "no". The feedback would need to be to the processor's parameters which it used to make the change.

But the sensor would detect the change and this would be fed to Akida. So maybe the long answer is sort of "yes".

Akida can learn from the sensor data, so if a new shape is detected as a dog, it can be added to the Akida model library.

Also if data is collected in the cloud from all similar vehicles (for example) this can be used to update the Akida model libraries of all the
connected vehicles, so all can learn from one. I think this is called "federated learning".

Thanks anew Dio. I (and I hope some others) are learning from you.

Thanks also for the diplomacy in your answers but, in my case, never be afraid to say outright if I'm stupid. Ditto anyone else on this forum who wants to join in.

 

[Diogenese]

Thanks anew Dio. I (and I hope some others) are learning from you.

Thanks also for the diplomacy in your answers but, in my case, never be afraid to say outright if I'm stupid. Ditto anyone else on this forum who wants to join in.

I live in a glass house ...

 

[MADX]

A device for detecting skin cancers with an Akida in a phone, an app, and using the camera.
Google Lens already does a good crude job. Add refinements from the app and you've got an hypochondriac delight.

 

[MADX]

I watched "Clarkson's Farm" (the ex Top Gear Clarkson) where he herded sheep with a drone that barked.
It sounded like an application for Akida, so I asked ChatGPT. Here's the answer

"

Using Akida chips in a drone for herding sheep is an interesting concept. While traditional drones often rely on computer vision algorithms and sensors for navigation and obstacle avoidance, integrating neuromorphic computing capabilities could offer some advantages. Here's how Akida chips might be applied in such a scenario:

1. Obstacle Avoidance and Navigation:
   
   • The Akida chip's event-driven processing could be beneficial for real-time obstacle avoidance. The drone could process visual data efficiently, making quick decisions based on changes in the environment. This is crucial when herding sheep in varied terrain where obstacles may be encountered.
2. Behavioral Adaptation:
   
   • The drone could be equipped with sensors to detect the behavior of the sheep. The Akida chip's ability to process spiking neural networks in real time might allow the drone to adapt its actions based on the behavior of the sheep, helping in more effective herding.
3. Energy Efficiency:
   
   • Akida's low-power design could be advantageous for drone applications where energy efficiency is crucial. Drones herding sheep may need to cover large areas, and a low-power processor can extend the drone's operational time without frequent recharging.
4. Learning and Adaptation:
   
   • The Akida chip's neuromorphic architecture is designed to support online learning. The drone could potentially learn and adapt its herding strategies over time based on the specific behaviors of the sheep and the terrain it encounters.

It's important to note that as of my last knowledge update in January 2022, specific implementations of Akida chips in drone technology for herding sheep might be more conceptual. Real-world deployment would require detailed engineering and testing to ensure the effectiveness, safety, and ethical considerations of using AI in herding practices. Additionally, advancements in technology and the availability of specific products may have occurred since then. Always check for the latest developments from relevant companies and research in the field."

I also asked it to suggest three devices for Akida and got generic answers e.g. autonomous cars etc.
I'm sure some more refined questions would come up with some specifics.
BTW, Google has just come out with their answer to ChatGPT. It is called Gemini and worth a look as it contains guidance for user questions relevant to the original question.
There may come a time when we can ask "Please design a device with Akida to....."

 

[wasMADX]

Post #72909 by skutza drew attention to what may be the single biggest reason why start-ups succeed i.e. TIMING of the launch when customers want a new product or service.
From that, in post #72915 HopalongPetrovski said (paraphrased)

" I think the timing is right for an Akida device for cybersecurity facilitated by some form of personal digital concierge.
Identity theft and the myriad ways it is conceived is a growing problem potentially affecting millions of us and is in desperate need of a viable solution.
Our unique Akida tech can be useful in this regard and a low-cost, easily implemented, reliable, and adaptable countermeasure incorporated at both the server level and where we reside, at the edge, may be a product that could gain wide exposure and acceptance."